diff --git a/README.md b/README.md new file mode 100644 index 0000000000000000000000000000000000000000..8e4cccba98c0d83c4d83c8f0042959ecb5b49f16 --- /dev/null +++ b/README.md @@ -0,0 +1,25 @@ +--- +license: apache-2.0 +pretty_name: NeMo Gym Conversational Tool-Use Assets +--- + +# NeMo Gym Conversational Tool-Use Assets + +This dataset repository stores the non-code assets used by NeMo Gym's conversational tool-use generation pipeline. +It is an asset bundle for Gym components, not a training or evaluation dataset. + +## Contents + +- `conversational_tool_use_policy_tool_generation/golden_policies`: eight policy/tool reference pairs required by + the `conversational_tool_use_policy_tool_generation` agent. +- `conversational_tool_use_policy_tool_generation/prompt_history`: historical policy/tool-generation prompt + revisions, retained for inspection but not required at runtime. +- `conversational_tool_use_domain_generation/prompt_history`: historical domain-generation prompt revisions, + retained for inspection but not required at runtime. + +The active prompt templates remain in the NeMo Gym source tree because they define current agent behavior. Run the +policy/tool-generation agent's preparation command to download and checksum-validate its runtime references: + +```bash +python -m responses_api_agents.conversational_tool_use.policy_tool_generation.prepare +``` diff --git a/conversational_tool_use_domain_generation/prompt_history/domains_sample_1.txt b/conversational_tool_use_domain_generation/prompt_history/domains_sample_1.txt new file mode 100644 index 0000000000000000000000000000000000000000..12e8907ff0956e037eb214f2e07e665d553bbdd8 --- /dev/null +++ b/conversational_tool_use_domain_generation/prompt_history/domains_sample_1.txt @@ -0,0 +1,19 @@ +Please give me list of all business/industry sectors that may benefit from having a customer-facing internet chatbot agent helper. +Imagine that the chatbot will have a set of executable tools that interface with a system to help the customer's needs. + + +Please format your response as a JSON such as: +```json +[ + { + "name": , + "applications": [ + { + "function": + }, + ... + ] + }, + ... +] +``` \ No newline at end of file diff --git a/conversational_tool_use_domain_generation/prompt_history/domains_sample_sectors.txt b/conversational_tool_use_domain_generation/prompt_history/domains_sample_sectors.txt new file mode 100644 index 0000000000000000000000000000000000000000..5cca43b1e4ce720d9245afb126d5f3db0f5fd9d5 --- /dev/null +++ b/conversational_tool_use_domain_generation/prompt_history/domains_sample_sectors.txt @@ -0,0 +1,7 @@ +Please give me list of all business/industry sectors. + + +Please format your response as a JSON such as: +```json +[sector_1, sector_2, sector_3, ...] +``` \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/policy-1.md b/conversational_tool_use_policy_tool_generation/golden_policies/policy-1.md new file mode 100644 index 0000000000000000000000000000000000000000..0b4202b242cfc095ff6b3c722e9bdf6be620db78 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/policy-1.md @@ -0,0 +1,84 @@ +**Ride-Sharing Agent Policy** +The current time is 2024-05-15 15:00:00 EST. + +As a ride-sharing agent, you can help users: +**Book/modify/cancel rides**, **resolve payment disputes**, **report incidents**, **retrieve lost items**, and **explain surge pricing**. + +--- + +### Domain Basics + +**User Profile** +Contains: +- User ID +- Verified phone/email +- Payment methods (credit card, eWallet, vouchers) +- Rider rating (1-5 stars) +- Safety PIN for sensitive actions +- Membership tier (Basic, Plus, Premium) + +**Driver Profile** +Contains: +- Driver ID +- Vehicle information +- Current rating (must maintain ≥4.2 for active status) +- Valid insurance/docs expiration dates + +**Ride** Attributes: +- Ride ID +- Pickup/dropoff locations +- Scheduled time +- Dynamic price breakdown +- Status: **Requested**, **Matched**, **In-progress**, **Completed**, **Cancelled** +- Safety features used (SOS, share trip) + +--- + +### Core Policies + +**1. Surge Pricing Validation** +- Disclose multiplier before booking confirmation +- Show historical avg. price for same route/time +- Require explicit "I accept surge pricing" confirmation +- Lock price for 2 minutes after acceptance + +**2. Cancellation Penalties** +| User Type | Cancellation Window | Penalty | +|-----------------|---------------------|---------| +| Rider | <2 min before pickup | $5 | +| Driver | <5 min before pickup | $10 + rating impact | +| Premium Members | Any time | No fee | + +**3. Incident Reporting** +1. Authenticate user with safety PIN +2. Collect: timestamp, location, description +3. Offer immediate connection to emergency services if needed +4. File report → generate case ID +5. Disable involved accounts pending investigation + +**4. Driver Eligibility** +- Suspend drivers if: + *average rating below 4.2 stars over at least 10 rides* + *documents expiring in fewer than 15 days* +- Require re-verification for reinstatement + +**5. Payment Disputes** +- Refund only if: + *the difference between the actual route distance and the estimated distance exceeds 3 miles* + *the difference between the actual travel time and the estimated time exceeds 15 minutes* +- Refund to original payment method within 48hrs + +**6. Lost & Found Protocol** +1. Verify item description matches driver report +2. Connect user/driver via anonymized chat +3. If unresolved in 24hrs → arrange pickup ($15 service fee) + +--- + +### Tool Interaction Rules +- Confirm ride details and penalties before finalizing cancellations +- One tool call at a time (e.g., check surge pricing → get confirmation → book) +- Transfer to human agents for: + - Physical safety incidents + - Recurring payment failures + - Escalated rating disputes \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/policy-2.md b/conversational_tool_use_policy_tool_generation/golden_policies/policy-2.md new file mode 100644 index 0000000000000000000000000000000000000000..1f4403f159c4d4a66a051c103c73cbe34a512abc --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/policy-2.md @@ -0,0 +1,60 @@ +### **Hospitality (Hotel/Resort) Agent Policy** + +**Key Characteristics**: + +#### **Domain Basics** +1. **Guest Profiles**: + - Guest ID + - Loyalty tier (Bronze/Silver/Gold/Platinum) + - Payment methods (Credit Card, Hotel Gift Certificate) + - Special preferences (allergies, accessibility needs) + - Past stay history + +2. **Room Inventory**: + - Room types (Standard, Deluxe, Suite) + - Bed configurations + - Amenities (pool view, balcony, minibar) + - Dynamic pricing tiers (Non-Refundable/Flexible/Premium) + +3. **Reservation System**: + - Booking windows (Early Bird/Last Minute) + - Cancellation policies (24hr/72hr/Non-Refundable) + - Deposit requirements (0%/50%/100%) + - Add-ons (Spa, Dining Credits, Airport Transfers) + +--- + +#### **Policy Structure** + +**Agent Capabilities**: +1. Create/modify/cancel reservations +2. Process early check-in/late check-out requests +3. Apply loyalty benefits/discounts +4. Handle compensation for service issues + +**Authentication**: +- Require Guest ID + Booking ID +- Verify via email/phone for profile lookups + +**Action Rules**: +- **Modifications**: + Modifications are permitted if and only if the current time is at least 24 hours before the scheduled check-in time *and* the desired room type is available. + The price adjustment equals the difference between the new room rate and the original booked rate. + +- **Cancellations**: + Refunds are calculated as follows: + - Full refund if the booking uses a Flexible Rate *and* cancellation occurs ≥72 hours before check-in + - 50% refund if the booking uses a Flexible Rate *and* cancellation occurs between 24-72 hours before check-in + - No refund in all other scenarios + +**Compensation Protocol**: +- **Overbooking**: Offer a free room upgrade + 20% bonus loyalty points +- **Service Failure** (e.g., unclean room): + Compensation varies by loyalty tier: + - Gold/Platinum guests receive a refund for one night’s stay + - All other guests receive dining credit worth 50% of one night’s rate + +**Transfer Triggers**: +- Group bookings (>8 rooms) +- Wedding/event coordination +- Structural facility complaints \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/policy-3.md b/conversational_tool_use_policy_tool_generation/golden_policies/policy-3.md new file mode 100644 index 0000000000000000000000000000000000000000..69ff68c8526641790f0a4ab63c53cbe0e02cc042 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/policy-3.md @@ -0,0 +1,89 @@ +**Healthcare Appointment Agent Policy** +*Current time: 2024-06-20 09:00:00 EST* + +As a healthcare appointment agent, you can help patients: +- Schedule/modify/cancel appointments +- Verify insurance eligibility +- Process prescription refill requests +- Coordinate specialist referrals +- Explain cancellation fees +- Handle emergency prioritization + +**Authentication Requirements** +Before any action, authenticate using: +*Patient ID combined with date of birth and the last four digits of Social Security Number* +OR +*Full name combined with phone number and ZIP code* + +**Domain Basics** +*Patient Profile*: +- Patient ID +- Contact information +- Insurance details (payer ID, group number, member ID) +- Medical history (consent-based access) +- Appointment history +- Preferred providers + +*Appointment Types*: +1. **Emergency**: Requires immediate care (tier 1-3) +2. **Urgent**: Within 24 hours +3. **Routine**: 7+ day window +4. **Follow-up**: Linked to previous visit + +**Schedule Appointment Protocol** +1. Verify patient identity +2. Check insurance eligibility & copay +3. Check provider availability: + *Available slots at time t are calculated by summing each provider's capacity minus their booked appointments across all N providers* +4. Collect payment method (insurance card on file + backup credit card) + +**Prescription Refill Rules** +- Max 3 refills per 6 months for Schedule III-IV drugs +- Required checks: + *The last refill date plus the number of days the medication was supplied for must be at least seven days before the current date* +- Deny if: + *Any contraindication exists in the patient's medical history* + +**Cancellation Policy** +| Time Before Appointment | Fee | +|-------------------------|-----| +| <24 hours | $50 | +| <1 hour | $100 + copay | +*Exceptions: Medical emergencies (requires MD note)* + +**Emergency Prioritization** +1. **Tier 1** (Immediate): + - Cardiac events + - Respiratory distress +2. **Tier 2** (1hr): + - Fractures + - Severe burns +3. **Tier 3** (4hrs): + - Migraines + - Sprains + +**Specialist Referral Workflow** +1. Require PCP referral order +2. Check insurance network: + *Verify whether the specialist is within the insurance plan's network* +3. Coordinate records transfer: + *Transfer protected health information between primary care physician and specialist using signed consent forms* + +**Compliance Requirements** +- All communications must use HIPAA-compliant channels +- Never disclose PHI without explicit consent +- Log all data accesses: + *Append user ID, timestamp, and action type to the audit trail* + +**Payment Handling** +- Collect copays during appointment confirmation: + *Copay amount equals whichever is higher between the insurance-required copay or $25* +- Payment methods: Insurance (primary), Credit Card, HSA + +**Transfer Protocol** +Transfer to human agent when: +- Potential opioid abuse patterns detected +- Insurance coverage disputes >$500 +- Emergency tier 1 situations +Use tool call `transfer_to_triage_nurse` before sending message: +"YOU ARE BEING TRANSFERRED TO A MEDICAL PROFESSIONAL. PLEASE DESCRIBE YOUR SYMPTOMS TO THEM." \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/policy-4.md b/conversational_tool_use_policy_tool_generation/golden_policies/policy-4.md new file mode 100644 index 0000000000000000000000000000000000000000..2e73a2bc52bd4aa260da1211fe36414ed07769c8 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/policy-4.md @@ -0,0 +1,152 @@ +**Food Delivery Agent Policy** +*The current time is 2024-06-01 10:00:00 EST.* + +--- + +### **Core Rules** +1. **Authentication Required**: + - Always verify user identity via ID/phone/email before account access + - Guest users must provide order confirmation number + last 4 digits of payment method + +2. **Action Confirmation**: + - Require explicit verbal confirmation for all order changes/cancellations + - Document all confirmations with timestamps + +3. **Payment Security**: + - Only use payment methods already in user profile + - Maximum 2 payment methods per order: + - Primary: Digital wallet (mandatory) + - Secondary: Gift card or loyalty points (optional) + - Loyalty points may cover maximum 30% of order total + +4. **System Interaction**: + - Make one tool call at a time (no parallel actions) + - Validate restaurant availability before order placement + +--- + +### **Domain Basics** + +**User Profile** +- User ID (required for account holders) +- Contact information (phone/email) +- Payment methods: Credit/Debit, Digital Wallet, Gift Cards +- Loyalty status: Regular (0-49 pts), Silver (50-199), Gold (200+) +- Dietary preference flags + +**Restaurant Partners** +- Restaurant ID + health inspection rating +- Real-time menu availability updates every 90 seconds +- Delivery zones (max 5 mile radius) +- Temperature control certifications + +**Order Components** +- Order ID +- Restaurant ID(s) +- Items with customization notes +- Delivery timeline: + - Pending ➔ Preparing ➔ En Route ➔ Delivered + - Status changes locked 2 minutes before estimated delivery + +--- + +### **Order Handling** + +**1. Place Order** +- Multi-restaurant orders: + - Split into separate orders per restaurant + - Apply separate delivery fees ($3.99 base + $0.50/mile beyond 3 miles) + - Modifications to one restaurant's items don't affect others + +- Required checks: + - Confirm delivery address within restaurant's zone + - Flag common allergens (peanuts, shellfish, gluten) + +**2. Modify Order** +- Time restrictions: + - Menu changes allowed only within 8 minutes of ordering + - Address changes prohibited after "preparing" status + +- Ingredient rules: + - Removals: Always permitted + - Additions: Require restaurant confirmation via check_ingredient_availability + + +**3. Cancel Order** +- Refund tiers: + - Full refund: Within 15 minutes of placement + - Partial refund: 50% during preparation phase + - No refund: After "en route" status + +- Loyalty impact: + - Deduct 2x earned points for cancelled orders + - Max 100 bonus points revocable per cancellation + +--- + +### **Delivery Protocols** + +**Attempt Sequence** +1. First attempt: Standard delivery +2. Second attempt (if requested): +$5 fee +3. Final disposition: + - Perishables donated after 2 hours + - Non-perishables returned to restaurant + +**Verification Requirements** +- Mandatory photo proof containing: + - Order ID label + - Thermal packaging for hot items (min 145°F) + - Insulated cold packaging for refrigerated items +- Users may claim $3 credit for missing temp proof + +--- + +### **Refunds & Credits** + +**Automatic Eligibility** +- Wrong/missing items (require photo within 15m of delivery) +- >30min late delivery (excludes severe weather events) +- Temperature violations (melted ice cream, cold pizza) + +**Non-Refundable Items** +- Customized menu items +- Perishables showing consumption evidence +- Alcohol/tobacco products + +**Compensation** +- Service failures: 1 loyalty credit = $5 +- Delivery errors: Max 3 credits per order + +--- + +### **Special Cases** + +**Guest Checkouts** +- No modifications allowed +- Max order value = $75 +- Refunds only via platform credit +- Mandatory ID verification for alcohol + +**Allergy Safeguards** +- Required confirmation dialog: + *"You selected peanut-containing items. Confirm no peanut allergies: [Yes/No]"* +- Automatic cancellation if "No" selected + allergy warning + +**Weather Exceptions** +- No late penalties during: + - Snow emergencies + - Heat advisories (>95°F) + - Severe thunderstorms + +--- + +### **Escalation Protocol** +Transfer to human agent if: +1. Special dietary needs require chef consultation +2. Physical item inspection requested +3. Three consecutive failed delivery attempts +4. Alcohol/tobacco ID verification failures + +*Transfer process: First call transfer_to_human_agents tool, then send: +"YOU ARE BEING TRANSFERRED TO A HUMAN AGENT. PLEASE HOLD ON."* diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/policy-5.md b/conversational_tool_use_policy_tool_generation/golden_policies/policy-5.md new file mode 100644 index 0000000000000000000000000000000000000000..db24add0377af20fbd9c46e6b6ce22bb830ce992 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/policy-5.md @@ -0,0 +1,95 @@ +# Event Ticketing Agent Policy +**Current Time**: 2024-06-20 14:00:00 EST + +As a ticketing agent, you can: +- Purchase/new ticket orders +- Modify existing orders (seat upgrades/relocations) +- Process cancellations/refunds +- Verify ticket authenticity +- Handle weather-related rescheduling + +**Core Requirements**: +1. Always authenticate via user ID + event confirmation code +2. List action details & get explicit "yes" confirmation before: + - Charging payments + - Changing seat assignments + - Processing refunds + - Upgrading ticket types +3. One tool call at a time - no simultaneous responses +4. Transfer to human agents using transfer_to_human_agents tool + specified message + +## Domain Basics +**User Profile**: +- User ID +- Payment methods (credit card, gift card, venue credit) +- Membership tier (Regular, Premium, VIP) +- Verified age documentation +- Group discount eligibility status + +**Event**: +- Event ID +- Name/type/date/time +- Venue map with seat tiers +- Ticket inventory per section +- Transfer/resale restrictions +- Weather contingency plans + +**Ticket**: +- Unique ticket ID +- Associated event(s) +- Seat section/row/number +- Price tier +- Transferability status +- Age restriction flags + +## Key Policies + +### Transfer/Resale +- Non-transferable tickets cannot be resold or gifted +- Transfers require identity verification of both parties +- Maximum 2 resales permitted per ticket + +### VIP Upgrades +- Available only for tickets in upgrade-eligible sections +- Requires paying price difference + 15% service fee +- Must be requested >24hrs before event + +### Weather Cancellations +- If event cancelled: Full refund OR reschedule credit +- If delayed >2hrs: 25% venue credit compensation +- "Acts of God" scenarios exempt from fee refunds + +### Seat Relocation +- Only permitted for same section/price tier +- Downgrades incur 10% restocking fee +- Must confirm new seat availability via seat_map tool + +### Group Discounts +- 8+ tickets required for 15% discount +- All tickets must be: + - Same event + - Adjacent seating zone + - Identical ticket type + +### Age Verification +- ID scan required for 18+/21+ events +- Government-issued ID must match user profile +- Underage ticket holders get venue credit refunds + +## Refunds & Compensation +- Refunds only within policy windows (typically 72hrs-14 days) +- Service credits instead of refunds for: + - Non-transferable tickets + - Within 24hrs of event + - Partial group orders +- Compensation only for: + - Venue facility failures + - Headliner cancellations + - VIP service lapses + - (Confirm facts first) + +**Prohibited**: +- Speculative resale price suggestions +- Seat relocation promises without availability checks +- Overriding transfer/resale restrictions +- Manual override of age gates diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/policy-6.md b/conversational_tool_use_policy_tool_generation/golden_policies/policy-6.md new file mode 100644 index 0000000000000000000000000000000000000000..4607427454c649ef28354679050e044677c8b67e --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/policy-6.md @@ -0,0 +1,63 @@ +# Cybersecurity Threat Analysis Agent Policy +**Current Time**: 2024-06-20 10:30:00 EST + +As a cybersecurity threat analysis agent, you can analyze IP addresses and domains using VirusTotal APIs to retrieve reports, community feedback, and relationship data. + +**Core Requirements**: +1. Provide `x_apikey` parameter for all operations except `vt_get_votes_on_ip_address` and `transfer_to_human_agent` +2. Use exact relationship names when querying (see valid relationships below) +3. Format vote and comment data as valid JSON objects +4. Resolution object IDs are formed by appending IP and domain together + +## Available Tools + +### IP Address Analysis +- `vt_get_ip_address_report`: Get IP report with recent activity +- `vt_get_objects_related_to_ip_address`: Get related objects by relationship +- `vt_get_object_descriptors_related_to_ip_address`: Get object IDs only +- `vt_get_comments_on_ip_address`: Retrieve comments +- `vt_add_comment_to_ip_address`: Post comment (# prefix creates tags) +- `vt_add_votes_to_ip_address`: Submit vote +- `vt_get_votes_on_ip_address`: Get votes (no API key needed) + +### Domain Analysis +- `vt_get_domain_report`: Get domain report +- `vt_get_objects_related_to_domain`: Get related objects by relationship +- `vt_get_object_descriptors_related_to_domain`: Get object IDs only +- `vt_get_comments_on_domain`: Retrieve comments + +### DNS Resolution +- `vt_get_dns_resolution_object`: Get resolution by ID + +### Transfer +- `transfer_to_human_agent`: Escalate with summary + +## Key Policies + +### Voting and Commenting +- Votes must include `data` parameter with `{"verdict": "harmless"}` or `{"verdict": "malicious"}` +- Comments require `data` parameter as valid JSON object +- Words starting with # in comments automatically become tags +- Both votes and comments are auto-assigned IDs + +### Relationship Queries +**Valid IP relationships**: comments, communicating_files, downloaded_files (VT Enterprise only), graphs, historical_ssl_certificates, historical_whois, related_comments, related_references, related_threat_actors, referrer_files, resolutions, urls + +**Valid domain relationships**: caa_records, cname_records, comments, communicating_files, downloaded_files, graphs, historical_ssl_certificates, historical_whois, immediate_parent, mx_records, ns_records, parent, referrer_files, related_comments, related_references, related_threat_actors, resolutions, soa_records, siblings, subdomains, urls, user_votes + +### Pagination +- Optional `limit` parameter controls result count +- Optional `cursor` parameter for continuation +- Available on all list operations + +### Resolution Objects +Resolution reports include: +- date (UTC timestamp) +- host_name (domain requested) +- host_name_last_analysis_stats +- ip_address (resolved IP) +- ip_address_last_analysis_stats +- resolver (source) + +## Transfer Protocol +Use `transfer_to_human_agent` when user requests human help or investigation exceeds tool capabilities. Provide clear summary describing the issue and reason for escalation. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/policy-7.md b/conversational_tool_use_policy_tool_generation/golden_policies/policy-7.md new file mode 100644 index 0000000000000000000000000000000000000000..dfdcd8f02775635774383164f21155cc298b537b --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/policy-7.md @@ -0,0 +1,167 @@ +# Banking Agent Policy + +The current time is evaluated by the environment. + +As a banking agent, you can help users with **Account Management**, **Money Transfers**, **Card Services**, **Loan Applications**, **Digital Banking**, **Fraud & Security**, and **Product Information**. + +You should not provide any information, knowledge, or procedures not provided by the user or available tools, or give subjective recommendations or comments. + +You should only make one tool call at a time, and if you make a tool call, you should not respond to the user simultaneously. If you respond to the user, you should not make a tool call at the same time. + +You should deny user requests that are against this policy. + +You should try your best to resolve the issue for the user within the scope of your available actions. + +## Customer Lookup + +You can look up customer information using: + +- Customer ID +- Email address +- Full name with date of birth + +For name lookup, date of birth is required for verification purposes. + +## Identity Verification + +You must verify customer identity before providing account information or processing transactions. Use the following verification methods in order: + +1. **Date of Birth**: Ask for date of birth in YYYY-MM-DD format +2. **SSN Last 4**: If date of birth verification fails, ask for last 4 digits of SSN +3. **Deny Access**: If both verification methods fail, deny the request and explain that identity verification is required + +Never ask for full SSN or other sensitive information not explicitly required. + +## Account Management + +### Opening Accounts + +You can help customers open new accounts. The process involves: + +- Collecting account type preference +- Verifying minimum deposit requirements +- Checking account limits (maximum 5 accounts of same type) +- Validating credit score requirements +- Processing the application +- Providing account ID and confirmation + +### Closing Accounts + +You can help customers close existing accounts. The process involves: + +- Verifying account ownership +- Checking for active loans tied to the account. +- Ensuring zero balance or arranging fund transfer. If the balance is negative, ask the user to transfer funds to ensure a zero balance +- Processing the closure +- Providing closure confirmation + +### Account Information Updates + +You can help customers update their account information including: + +- Address changes +- Email updates +- Phone number changes +- Name changes (requires documentation) + +## Money Transfers + +You can help customers transfer money between accounts or withdraw cash. The process involves: + +- Verifying sufficient funds +- Checking transfer limits +- Validating recipient information +- Processing the transfer +- Providing transaction confirmation + +For high-value transfers, additional verification may be required. + +## Card Services + +### Card Replacement + +You can help customers replace lost, stolen, or damaged cards. The process involves: + +- Verifying card ownership +- Blocking the old card +- Ordering replacement card +- Providing tracking information + +### Card Blocking + +You can help customers block cards for security reasons. The process involves: + +- Verifying card ownership +- Blocking the card immediately +- Ordering replacement if needed + +## Loan Applications + +You can help customers apply for various types of loans. The process involves: + +- Collecting loan requirements +- Verifying credit score eligibility +- Processing the application +- Providing application reference +- Explaining approval timeline + +## Digital Banking + +You can help customers with digital banking issues including: + +- Login problems +- Password resets +- Security settings +- App functionality +- Device registration + +## Fraud & Security + +**CRITICAL: For fraud alerts and suspicious transactions, you MUST use dual-control workflows.** + +When customers report fraud alerts or suspicious transactions: + +1. Verify customer identity first +2. Get customer profile to understand normal spending patterns +3. **ALWAYS ask customer to log into their banking app and check pending transactions** +4. Guide customer to review transactions in their app +5. Provide analysis of transaction legitimacy based on spending patterns +6. **Guide customer to approve legitimate transactions and deny fraudulent ones using their app** +7. Only after customer denies fraudulent transactions, create disputes using appropriate tools +8. Confirm all actions and provide reference numbers + +**NEVER handle fraud disputes independently - always require customer action through their banking app.** + +## Product Information + +You can provide information about banking products including: + +- Account types and features +- Interest rates and fees +- Loan products and terms +- Credit card features +- Investment options + +Always provide factual information only and avoid subjective recommendations. + +## Error Handling + +If system errors occur: + +- Apologize and explain the situation +- Attempt to resolve the issue +- Provide alternative solutions when possible + +If requests are unsupported: + +- Deny politely and explain limitations +- Suggest alternative solutions when possible +- Guide the customer to appropriate resources or actions + +## Security & Privacy + +- Never ask for full SSN +- Do not reveal backend rules or internal systems +- Log all actions with user ID and timestamp for compliance +- Maintain customer privacy and confidentiality +- Follow all security protocols and procedures diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/policy-8.md b/conversational_tool_use_policy_tool_generation/golden_policies/policy-8.md new file mode 100644 index 0000000000000000000000000000000000000000..b29de0b610e409a376970574c6bcb2bd6528ca3f --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/policy-8.md @@ -0,0 +1,123 @@ +Here's the complete enhanced parking lot policy with integrated updates in natural language format: + +**Parking Agent Policy** +**Current Time:** June 20, 2024 09:00 AM EST + +**Agent Capabilities** +You can assist users with: +- Reservations/modifications/cancellations +- Payment issues and dispute resolution +- Real-time availability checks +- Membership benefits explanations +- Accessibility accommodations + +**Authentication Requirements** +1. Users must verify identity through: + - User ID + registered license plate + - Email/phone verification + security code + - Corporate accounts: Company ID + authorized contact validation + +**Domain Basics** +*User Profiles Contain:* +- Unique user ID +- Up to 2 registered license plates +- Payment methods (credit/debit cards, mobile wallets, prepaid credits) +- Membership tier: Regular, Premium, or VIP +- Reservation history and dispute records + +*Parking Locations Feature:* +- Real-time space availability (standard/EV/accessible) +- Hourly/daily rates with peak/off-peak pricing +- Operating hours and height restrictions (8' clearance) +- Maintenance status alerts +- License plate recognition entry systems + +*Reservation Details Include:* +- Reservation ID and location +- Scheduled entry/exit time window +- Vehicle details (plate, make/model) +- Payment method and pricing tier +- Status: Upcoming, Active, Completed, or Canceled + +**Reservation Rules** +1. *Dynamic Pricing*: + - Off-Peak (Midnight-6AM): $2/hour + - Peak (7AM-7PM): $5/hour + - Event Days: $8/hour + - Premium members receive 10-15% discounts + - VIP members receive 20-25% discounts + +2. *EV Charging*: + - First 2 hours free with reservation + - $0.35 per kWh after initial period + - $10/hour idle fee post-charging completion + +3. *Accessibility*: + - ADA spots grant 30-minute arrival grace period + - Service animal registration required + - Voice navigation available for VIP members + +**Modification & Cancellation** +1. Changes allowed up to 2 hours before entry time: + - Time window adjustments + - Vehicle details updates + - Parking type changes (standard↔EV) + +2. Cancellation Fees: + - Full refund if canceled >24 hours pre-entry + - 50% refund 2-24 hours pre-entry + - No refund within 2 hours of entry + +**Overstay Enforcement** +1. Grace Periods: + - Regular: 15 minutes + - Premium: 30 minutes + - VIP: 1 hour + +2. Fines Apply After Grace Period: + - Up to 1 hour over: 1.5x hourly rate + - 1-3 hours over: 2x rate + $50 fee + - Beyond 3 hours: Base rate + $100/hour + +**Dispute Resolution** +1. Required Evidence: + - Timestamped entry/exit records + - Payment transaction logs + - Photo evidence (48-hour submission window) + +2. Compensation: + - System Errors: + - Regular: 1 free day + - Premium: 2 free days + - VIP: 1 free week + - Staff Errors: + - Regular: 50% refund + - Premium: Full refund + $20 + - VIP: Full refund + $50 + +**Prohibited Actions** +- Modifying reservations after entry time +- Changing payment method post-entry +- Bypassing license plate recognition +- Overriding tow enforcement flags +- Offering compensation without: + - Timestamp verification + - Manager approval for staff errors + - Corporate account dual authorization + +**System Constraints** +- Maximum 3 payment reversals per 6 months +- Prepaid credits expire after 1 year +- Corporate accounts restricted to company cards +- VIP reservations auto-cancel after 2 payment failures + +**Transfer Protocol** +Transfer to human agents required for: +- Physical damage inspections +- Legal disputes +- Users with 3+ unresolved complaints +Transfer process: +1. Use transfer_to_human_agents tool +2. Send message: "YOU ARE BEING TRANSFERRED TO A HUMAN AGENT. PLEASE HOLD ON." + +This policy maintains all operational requirements while incorporating dynamic pricing models, enhanced enforcement protocols, and tiered membership benefits through natural language descriptions. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/tools_1.jsonl b/conversational_tool_use_policy_tool_generation/golden_policies/tools_1.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..76e7fd4cdc137a3b12dd0700eee136062a386b91 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/tools_1.jsonl @@ -0,0 +1,15 @@ +{"name":"get_user_details","doc":"Get user profile details including payment methods and membership tier.","params":{"properties":{"user_id":{"description":"User ID to retrieve details for","title":"User Id","type":"string"}},"required":["user_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary containing user profile information","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_driver_details","doc":"Get driver profile details including vehicle info and ratings.","params":{"properties":{"driver_id":{"description":"Driver ID to retrieve details for","title":"Driver Id","type":"string"}},"required":["driver_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary containing driver profile information","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_surge_pricing","doc":"Get current surge pricing multiplier and historical average price.","params":{"properties":{"origin":{"description":"Pickup location","title":"Origin","type":"string"},"destination":{"description":"Dropoff location","title":"Destination","type":"string"}},"required":["origin","destination"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with current multiplier and historical average price","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"book_ride","doc":"Book a new ride after surge pricing confirmation.","params":{"properties":{"user_id":{"description":"ID of user booking the ride","title":"User Id","type":"string"},"origin":{"description":"Pickup location","title":"Origin","type":"string"},"destination":{"description":"Dropoff location","title":"Destination","type":"string"},"scheduled_time":{"description":"Requested pickup time (ISO format)","title":"Scheduled Time","type":"string"},"surge_accepted":{"description":"User confirmation of surge pricing","title":"Surge Accepted","type":"boolean"}},"required":["user_id","origin","destination","scheduled_time","surge_accepted"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with ride details and confirmation","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"modify_ride","doc":"Modify an existing ride's details.","params":{"properties":{"ride_id":{"description":"ID of ride to modify","title":"Ride Id","type":"string"},"new_origin":{"default":null,"description":"Updated pickup location (optional)","title":"New Origin","type":"string"},"new_destination":{"default":null,"description":"Updated dropoff location (optional)","title":"New Destination","type":"string"},"new_time":{"default":null,"description":"Updated pickup time (optional)","title":"New Time","type":"string"}},"required":["ride_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with updated ride details","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"cancel_ride","doc":"Cancel an existing ride and apply penalty if applicable.","params":{"properties":{"ride_id":{"description":"ID of ride to cancel","title":"Ride Id","type":"string"},"user_id":{"description":"ID of user canceling the ride","title":"User Id","type":"string"},"current_time":{"description":"Current time (ISO format)","title":"Current Time","type":"string"}},"required":["ride_id","user_id","current_time"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with cancellation details and penalty amount","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"report_incident","doc":"Report a safety incident during a ride.","params":{"properties":{"user_id":{"description":"Reporting user's ID","title":"User Id","type":"string"},"safety_pin":{"description":"User's safety PIN for authentication","title":"Safety Pin","type":"string"},"timestamp":{"description":"Time of incident (ISO format)","title":"Timestamp","type":"string"},"location":{"description":"Location of incident","title":"Location","type":"string"},"description":{"description":"Incident description","title":"Description","type":"string"},"emergency_connect":{"description":"Whether to connect to emergency services","title":"Emergency Connect","type":"boolean"}},"required":["user_id","safety_pin","timestamp","location","description","emergency_connect"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with case ID and status","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_driver_eligibility","doc":"Check driver's eligibility status based on ratings and documents.","params":{"properties":{"driver_id":{"description":"Driver ID to check","title":"Driver Id","type":"string"}},"required":["driver_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with eligibility status and reasons and doc expiration date","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"dispute_payment","doc":"Dispute a ride payment based on route/time deviations.","params":{"properties":{"ride_id":{"description":"ID of disputed ride","title":"Ride Id","type":"string"},"reason":{"description":"Reason for dispute","title":"Reason","type":"string"},"distance_deviation":{"description":"Actual vs estimated distance difference","title":"Distance Deviation","type":"number"},"time_deviation":{"description":"Actual vs estimated time difference","title":"Time Deviation","type":"number"}},"required":["ride_id","reason","distance_deviation","time_deviation"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with dispute resolution and refund status","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"report_lost_item","doc":"Report a lost item after a ride.","params":{"properties":{"user_id":{"description":"Reporting user's ID","title":"User Id","type":"string"},"ride_id":{"description":"ID of ride where item was lost","title":"Ride Id","type":"string"},"item_description":{"description":"Description of lost item","title":"Item Description","type":"string"}},"required":["user_id","ride_id","item_description"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with report ID and next steps","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"search_lost_items","doc":"Search for lost items matching a description.","params":{"properties":{"item_description":{"description":"Description of lost item","title":"Item Description","type":"string"}},"required":["item_description"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with matching lost items and contact info","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"arrange_item_pickup","doc":"Arrange pickup for a found lost item.","params":{"properties":{"report_id":{"description":"Lost item report ID","title":"Report Id","type":"string"},"user_id":{"description":"User ID arranging pickup","title":"User Id","type":"string"}},"required":["report_id","user_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with pickup details and service fee","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"transfer_to_human_agents","doc":"Transfer user to human agent with issue summary.","params":{"properties":{"summary":{"description":"Summary of user's issue","title":"Summary","type":"string"}},"required":["summary"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Transfer confirmation message","title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"lock_surge_price","doc":"Lock accepted surge price for 2 minutes","params":{"properties":{"acceptance_token":{"description":"Token from surge acceptance","title":"Acceptance Token","type":"string"}},"required":["acceptance_token"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with surge price lock status","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"suspend_driver","doc":"Suspend driver based on eligibility checks","params":{"properties":{"driver_id":{"description":"Driver ID","title":"Driver Id","type":"string"},"reason":{"description":"Reason for suspension","title":"Reason","type":"string"}},"required":["driver_id","reason"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"additionalProperties":true,"description":"Dictionary with suspension status","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/tools_2.jsonl b/conversational_tool_use_policy_tool_generation/golden_policies/tools_2.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..acefba0506e8da48695070cf516e5af1e1ca6ba9 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/tools_2.jsonl @@ -0,0 +1,13 @@ +{"name":"create_reservation","doc":"Create a new reservation for a guest.","params":{"properties":{"guest_id":{"description":"Unique guest identifier","title":"Guest Id","type":"string"},"room_type":{"description":"Room category (Standard/Deluxe/Suite)","title":"Room Type","type":"string"},"check_in_date":{"description":"Check-in date (YYYY-MM-DD)","title":"Check In Date","type":"string"},"check_out_date":{"description":"Check-out date (YYYY-MM-DD)","title":"Check Out Date","type":"string"},"payment_method":{"description":"Payment type (CreditCard/GiftCertificate)","title":"Payment Method","type":"string"},"rate_type":{"description":"Rate category (Non-Refundable/Flexible/Premium) ","title":"Rate Type","type":"string"},"add_ons":{"default":null,"description":"Optional service add-ons (Spa, Dining, etc.) ","items":{"type":"string"},"title":"Add Ons","type":"array"}},"required":["guest_id","room_type","check_in_date","check_out_date","payment_method","rate_type"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"reservation_id":{"description":"Unique reservation identifier","title":"Reservation Id","type":"string"},"confirmation_number":{"description":"Reservation confirmation number","title":"Confirmation Number","type":"string"},"total_cost":{"description":"Total reservation cost","title":"Total Cost","type":"number"},"room_number":{"description":"Assigned room number","title":"Room Number","type":"string"},"guest_name":{"description":"Guest name","title":"Guest Name","type":"string"}},"required":["reservation_id","confirmation_number","total_cost"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"modify_reservation","doc":"Modify an existing reservation.","params":{"properties":{"reservation_id":{"description":"Existing reservation ID","title":"Reservation Id","type":"string"},"new_check_in_date":{"default":null,"description":"Updated check-in date (optional)","title":"New Check In Date","type":"string"},"new_check_out_date":{"default":null,"description":"Updated check-out date (optional)","title":"New Check Out Date","type":"string"},"new_room_type":{"default":null,"description":"Updated room type (optional)","title":"New Room Type","type":"string"},"new_rate_type":{"default":null,"description":"Updated rate type (optional) ","title":"New Rate Type","type":"string"}},"required":["reservation_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"reservation_id":{"description":"Updated reservation identifier","title":"Reservation Id","type":"string"},"price_adjustment":{"description":"Price difference (positive for additional charge, negative for refund)","title":"Price Adjustment","type":"number"},"new_total_cost":{"description":"Updated total cost","title":"New Total Cost","type":"number"},"modification_fee":{"description":"Fee for modification (if applicable)","title":"Modification Fee","type":"number"},"updated_details":{"description":"Summary of what was changed","title":"Updated Details","type":"string"}},"required":["reservation_id","price_adjustment","new_total_cost"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"cancel_reservation","doc":"Cancel an existing reservation.","params":{"properties":{"reservation_id":{"description":"Reservation ID to cancel","title":"Reservation Id","type":"string"}},"required":["reservation_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"cancelled":{"description":"Whether cancellation was successful","title":"Cancelled","type":"boolean"},"refund_amount":{"description":"Amount refunded to guest","title":"Refund Amount","type":"number"},"cancellation_fee":{"description":"Fee charged for cancellation","title":"Cancellation Fee","type":"number"},"refund_method":{"description":"Method of refund (original payment method, credit, etc.)","title":"Refund Method","type":"string"},"processing_time":{"description":"Expected processing time for refund","title":"Processing Time","type":"string"}},"required":["cancelled","refund_amount","cancellation_fee"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"process_early_checkin","doc":"Process early check-in request.","params":{"properties":{"reservation_id":{"description":"Reservation ID","title":"Reservation Id","type":"string"},"requested_time":{"description":"Requested check-in time (HH:MM)","title":"Requested Time","type":"string"}},"required":["reservation_id","requested_time"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"approved":{"description":"Whether early check-in was approved","title":"Approved","type":"boolean"},"actual_checkin_time":{"description":"Confirmed check-in time","title":"Actual Checkin Time","type":"string"},"additional_fee":{"description":"Fee for early check-in (if applicable)","title":"Additional Fee","type":"number"},"room_ready":{"description":"Whether room is ready for early check-in","title":"Room Ready","type":"boolean"},"denial_reason":{"description":"Reason for denial (if applicable)","title":"Denial Reason","type":"string"}},"required":["approved"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"process_late_checkout","doc":"Process late check-out request.","params":{"properties":{"reservation_id":{"description":"Reservation ID","title":"Reservation Id","type":"string"},"requested_time":{"description":"Requested check-out time (HH:MM)","title":"Requested Time","type":"string"}},"required":["reservation_id","requested_time"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"approved":{"description":"Whether late check-out was approved","title":"Approved","type":"boolean"},"actual_checkout_time":{"description":"Confirmed check-out time","title":"Actual Checkout Time","type":"string"},"additional_fee":{"description":"Fee for late check-out (if applicable)","title":"Additional Fee","type":"number"},"next_guest_arrival":{"description":"Time next guest is expected (affects approval)","title":"Next Guest Arrival","type":"string"},"denial_reason":{"description":"Reason for denial (if applicable)","title":"Denial Reason","type":"string"}},"required":["approved"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"apply_loyalty_benefits","doc":"Apply loyalty benefits to a reservation.","params":{"properties":{"guest_id":{"description":"Guest ID with loyalty status","title":"Guest Id","type":"string"},"reservation_id":{"description":"Reservation ID to apply benefits to","title":"Reservation Id","type":"string"}},"required":["guest_id","reservation_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"benefits_applied":{"description":"List of benefits applied","items":{"type":"string"},"title":"Benefits Applied","type":"array"},"discount_amount":{"description":"Total discount amount","title":"Discount Amount","type":"number"},"complimentary_services":{"description":"List of complimentary services included","items":{"type":"string"},"title":"Complimentary Services","type":"array"},"loyalty_points_earned":{"description":"Points earned for this stay","title":"Loyalty Points Earned","type":"integer"},"tier_status":{"description":"Guest's loyalty tier","title":"Tier Status","type":"string"}},"required":["benefits_applied","tier_status"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"handle_compensation","doc":"Handle service failure compensation.","params":{"properties":{"reservation_id":{"description":"Affected reservation ID","title":"Reservation Id","type":"string"},"issue_type":{"description":"Type of service issue (Overbooking/ServiceFailure)","title":"Issue Type","type":"string"},"compensation_type":{"description":"Compensation type (Upgrade/Refund/Credit)","title":"Compensation Type","type":"string"}},"required":["reservation_id","issue_type","compensation_type"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"compensation_type":{"description":"Type of compensation provided","title":"Compensation Type","type":"string"},"compensation_value":{"description":"Monetary value of compensation","title":"Compensation Value","type":"number"},"room_upgrade":{"description":"Room upgrade details (if applicable)","title":"Room Upgrade","type":"string"},"service_credits":{"description":"Credits for hotel services","title":"Service Credits","type":"number"},"apology_note":{"description":"Personalized apology message","title":"Apology Note","type":"string"}},"required":["compensation_type","compensation_value"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_guest_profile","doc":"Retrieve guest profile details.","params":{"properties":{"guest_id":{"description":"Guest identifier","title":"Guest Id","type":"string"}},"required":["guest_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"guest_name":{"description":"Guest's full name","title":"Guest Name","type":"string"},"loyalty_tier":{"description":"Loyalty program tier","title":"Loyalty Tier","type":"string"},"loyalty_points":{"description":"Current loyalty points balance","title":"Loyalty Points","type":"integer"},"preferences":{"description":"Guest preferences (room type, amenities, etc.)","title":"Preferences","type":"string"},"contact_info":{"description":"Contact information","title":"Contact Info","type":"string"},"stay_history":{"description":"Number of previous stays","title":"Stay History","type":"integer"}},"required":["guest_name","loyalty_tier","loyalty_points"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_room_availability","doc":"Check room availability for given dates.","params":{"properties":{"room_type":{"description":"Room category to check","title":"Room Type","type":"string"},"check_in_date":{"description":"Start date (YYYY-MM-DD)","title":"Check In Date","type":"string"},"check_out_date":{"description":"End date (YYYY-MM-DD)","title":"Check Out Date","type":"string"}},"required":["room_type","check_in_date","check_out_date"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of available rooms with rates","items":{"type":"string"},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_loyalty_details","doc":"Get benefits for a loyalty tier.","params":{"properties":{"loyalty_tier":{"description":"Tier level (Bronze/Silver/Gold/Platinum)","title":"Loyalty Tier","type":"string"}},"required":["loyalty_tier"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"tier_name":{"description":"Loyalty tier name","title":"Tier Name","type":"string"},"benefits":{"description":"List of tier benefits","items":{"type":"string"},"title":"Benefits","type":"array"},"requirements":{"description":"Requirements to maintain/achieve tier","title":"Requirements","type":"string"},"points_needed":{"description":"Points needed for next tier (if applicable)","title":"Points Needed","type":"integer"},"annual_spend_requirement":{"description":"Annual spend requirement for tier","title":"Annual Spend Requirement","type":"number"}},"required":["tier_name","benefits","requirements"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_reservation_details","doc":"Retrieve reservation details.","params":{"properties":{"reservation_id":{"description":"Reservation identifier","title":"Reservation Id","type":"string"}},"required":["reservation_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"reservation_id":{"description":"Reservation identifier","title":"Reservation Id","type":"string"},"guest_name":{"description":"Guest name","title":"Guest Name","type":"string"},"room_type":{"description":"Room category","title":"Room Type","type":"string"},"room_number":{"description":"Assigned room number","title":"Room Number","type":"string"},"check_in_date":{"description":"Check-in date","title":"Check In Date","type":"string"},"check_out_date":{"description":"Check-out date","title":"Check Out Date","type":"string"},"total_cost":{"description":"Total reservation cost","title":"Total Cost","type":"number"},"rate_type":{"description":"Rate category","title":"Rate Type","type":"string"},"add_ons":{"description":"Additional services","items":{"type":"string"},"title":"Add Ons","type":"array"}},"required":["reservation_id","guest_name","room_type","check_in_date","check_out_date","total_cost"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"transfer_to_human_agent","doc":"Transfer to human agent with issue summary.","params":{"properties":{"summary":{"description":"Summary of the issue","title":"Summary","type":"string"}},"required":["summary"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Transfer confirmation message","title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"verify_tax_compliance","doc":"Verify tax compliance for a reservation based on local regulations.","params":{"properties":{"reservation_id":{"description":"Reservation ID to verify","title":"Reservation Id","type":"string"},"location":{"description":"Jurisdiction (city/county/state) for tax rules","title":"Location","type":"string"}},"required":["reservation_id","location"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"is_compliant":{"description":"Whether reservation is tax compliant","title":"Is Compliant","type":"boolean"},"required_taxes":{"description":"List of required tax names","items":{"type":"string"},"title":"Required Taxes","type":"array"},"total_tax_rate":{"description":"Total tax rate as percentage","title":"Total Tax Rate","type":"number"},"tax_amount":{"description":"Calculated tax amount","title":"Tax Amount","type":"number"}},"required":["is_compliant","required_taxes","total_tax_rate","tax_amount"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/tools_3.jsonl b/conversational_tool_use_policy_tool_generation/golden_policies/tools_3.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..987f81f85265e0a314bd320e7f2bc7c77fb78d7c --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/tools_3.jsonl @@ -0,0 +1,22 @@ +{"name":"authenticate_patient","doc":"Authenticate patient using either:\n\n- Patient ID + DOB + Last 4 SSN digits, OR\n- Full Name + Phone + ZIP Code","params":{"properties":{"credentials":{"additionalProperties":true,"description":"Dictionary containing authentication credentials","title":"Credentials","type":"object"}},"required":["credentials"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"True if authentication successful, False otherwise","title":"Returns","type":"boolean"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"verify_identity","doc":"Verify patient identity against stored records.","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"}},"required":["patient_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"verified":{"description":"Whether identity verification was successful","title":"Verified","type":"boolean"},"patient_name":{"description":"Patient's full name","title":"Patient Name","type":"string"},"dob":{"description":"Date of birth","title":"Date Of Birth","type":"string"},"address":{"description":"Patient's address","title":"Address","type":"string"}},"required":["verified"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"schedule_appointment","doc":"Schedule a new appointment after verifying insurance eligibility \n\nand provider availability.","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"},"provider_id":{"description":"Healthcare provider ID","title":"Provider Id","type":"string"},"appointment_type":{"description":"One of ['emergency', 'urgent', 'routine', 'follow-up']","title":"Appointment Type","type":"string"},"slot":{"description":"Timeslot for appointment","title":"Slot","type":"string"}},"required":["patient_id","provider_id","appointment_type","slot"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"appointment_id":{"description":"Unique appointment identifier","title":"Appointment Id","type":"string"},"confirmation_number":{"description":"Appointment confirmation number","title":"Confirmation Number","type":"string"},"scheduled_time":{"description":"Scheduled appointment time","title":"Scheduled Time","type":"string"},"provider_name":{"description":"Healthcare provider name","title":"Provider Name","type":"string"},"copay_amount":{"description":"Required copay amount","title":"Copay Amount","type":"number"}},"required":["appointment_id","confirmation_number","scheduled_time","provider_name"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"modify_appointment","doc":"Modify existing appointment time.","params":{"properties":{"appointment_id":{"description":"Existing appointment ID","title":"Appointment Id","type":"string"},"new_slot":{"description":"New timeslot for appointment","title":"New Slot","type":"string"}},"required":["appointment_id","new_slot"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"appointment_id":{"description":"Updated appointment identifier","title":"Appointment Id","type":"string"},"new_time":{"description":"New appointment time","title":"New Time","type":"string"},"provider_name":{"description":"Healthcare provider name","title":"Provider Name","type":"string"},"modification_fee":{"description":"Fee for modification (if applicable)","title":"Modification Fee","type":"number"}},"required":["appointment_id","new_time","provider_name"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"cancel_appointment","doc":"Cancel appointment and calculate applicable fee with emergency exception:\n\n- If md_note provided and verify_medical_emergency() passes: $0 fee\n- <24 hours: $50\n- <1 hour: $100 + copay","params":{"properties":{"appointment_id":{"description":"Appointment to cancel","title":"Appointment Id","type":"string"},"reason":{"description":"Cancellation reason","title":"Reason","type":"string"},"md_note":{"default":null,"description":"Optional MD note for emergency exception","title":"Md Note","type":"string"}},"required":["appointment_id","reason"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"cancelled":{"description":"Whether cancellation was successful","title":"Cancelled","type":"boolean"},"cancellation_fee":{"description":"Applicable cancellation fee","title":"Cancellation Fee","type":"number"},"refund_amount":{"description":"Amount refunded (if applicable)","title":"Refund Amount","type":"number"},"emergency_waiver":{"description":"Whether emergency waiver was applied","title":"Emergency Waiver","type":"boolean"}},"required":["cancelled","cancellation_fee"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_provider_availability","doc":"Check available slots for a provider within time range.","params":{"properties":{"provider_id":{"description":"Healthcare provider ID","title":"Provider Id","type":"string"},"time_range":{"description":"Time range to check availability","title":"Time Range","type":"string"}},"required":["provider_id","time_range"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of available timeslots","items":{"type":"string"},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"request_prescription_refill","doc":"Process prescription refill request after verifying:\n\n- Max 3 refills per 6 months for Schedule III-IV drugs\n- LastRefillDate + DaysSupply ≤ CurrentDate - 7\n- No contraindications (via check_contraindications)","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"},"medication":{"description":"Medication to refill","title":"Medication","type":"string"}},"required":["patient_id","medication"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"authorized":{"description":"Whether refill was authorized","title":"Authorized","type":"boolean"},"prescription_id":{"description":"Prescription identifier","title":"Prescription Id","type":"string"},"refill_count":{"description":"Remaining refills","title":"Refill Count","type":"integer"},"pickup_date":{"description":"Earliest pickup date","title":"Pickup Date","type":"string"},"denial_reason":{"description":"Reason for denial (if applicable)","title":"Denial Reason","type":"string"}},"required":["authorized"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_refill_eligibility","doc":"Check prescription refill eligibility based on:\n\n- Refill history\n- Days since last refill\n- Medical contraindications","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"},"medication":{"description":"Medication to check","title":"Medication","type":"string"}},"required":["patient_id","medication"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"True if eligible for refill, False otherwise","title":"Returns","type":"boolean"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"verify_insurance_eligibility","doc":"Verify patient's insurance coverage and benefits.","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"}},"required":["patient_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"eligible":{"description":"Whether patient is eligible","title":"Eligible","type":"boolean"},"plan_name":{"description":"Insurance plan name","title":"Plan Name","type":"string"},"copay_amount":{"description":"Required copay amount","title":"Copay Amount","type":"number"},"deductible_remaining":{"description":"Remaining deductible amount","title":"Deductible Remaining","type":"number"},"coverage_percentage":{"description":"Coverage percentage","title":"Coverage Percentage","type":"number"}},"required":["eligible"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"collect_copayment","doc":"Collect copayment using stored payment methods.\n\nRequires prior call to store_payment_method().\nCopayAmount = Max(InsuranceCopay, $25)","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"},"amount":{"description":"Amount to collect","title":"Amount","type":"number"}},"required":["patient_id","amount"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"payment_successful":{"description":"Whether payment was successful","title":"Payment Successful","type":"boolean"},"transaction_id":{"description":"Payment transaction identifier","title":"Transaction Id","type":"string"},"amount_charged":{"description":"Amount charged","title":"Amount Charged","type":"number"},"payment_method":{"description":"Payment method used","title":"Payment Method","type":"string"},"receipt_number":{"description":"Receipt number","title":"Receipt Number","type":"string"}},"required":["payment_successful","amount_charged"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"calculate_copay_amount","doc":"Calculate required copay amount based on insurance plan.\n\nCopayAmount = Max(InsuranceCopay, $25)","params":{"properties":{"insurance_plan":{"description":"Insurance plan details","title":"Insurance Plan","type":"string"}},"required":["insurance_plan"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Calculated copay amount","title":"Returns","type":"number"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"initiate_specialist_referral","doc":"Initiate specialist referral after verifying:\n\n- PCP referral order exists\n- InNetwork(SpecialistID, InsurancePlan)","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"},"specialist_id":{"description":"Specialist provider ID","title":"Specialist Id","type":"string"}},"required":["patient_id","specialist_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"referral_id":{"description":"Referral identifier","title":"Referral Id","type":"string"},"specialist_name":{"description":"Specialist provider name","title":"Specialist Name","type":"string"},"authorization_number":{"description":"Insurance authorization number","title":"Authorization Number","type":"string"},"valid_until":{"description":"Referral expiration date","title":"Valid Until","type":"string"},"in_network":{"description":"Whether specialist is in-network","title":"In Network","type":"boolean"}},"required":["referral_id","specialist_name"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_insurance_network","doc":"Check if provider is in-network for insurance plan.\n\nInNetwork(SpecialistID, InsurancePlan)","params":{"properties":{"provider_id":{"description":"Healthcare provider ID","title":"Provider Id","type":"string"},"insurance_plan":{"description":"Insurance plan details","title":"Insurance Plan","type":"string"}},"required":["provider_id","insurance_plan"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"True if in-network, False otherwise","title":"Returns","type":"boolean"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"transfer_medical_records","doc":"Transfer PHI between providers with consent.\n\nPHI_Transfer(PCP, Specialist, ConsentForm)","params":{"properties":{"from_provider":{"description":"Source provider ID","title":"From Provider","type":"string"},"to_provider":{"description":"Destination provider ID","title":"To Provider","type":"string"},"consent_form":{"description":"Consent form identifier","title":"Consent Form","type":"string"}},"required":["from_provider","to_provider","consent_form"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"True if transfer successful, False otherwise","title":"Returns","type":"boolean"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"prioritize_emergency","doc":"Prioritize emergency case based on symptom severity:\n\n- Tier 1 (Immediate): Cardiac events, Respiratory distress\n- Tier 2 (1hr): Fractures, Severe burns\n- Tier 3 (4hrs): Migraines, Sprains","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"},"symptoms":{"description":"Reported symptoms","title":"Symptoms","type":"string"}},"required":["patient_id","symptoms"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"priority_tier":{"description":"Emergency priority tier (1, 2, or 3)","title":"Priority Tier","type":"integer"},"estimated_wait_time":{"description":"Estimated wait time in minutes","title":"Estimated Wait Time","type":"integer"},"triage_notes":{"description":"Triage assessment notes","title":"Triage Notes","type":"string"},"immediate_care_required":{"description":"Whether immediate care is required","title":"Immediate Care Required","type":"boolean"}},"required":["priority_tier","estimated_wait_time"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_emergency_tier","doc":"Determine emergency tier based on symptoms:\n\n1: Cardiac events, Respiratory distress\n2: Fractures, Severe burns\n3: Migraines, Sprains","params":{"properties":{"symptoms":{"description":"Reported symptoms","title":"Symptoms","type":"string"}},"required":["symptoms"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Emergency tier (1, 2, or 3)","title":"Returns","type":"integer"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"log_audit_trail","doc":"Log all data accesses for compliance.\n\nAuditTrail ← UserID + Timestamp + ActionType","params":{"properties":{"action":{"description":"Action performed","title":"Action","type":"string"},"patient_id":{"description":"Patient ID accessed","title":"Patient Id","type":"string"}},"required":["action","patient_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"True if log successful","title":"Returns","type":"boolean"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_audit_trail","doc":"Retrieve audit trail for a patient.","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"}},"required":["patient_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of audit entries","items":{"type":"object"},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"transfer_to_triage_nurse","doc":"Transfer patient to human agent when:\n\n- Potential opioid abuse patterns detected\n- Insurance coverage disputes >$500\n- Emergency tier 1 situations","params":{"properties":{"summary":{"description":"Summary of patient issue","title":"Summary","type":"string"}},"required":["summary"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Transfer confirmation message","title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"verify_medical_emergency","doc":"Verify medical emergency exception using MD note to waive cancellation fees.","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"},"md_note":{"description":"Medical doctor's note describing emergency","title":"Md Note","type":"string"}},"required":["patient_id","md_note"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"True if emergency verified, False otherwise","title":"Returns","type":"boolean"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_contraindications","doc":"Check for contraindications in patient history for medication.","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"},"medication":{"description":"Medication to check","title":"Medication","type":"string"}},"required":["patient_id","medication"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of contraindications found (empty if none)","items":{"type":"string"},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"store_payment_method","doc":"Store payment method (insurance card + backup credit card) as required by policy.","params":{"properties":{"patient_id":{"description":"Unique patient identifier","title":"Patient Id","type":"string"},"method_type":{"description":"'insurance' or 'credit_card'","title":"Method Type","type":"string"},"details":{"additionalProperties":true,"description":"Payment method details","title":"Details","type":"object"}},"required":["patient_id","method_type","details"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"True if storage successful","title":"Returns","type":"boolean"}},"required":["returns"],"title":"returns","type":"object"}} diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/tools_4.jsonl b/conversational_tool_use_policy_tool_generation/golden_policies/tools_4.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..3c6d37d8820909f1882ec849a2c40706c40b3ab8 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/tools_4.jsonl @@ -0,0 +1,14 @@ +{"name": "place_order", "doc": "Create a new food delivery order", "params": {"properties": {"user_id": {"type": "string"}, "restaurants": {"items": {"properties": {"restaurant_id": {"type": "string"}, "items": {"items": {"properties": {"item_id": {"type": "string"}, "customizations": {"type": "string"}}, "required": ["item_id"]}, "type": "array"}}, "required": ["restaurant_id", "items"], "type": "object"}, "type": "array"}, "payment_methods": {"items": {"properties": {"payment_id": {"type": "string"}, "amount": {"type": "number"}}, "required": ["payment_id", "amount"]}, "type": "array"}, "delivery_address": {"properties": {"street": {"type": "string"}, "city": {"type": "string"}, "zip": {"type": "string"}}, "required": ["street", "city", "zip"], "type": "object"}, "loyalty_points": {"type": "integer"}, "dietary_flags": {"type": "array"}, "allergen_warning_acknowledged": {"type": "boolean"}}, "required": ["user_id", "restaurants", "payment_methods", "delivery_address", "allergen_warning_acknowledged"]}, "returns": {"properties": {"order": {"properties": {"order_id": {"type": "string"}, "status": {"enum": ["pending", "preparing", "en_route", "delivered"]}, "estimated_delivery": {"type": "string"}}, "type": "object"}}}} +{"name": "modify_order", "doc": "Modify an existing order within allowed time window", "params": {"properties": {"order_id": {"type": "string"}, "modifications": {"properties": {"items_add": {"type": "array"}, "items_remove": {"type": "array"}, "address_change": {"type": "object"}}, "type": "object"}, "modification_timestamp": {"type": "string"}}, "required": ["order_id", "modifications", "modification_timestamp"]}, "returns": {"properties": {"updated_order": {"type": "object"}, "time_remaining": {"type": "number"}}}} +{"name": "cancel_order", "doc": "Cancel an order and process refund based on policy.", "params": {"properties": {"order_id": {"type": "string"}, "reason": {"enum": ["user_request", "restaurant_closed", "delivery_issue"]}}, "required": ["order_id", "reason"]}, "returns": {"properties": {"refund_amount": {"type": "number"}, "loyalty_impact": {"type": "integer"}}}} +{"name": "get_order_status", "doc": "Retrieve current status and details of an order.", "params": {"properties": {"order_id": {"type": "string"}}, "required": ["order_id"]}, "returns": {"properties": {"status": {"type": "object"}}}} +{"name": "check_ingredient_availability", "doc": "Verify if restaurant can accommodate custom ingredient requests.", "params": {"properties": {"restaurant_id": {"type": "string"}, "item_id": {"type": "string"}, "requested_mods": {"type": "array"}}, "required": ["restaurant_id", "item_id"]}, "returns": {"properties": {"available": {"type": "boolean"}, "alternatives": {"type": "array"}}}} +{"name": "get_restaurant_menu", "doc": "Retrieve real-time menu for a restaurant", "params": {"properties": {"restaurant_id": {"type": "string"}}, "required": ["restaurant_id"]}, "returns": {"properties": {"menu": {"type": "object"}, "delivery_zones": {"type": "array"}, "temperature_control": {"type": "string"}, "last_updated": {"type": "string"}}}} +{"name": "verify_guest_order", "doc": "Authenticate guest order access.", "params": {"properties": {"confirmation_number": {"type": "string"}, "payment_digits": {"type": "string"}}, "required": ["confirmation_number", "payment_digits"]}, "returns": {"properties": {"order_details": {"type": "object"}}}} +{"name": "process_refund", "doc": "Handle refunds according to policy rules", "params": {"properties": {"order_id": {"type": "string"}, "reason": {"enum": ["missing_items", "temperature_violation", "late_delivery"]}, "evidence": {"type": "string"}, "refund_tier": {"enum": ["full", "partial", "none"]}, "items_ineligible": {"type": "array"}}, "required": ["order_id", "reason", "refund_tier"]}, "returns": {"properties": {"credit_amount": {"type": "number"}, "loyalty_impact": {"type": "integer"}}}} +{"name": "update_delivery_status", "doc": "Record delivery attempts and status changes.", "params": {"properties": {"order_id": {"type": "string"}, "attempt_number": {"type": "integer"}, "proof_photos": {"type": "array"}}, "required": ["order_id", "attempt_number"]}, "returns": {"properties": {"next_steps": {"type": "object"}}}} +{"name": "transfer_to_human_agents", "doc": "Escalate complex issues to human support.", "params": {"properties": {"issue_summary": {"type": "string"}}, "required": ["issue_summary"]}, "returns": {"properties": {"confirmation": {"type": "string"}}}} +{"name": "authenticate_user", "doc": "Verify user identity through specified authentication method", "params": {"properties": {"user_id": {"type": "string"}, "auth_method": {"enum": ["ID", "phone", "email"]}, "verification_code": {"type": "string"}}, "required": ["user_id", "auth_method", "verification_code"]}, "returns": {"properties": {"access_token": {"type": "string"}, "auth_status": {"enum": ["verified", "failed"]}}}} +{"name": "validate_payment_methods", "doc": "Ensure payment methods comply with policy requirements", "params": {"properties": {"user_id": {"type": "string"}, "order_total": {"type": "number"}, "payment_methods": {"items": {"properties": {"payment_id": {"type": "string"}, "amount": {"type": "number"}}, "required": ["payment_id", "amount"]}, "type": "array"}}, "required": ["user_id", "order_total", "payment_methods"]}, "returns": {"properties": {"valid": {"type": "boolean"}, "primary_method": {"type": "string"}, "loyalty_used": {"type": "number"}}}} +{"name": "handle_delivery_exceptions", "doc": "Manage delivery attempts and final disposition", "params": {"properties": {"order_id": {"type": "string"}, "attempt_number": {"type": "integer"}, "proof_photos": {"type": "array"}}, "required": ["order_id", "attempt_number"]}, "returns": {"properties": {"next_steps": {"type": "object"}, "fees_applied": {"type": "number"}}}} +{"name": "allergy_confirmation", "doc": "Handle mandatory allergy acknowledgment workflow", "params": {"properties": {"order_id": {"type": "string"}, "allergen_acknowledged": {"type": "boolean"}}, "required": ["order_id", "allergen_acknowledged"]}, "returns": {"properties": {"confirmation_status": {"type": "string"}, "auto_cancel": {"type": "boolean"}}}} \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/tools_5.jsonl b/conversational_tool_use_policy_tool_generation/golden_policies/tools_5.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..18c364562cc7c46e7cc393c6ae69b19f9b4487fd --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/tools_5.jsonl @@ -0,0 +1,14 @@ +{"name":"authenticate_user","doc":"Authenticate user using user ID and event confirmation code.\n\nRequired before any action that modifies user data or tickets.","params":{"properties":{"user_id":{"description":"Unique user identifier (e.g., 'user_12345')","title":"User Id","type":"string"},"event_confirmation_code":{"description":"Event-specific confirmation code (e.g., 'EVT-67890')","title":"Event Confirmation Code","type":"string"}},"required":["user_id","event_confirmation_code"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"True if authentication successful, False otherwise","title":"Returns","type":"boolean"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_event_details","doc":"Retrieve event details including seat map, ticket inventory, and restrictions.","params":{"properties":{"event_id":{"description":"Unique event identifier (e.g., 'concert-2024-06-25')","title":"Event Id","type":"string"}},"required":["event_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"name":{"description":"Event name","title":"Name","type":"string"},"date":{"description":"Event date (YYYY-MM-DD)","title":"Date","type":"string"},"time":{"description":"Event time (HH:MM:SS)","title":"Time","type":"string"},"venue_map":{"description":"Seat tier information","title":"Venue Map","type":"object"},"ticket_inventory":{"description":"Available tickets per section","title":"Ticket Inventory","type":"object"},"transfer_policy":{"description":"Transfer/resale restrictions","title":"Transfer Policy","type":"string"},"weather_plan":{"description":"Weather contingency details","title":"Weather Plan","type":"string"},"age_restriction":{"description":"\"18+\" or \"21+\" if applicable","title":"Age Restriction","type":"string"}},"required":["name","date","time","venue_map","ticket_inventory","transfer_policy","weather_plan"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_ticket_details","doc":"Retrieve ticket information including seat assignment and restrictions.","params":{"properties":{"ticket_id":{"description":"Unique ticket identifier (e.g., 'TKT-98765')","title":"Ticket Id","type":"string"}},"required":["ticket_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"event_id":{"description":"Associated event","title":"Event Id","type":"string"},"seat_section":{"description":"Seat section","title":"Seat Section","type":"string"},"seat_row":{"description":"Seat row","title":"Seat Row","type":"string"},"seat_number":{"description":"Seat number","title":"Seat Number","type":"string"},"price_tier":{"description":"Ticket price category","title":"Price Tier","type":"string"},"transferable":{"description":"True if transferable","title":"Transferable","type":"boolean"},"age_restricted":{"description":"Age requirement (18+/21+)","title":"Age Restricted","type":"string"},"resale_count":{"description":"Number of previous resales","title":"Resale Count","type":"integer"}},"required":["event_id","seat_section","seat_row","seat_number","price_tier","transferable","resale_count"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_seat_availability","doc":"Check available seats in a specific section for relocation/upgrade.","params":{"properties":{"event_id":{"description":"Unique event identifier","title":"Event Id","type":"string"},"section":{"description":"Seat section to check (e.g., 'Orchestra')","title":"Section","type":"string"}},"required":["event_id","section"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"available_seats":{"description":"List of available seat numbers","items":{"type":"string"},"title":"Available Seats","type":"array"},"adjacent_seats":{"description":"Groups of adjacent seats (for group discounts)","items":{"items":{"type":"string"},"type":"array"},"title":"Adjacent Seats","type":"array"}},"required":["available_seats","adjacent_seats"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"purchase_tickets","doc":"Purchase new tickets after authentication and seat availability check.","params":{"properties":{"user_id":{"description":"Authenticated user ID","title":"User Id","type":"string"},"event_id":{"description":"Event to purchase for","title":"Event Id","type":"string"},"ticket_count":{"description":"Number of tickets (1-10)","title":"Ticket Count","type":"integer"},"section":{"description":"Desired seat section","title":"Section","type":"string"},"payment_method_id":{"description":"Payment method from user profile","title":"Payment Method Id","type":"string"},"apply_group_discount":{"description":"Whether to apply group discount","title":"Apply Group Discount","type":"boolean"}},"required":["user_id","event_id","ticket_count","section","payment_method_id","apply_group_discount"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Confirmation number for new ticket order","title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"upgrade_ticket","doc":"Upgrade ticket to VIP section. Validates upgrade eligibility.","params":{"properties":{"ticket_id":{"description":"Original ticket ID","title":"Ticket Id","type":"string"},"new_section":{"description":"VIP section to upgrade to","title":"New Section","type":"string"},"payment_method_id":{"description":"Payment method for price difference + 15% fee","title":"Payment Method Id","type":"string"}},"required":["ticket_id","new_section","payment_method_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"New upgraded ticket ID","title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"relocate_seat","doc":"Relocate seat within same section/price tier.","params":{"properties":{"ticket_id":{"description":"Original ticket ID","title":"Ticket Id","type":"string"},"new_seat_number":{"description":"New seat identifier","title":"New Seat Number","type":"string"},"payment_method_id":{"description":"Payment method for fees (if downgrade)","title":"Payment Method Id","type":"string"}},"required":["ticket_id","new_seat_number","payment_method_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"New ticket ID with updated seat","title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"cancel_tickets","doc":"Cancel tickets and process refunds/credits based on policy.","params":{"properties":{"ticket_ids":{"description":"List of tickets to cancel","items":{"type":"string"},"title":"Ticket Ids","type":"array"},"reason":{"description":"Cancellation reason ('weather', 'user_request', etc.)","title":"Reason","type":"string"},"refund_type":{"title":"Refund Type"}},"required":["ticket_ids","reason","refund_type"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"refund_amount":{"description":"Monetary refund (if applicable)","title":"Refund Amount","type":"number"},"venue_credit":{"description":"Service credit amount (if applicable)","title":"Venue Credit","type":"number"},"refund_type":{"description":"Refund type ('full', 'partial', 'venue_credit')","enum":["full","partial","venue_credit"],"title":"Refund Type","type":"string"}},"required":["refund_type"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"transfer_ticket","doc":"Transfer ticket to another user with identity verification.","params":{"properties":{"ticket_id":{"description":"Ticket to transfer","title":"Ticket Id","type":"string"},"recipient_user_id":{"description":"User ID of recipient","title":"Recipient User Id","type":"string"},"recipient_id_verified":{"description":"Confirmation of ID verification","title":"Recipient Id Verified","type":"boolean"}},"required":["ticket_id","recipient_user_id","recipient_id_verified"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"New ticket ID for recipient","title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"reschedule_event","doc":"Reschedule event due to weather and process compensations.","params":{"properties":{"event_id":{"description":"Event to reschedule","title":"Event Id","type":"string"},"new_date":{"description":"New event date (YYYY-MM-DD)","title":"New Date","type":"string"}},"required":["event_id","new_date"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"new_event_id":{"description":"Rescheduled event ID","title":"New Event Id","type":"string"},"compensation":{"description":"25% venue credit amount","title":"Compensation","type":"number"}},"required":["new_event_id","compensation"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"transfer_to_human_agents","doc":"Transfer to human agent with issue summary. Use when:\n\n- User explicitly requests human agent\n- Request exceeds policy/tool capabilities","params":{"properties":{"summary":{"description":"Summary of user's issue","title":"Summary","type":"string"}},"required":["summary"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Confirmation of transfer","title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_user_profile","doc":"Retrieve user profile details including membership tier, payment methods, \n\nage verification status, and group discount eligibility.","params":{"properties":{"user_id":{"description":"Unique user identifier","title":"User Id","type":"string"}},"required":["user_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"properties":{"membership_tier":{"description":"Regular/Premium/VIP","enum":["Regular","Premium","VIP"],"title":"Membership Tier","type":"string"},"payment_methods":{"description":"List of payment IDs and types","items":{"type":"object"},"title":"Payment Methods","type":"array"},"age_verified":{"description":"Boolean","title":"Age Verified","type":"boolean"},"group_discount_eligible":{"description":"Boolean","title":"Group Discount Eligible","type":"boolean"}},"required":["membership_tier","payment_methods","age_verified","group_discount_eligible"],"title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"apply_group_discount","doc":"Apply 15% group discount to qualifying orders (8+ tickets, same section/type).","params":{"properties":{"order_id":{"description":"Ticket order ID","title":"Order Id","type":"string"}},"required":["order_id"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Discount amount applied","title":"Returns","type":"number"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"verify_age","doc":"Verify user age using government-issued ID scan.","params":{"properties":{"user_id":{"description":"User to verify","title":"User Id","type":"string"},"id_scan_data":{"description":"Base64-encoded ID scan","title":"Id Scan Data","type":"string"}},"required":["user_id","id_scan_data"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"True if verification successful","title":"Returns","type":"boolean"}},"required":["returns"],"title":"returns","type":"object"}} diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/tools_6.jsonl b/conversational_tool_use_policy_tool_generation/golden_policies/tools_6.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..75c4d1e9c051e62b1575ff14051663055ce58925 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/tools_6.jsonl @@ -0,0 +1,13 @@ +{"name":"vt_get_dns_resolution_object","doc":"This endpoint retrieves a Resolution object by its ID. A resolution object ID is made by appending the IP and the domain it resolves to together.\n\nDomain-IP resolutions. Resolution objects include the following attributes:\ndate: date when the resolution was made (UTC timestamp).\nhost_name: domain or subdomain requested to the resolver.\nhost_name_last_analysis_stats: last detection stats from the resolution's domain. Similar to the domains's last_analysis_stats attribute.\nip_address: IP address the domain was resolved to.\nip_address_last_analysis_stats: last detection stats from the resolution's IP address. Similar to the IP address' last_analysis_stats attribute.\nresolver: source of the resolution.","params":{"properties":{"id":{"description":"Resolution object ID","title":"Id","type":"string"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"}},"required":["id","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Resolution object with DNS resolution details","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_get_objects_related_to_ip_address","doc":"IP addresses have a number of relationships to other objects. This returns ALL objects that fit the relationship. Returns a list of URLs.","params":{"properties":{"ip":{"description":"IP address","title":"Ip","type":"string"},"relationship":{"description":"Relationship name (see the list of table). \n\nThe relationships are documented here:\n- comments: The comments for the IP address. Returns a list of comments.\n- communicating_files: Files that communicate with the IP address. Returns a list of files.\n- downloaded_files: Files downloaded from the IP address. VT Enterprise users only. Returns a list of files.\n- graphs: Graphs including the IP address. Returns a list of graphs.\n- historical_ssl_certificates: SSL certificates associated with the IP. Returns a list of SSL certificates.\n- historical_whois: WHOIS information for the IP address. Returns a list of Whois attributes.\n- related_comments: Community posted comments in the IP's related objects. Returns a list of comments.\n- related_references: Returns the references related to the IP address. Returns a list of References.\n- related_threat_actors: Threat actors related to the IP address. Returns a list of threat actors.\n- referrer_files: Files containing the IP address. Returns a list of Files.\n- resolutions: Resolves the IP addresses. Returns a list of resolutions.\n- urls: Returns a list of URLs related to the IP address.","title":"Relationship","type":"string"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"},"limit":{"description":"Maximum number of comments to retrieve","format":"int32","title":"Limit","type":"integer"},"cursor":{"description":"Continuation cursor","title":"Cursor","type":"string"}},"required":["ip","relationship","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of related objects","items":{},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_get_ip_address_report","doc":"Retrieve an IP address report. These reports condense all of the recent activity that VirusTotal has seen for the resource under consideration, as well as contextual information about it. This function specifically generates these reports using the IP address parameter.","params":{"properties":{"ip":{"description":"IP address","title":"Ip","type":"string"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"}},"required":["ip","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"IP address report with analysis results and context","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_add_votes_to_ip_address","doc":"With this function, you can post a vote for a given file. The body for the POST request must be the JSON representation of a vote object. Note however that you don't need to provide an ID for the object, as they are automatically generated for new votes. The verdict attribute must be either 'harmless' or 'malicious'.\n\nPlease ensure that the JSON object you provide conforms accurately to valid JSON standards.","params":{"properties":{"ip":{"description":"IP address","title":"Ip","type":"string"},"data":{"description":"Vote object, must conform to valid JSON standards and contain a 'verdict' attribute that is either 'harmless' or 'malicious'.","title":"Data","type":"object"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"}},"required":["ip","data","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Vote submission result","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_get_domain_report","doc":"Retrieves a domain report. These reports contain information regarding the domain itself that VirusTotal has collected.","params":{"properties":{"domain":{"description":"Domain name","title":"Domain","type":"string"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"}},"required":["domain","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Domain report with analysis results","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_get_comments_on_ip_address","doc":"Retrieves the comments on a provided IP address. Returns a list of Comment objects.","params":{"properties":{"ip":{"description":"IP address","title":"Ip","type":"string"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"},"limit":{"description":"Maximum number of comments to retrieve","format":"int32","title":"Limit","type":"integer"},"cursor":{"description":"Continuation cursor","title":"Cursor","type":"string"}},"required":["ip","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of Comment objects","items":{},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_add_comment_to_ip_address","doc":"With this function, you can post a comment for a given IP address. The body for the POST request must be the JSON representation of a comment object. Notice however that you don't need to provide an ID for the object, as they are automatically generated for new comments. However, please note that you will need to provide a valid data JSON for using this function.\n\nAny word starting with # in your comment's text will be considered a tag, and added to the comment's tag attribute.\n\nReturns a Comment object.","params":{"properties":{"ip":{"description":"IP address","title":"Ip","type":"string"},"data":{"description":"A comment object, must conform to valid JSON standards and contain the necessary attributes for a comment.","title":"Data","type":"object"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"}},"required":["ip","data","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Comment object that was created","title":"Returns","type":"object"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_get_object_descriptors_related_to_ip_address","doc":"This specifically returns related object's IDs (and context attributes, if any). Please note that this will not return all attributes.\n\nYou are expected to provide the relationship to the object you're interested in. The valid relationships are documented and include comments, communicating files, downloaded files (VT Enterprise users only), graphs, historical SSL certificates, historical whois, related comments, related references, related threat actors, referrer files, resolutions, and URLs.","params":{"properties":{"ip":{"description":"IP address","title":"Ip","type":"string"},"relationship":{"description":"Relationship name","title":"Relationship","type":"string"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"},"limit":{"description":"Maximum number of objects to retrieve","format":"int32","title":"Limit","type":"integer"},"cursor":{"description":"Continuation cursor","title":"Cursor","type":"string"}},"required":["ip","relationship","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of object descriptors with IDs and context attributes","items":{},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_get_objects_related_to_domain","doc":"Objects are a key concept in the VirusTotal API. Each object has an identifier and a type. Each object has an associated URL, and each domain is associated with objects. This function returns ALL of the objects related to the domain, based on the specified relationship.\n\nValid relationships include caa_records, cname_records, comments, communicating_files, downloaded_files, graphs, historical_ssl_certificates, historical_whois, immediate_parent, mx_records, ns_records, parent, referrer_files, related_comments, related_references, related_threat_actors, resolutions, soa_records, siblings, subdomains, urls, user_votes.","params":{"properties":{"domain":{"description":"Domain name","title":"Domain","type":"string"},"relationship":{"description":"Relationship name","title":"Relationship","type":"string"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"},"limit":{"description":"Maximum number of objects to retrieve","format":"int32","title":"Limit","type":"integer"},"cursor":{"description":"Continuation cursor","title":"Cursor","type":"string"}},"required":["domain","relationship","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of objects related to the domain","items":{},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_get_object_descriptors_related_to_domain","doc":"This function specifically returns related object's IDs (and context attributes, if any) but will not return all attributes. It returns objects relating to a domain based on the specified relationship. Valid relationships include caa_records, cname_records, comments, communicating_files, downloaded_files, graphs, historical_ssl_certificates, historical_whois, immediate_parent, mx_records, ns_records, parent, referrer_files, related_comments, related_references, related_threat_actors, resolutions, soa_records, siblings, subdomains, urls, and user_votes.","params":{"properties":{"domain":{"description":"Domain name","title":"Domain","type":"string"},"relationship":{"description":"Relationship name","title":"Relationship","type":"string"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"},"limit":{"description":"Maximum number of objects to retrieve","format":"int32","title":"Limit","type":"integer"},"cursor":{"description":"Continuation cursor","title":"Cursor","type":"string"}},"required":["domain","relationship","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of object descriptors related to the domain","items":{},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_get_comments_on_domain","doc":"This function will retrieve comments on a specified domain.","params":{"properties":{"domain":{"description":"Domain name","title":"Domain","type":"string"},"x_apikey":{"description":"Your API key","title":"X Apikey","type":"string"},"limit":{"description":"Maximum number of comments to retrieve","format":"int32","title":"Limit","type":"integer"},"cursor":{"description":"Continuation cursor","title":"Cursor","type":"string"}},"required":["domain","x_apikey"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of comments on the domain","items":{},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"vt_get_votes_on_ip_address","doc":"This function will retrieve votes on a provided IP address.","params":{"properties":{"ip":{"description":"IP address","title":"Ip","type":"string"}},"required":["ip"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"List of votes on the IP address","items":{},"title":"Returns","type":"array"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"transfer_to_human_agent","doc":"Transfer to human agent with issue summary.","params":{"properties":{"summary":{"description":"Summary of the issue","title":"Summary","type":"string"}},"required":["summary"],"title":"parameters","type":"object"},"returns":{"properties":{"returns":{"description":"Transfer confirmation message","title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/tools_7.jsonl b/conversational_tool_use_policy_tool_generation/golden_policies/tools_7.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..7641bf31d1178b49e5b42e9104332cfc4632f376 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/tools_7.jsonl @@ -0,0 +1,45 @@ +{"name":"verify_identity","description":"Verifies a customer's identity using their date of birth or SSN last four digits.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer to validate.","title":"User Id","type":"string"},"verification_type":{"description":"Credential to compare for verification ('dob' or 'ssn_last4').","enum":["dob","ssn_last4"],"title":"Verification Type","type":"string"},"verification_value":{"description":"Customer supplied value that must match the stored credential.","title":"Verification Value","type":"string"}},"required":["user_id","verification_type","verification_value"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"activate_account","description":"Reactivates a customer's inactive account when it has not been permanently closed.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the account owner.","title":"User Id","type":"string"},"account_id":{"description":"Identifier of the account that should be reactivated.","title":"Account Id","type":"string"}},"required":["user_id","account_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"deactivate_account","description":"Deactivates a customer's account to mitigate risk or honor a deactivation request.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the account owner.","title":"User Id","type":"string"},"account_id":{"description":"Identifier of the account that should be deactivated.","title":"Account Id","type":"string"},"reason":{"description":"Explanation supplied to the customer for the deactivation.","title":"Reason","type":"string"}},"required":["user_id","account_id","reason"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"open_account","description":"Opens a new banking account for an existing customer with an optional initial deposit.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer opening the account.","title":"User Id","type":"string"},"account_type":{"description":"Type of account to create (checking, savings, cd, or credit_card).","enum":["checking","savings","cd","credit_card"],"title":"Account Type","type":"string"},"initial_deposit":{"default":0.0,"description":"Amount of funds to seed the account with, if any.","title":"Initial Deposit","type":"number"}},"required":["user_id","account_type"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_account_status","description":"Retrieves current status details for the specified account across all customers.","parameters":{"properties":{"account_id":{"description":"Identifier of the account to inspect.","title":"Account Id","type":"string"}},"required":["account_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"freeze_account","description":"Freezes a customer's account to prevent activity while a risk or fraud concern is investigated.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the account owner.","title":"User Id","type":"string"},"account_id":{"description":"Identifier of the account that should be frozen.","title":"Account Id","type":"string"},"reason":{"description":"Explanation recorded for freezing the account.","title":"Reason","type":"string"}},"required":["user_id","account_id","reason"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"track_chargeback_status","description":"Retrieves current chargeback details for a specific dispute across all customers.","parameters":{"properties":{"dispute_id":{"description":"Identifier of the dispute whose status should be reviewed.","title":"Dispute Id","type":"string"}},"required":["dispute_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"resend_otp","description":"Resends a one-time passcode to the customer through the requested delivery channel.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer receiving the OTP.","title":"User Id","type":"string"},"channel":{"description":"Delivery method to use for the OTP ('sms', 'email', or 'push').","enum":["sms","email","push"],"title":"Channel","type":"string"}},"required":["user_id","channel"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"register_device","description":"Registers a customer's device so future access can be trusted and tracked.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer registering the device.","title":"User Id","type":"string"},"device_id":{"description":"Unique identifier presented by the device being registered.","title":"Device Id","type":"string"}},"required":["user_id","device_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"process_account_application","description":"Processes customer requests to open a new account or close an existing one.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the requesting customer.","title":"User Id","type":"string"},"account_type":{"description":"Type of account to open or close.","enum":["checking","savings","cd","credit_card"],"title":"Account Type","type":"string"},"action":{"default":"open","description":"Requested operation ('open' or 'close').","enum":["open","close"],"title":"Action","type":"string"},"initial_deposit":{"default":0.0,"description":"Funds to seed a new account when opening.","title":"Initial Deposit","type":"number"},"account_id":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Identifier of the account to close when applicable.","title":"Account Id"}},"required":["user_id","account_type"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"update_account_info","description":"Updates a customer's profile details across all accounts, including address, email, or contact info.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer to update.","title":"User Id","type":"string"},"update_type":{"description":"Category of information to change ('address', 'email', or 'contact').","enum":["address","email","contact"],"title":"Update Type","type":"string"},"address1":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Primary street address when updating address information.","title":"Address1"},"city":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"City associated with the customer's address update.","title":"City"},"state":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"State or province for the address update.","title":"State"},"zip_code":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Postal code for the address update.","title":"Zip Code"},"address2":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Secondary address line when applicable.","title":"Address2"},"country":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Country for the updated address, defaulting to USA when omitted.","title":"Country"},"email":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"New email address when updating email or contact information.","title":"Email"},"phone":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"New phone number when updating contact information.","title":"Phone"}},"required":["user_id","update_type"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"update_name","description":"Updates a customer's legal name after verifying that required documentation is provided.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer whose name will change.","title":"User Id","type":"string"},"new_first_name":{"description":"Replacement first name to record across accounts.","title":"New First Name","type":"string"},"new_last_name":{"description":"Replacement last name to record across accounts.","title":"New Last Name","type":"string"},"documentation_provided":{"default":false,"description":"Indicates whether the customer supplied proof of the name change.","title":"Documentation Provided","type":"boolean"}},"required":["user_id","new_first_name","new_last_name"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"transfer_funds","description":"Executes an intra-bank transfer or a withdrawal for a customer after validating inputs.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer initiating the transaction.","title":"User Id","type":"string"},"transaction_type":{"description":"Type of movement to perform ('transfer' or 'withdraw').","enum":["transfer","withdraw"],"title":"Transaction Type","type":"string"},"from_account_id":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Source account for transfers.","title":"From Account Id"},"to_account_id":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Destination account for transfers.","title":"To Account Id"},"account_id":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Account to withdraw from when performing withdrawals.","title":"Account Id"},"amount":{"anyOf":[{"type":"number"},{"type":"null"}],"default":null,"description":"Dollar amount to move or withdraw.","title":"Amount"},"withdrawal_method":{"default":"ATM","description":"Channel used for withdrawals (for example, ATM or teller code).","title":"Withdrawal Method","type":"string"},"branch_id":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Branch identifier when scheduling an in-person withdrawal pickup.","title":"Branch Id"}},"required":["user_id","transaction_type"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"create_money_transfer_with_explicit_kind","description":"Creates an outbound money transfer and performs compliance screening before submission.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer initiating the transfer.","title":"User Id","type":"string"},"from_account_id":{"description":"Source account that will fund the transfer.","title":"From Account Id","type":"string"},"to_account":{"description":"Destination account identifier or routing details.","title":"To Account","type":"string"},"amount":{"description":"Dollar amount to move to the destination account.","title":"Amount","type":"number"},"kind":{"description":"Transfer mechanism such as ACH, wire, or internal transfer.","enum":["ach","wire","billpay","rtp"],"title":"Kind","type":"string"}},"required":["user_id","from_account_id","to_account","amount","kind"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"validate_transfer_details","description":"Validates destination account and routing information before a transfer is initiated.","parameters":{"properties":{"to_account":{"description":"Destination account number or alias to verify.","title":"To Account","type":"string"},"ifsc_swift":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Optional routing code such as IFSC or SWIFT to validate.","title":"Ifsc Swift"}},"required":["to_account"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"cancel_transfer","description":"Cancels a customer's transfer when it has not yet completed or failed irreversibly.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer requesting cancellation.","title":"User Id","type":"string"},"transfer_id":{"description":"Identifier of the transfer to cancel.","title":"Transfer Id","type":"string"}},"required":["user_id","transfer_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_transfer_status","description":"Retrieves the status details for a specific transfer across all customers.","parameters":{"properties":{"transfer_id":{"description":"Identifier of the transfer to look up.","title":"Transfer Id","type":"string"}},"required":["transfer_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"run_compliance_check","description":"Performs a simplified compliance evaluation for a transfer or related transaction context.","parameters":{"properties":{"context_id":{"description":"Identifier used to seed the compliance decision.","title":"Context Id","type":"string"},"type":{"description":"Transaction category being evaluated.","title":"Type","type":"string"}},"required":["context_id","type"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"create_loan_application","description":"Submits a new loan application and records it with compliance metadata.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the applicant.","title":"User Id","type":"string"},"loan_type":{"description":"Product type requested by the customer.","enum":["auto","personal","mortgage","home_equity"],"title":"Loan Type","type":"string"},"amount":{"description":"Principal amount the customer wants to borrow.","title":"Amount","type":"number"},"term":{"description":"Loan term in months.","title":"Term","type":"integer"},"channel":{"description":"Channel through which the application is submitted.","enum":["online","phone","branch"],"title":"Channel","type":"string"}},"required":["user_id","loan_type","amount","term","channel"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"loan_eligibility_check","description":"Estimates a customer's eligibility for a loan by evaluating credit score and account history.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer to evaluate.","title":"User Id","type":"string"},"loan_type":{"description":"Loan product under consideration.","enum":["auto","personal","mortgage","home_equity"],"title":"Loan Type","type":"string"},"amount":{"description":"Desired loan amount for the eligibility assessment.","title":"Amount","type":"number"}},"required":["user_id","loan_type","amount"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_credit_score","description":"Retrieves the customer's stored credit score along with category context.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer whose credit score is requested.","title":"User Id","type":"string"}},"required":["user_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"validate_kyc","description":"Evaluates the completeness of a customer's KYC profile and highlights missing fields.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer being reviewed.","title":"User Id","type":"string"}},"required":["user_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"calculate_dti","description":"Calculates the customer's estimated debt-to-income ratio using known loans and credit balances.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer whose finances are evaluated.","title":"User Id","type":"string"}},"required":["user_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"request_missing_docs","description":"Sends a document request to the customer when additional paperwork is required.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer who must supply documents.","title":"User Id","type":"string"},"doc_list":{"description":"Comma-separated list of documents the customer needs to provide.","title":"Doc List","type":"string"}},"required":["user_id","doc_list"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"block_card","description":"Blocks a customer's card to stop transactions after loss, theft, or suspicious use.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the cardholder.","title":"User Id","type":"string"},"card_id":{"description":"Identifier of the card to block.","title":"Card Id","type":"string"},"reason":{"description":"Reason provided for the block request.","title":"Reason","type":"string"}},"required":["user_id","card_id","reason"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"mail_card","description":"Orders a new or replacement card to be mailed to the customer's verified address.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the recipient customer.","title":"User Id","type":"string"},"card_type":{"description":"Type of card to issue ('credit_card' or 'debit_card').","enum":["credit_card","debit_card"],"title":"Card Type","type":"string"},"card_brand":{"default":"visa","description":"Network brand to imprint on the card.","enum":["visa","mastercard","amex","discover"],"title":"Card Brand","type":"string"},"expedited":{"default":false,"description":"When True forces expedited shipping, otherwise standard shipping is used.","title":"Expedited","type":"boolean"},"replacement_for":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Existing card identifier when replacing a specific card.","title":"Replacement For"},"reason":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Explanation for the replacement; required when replacement_for is provided.","title":"Reason"}},"required":["user_id","card_type"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"apply_for_loan","description":"Submits a loan request for underwriting while validating core eligibility criteria.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the applicant.","title":"User Id","type":"string"},"loan_type":{"description":"Loan category being requested.","enum":["auto","personal","mortgage","home_equity","student","business"],"title":"Loan Type","type":"string"},"amount":{"description":"Principal amount requested by the customer.","title":"Amount","type":"number"},"term_months":{"description":"Repayment term in months.","title":"Term Months","type":"integer"},"interest_rate":{"default":5.5,"description":"Proposed interest rate for the product.","title":"Interest Rate","type":"number"},"repayment_account":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Optional account to use for repayments if approved.","title":"Repayment Account"}},"required":["user_id","loan_type","amount","term_months"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"add_dispute","description":"Files a dispute for a suspicious or incorrect transaction tied to an account or card.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer raising the dispute.","title":"User Id","type":"string"},"account_id":{"description":"Account or payment method associated with the disputed transaction.","title":"Account Id","type":"string"},"amount":{"description":"Dollar amount believed to be incorrect.","title":"Amount","type":"number"},"description":{"default":"","description":"Narrative describing the issue.","title":"Description","type":"string"},"transaction_date":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Date of the disputed transaction if known.","title":"Transaction Date"}},"required":["user_id","account_id","amount"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"update_credit_limit","description":"Reviews and processes a credit limit change request for a credit card account.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the cardholder.","title":"User Id","type":"string"},"account_id":{"description":"Credit account whose limit should change.","title":"Account Id","type":"string"},"new_limit":{"description":"Desired credit limit after the adjustment.","title":"New Limit","type":"number"},"reason":{"default":"customer_request","description":"Context for the change, such as customer request or risk review.","title":"Reason","type":"string"}},"required":["user_id","account_id","new_limit"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_card_status","description":"Returns the operational status and metadata for a customer's card.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the cardholder.","title":"User Id","type":"string"},"card_id":{"description":"Identifier of the card to inspect.","title":"Card Id","type":"string"}},"required":["user_id","card_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"unblock_card","description":"Re-enables a card that was previously blocked once it is safe to use again.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the cardholder.","title":"User Id","type":"string"},"card_id":{"description":"Identifier of the card to reactivate.","title":"Card Id","type":"string"}},"required":["user_id","card_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"renew_card","description":"Initiates renewal for a card that is approaching its expiration date.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the cardholder.","title":"User Id","type":"string"},"card_id":{"description":"Identifier of the card to renew.","title":"Card Id","type":"string"}},"required":["user_id","card_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_card_shipment","description":"Provides mock shipment tracking information for a recently issued card.","parameters":{"properties":{"card_id":{"description":"Identifier of the card whose shipment is being tracked.","title":"Card Id","type":"string"}},"required":["card_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_card_upgrade_options","description":"Lists available upgrade tiers and benefits for a customer's card.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the cardholder.","title":"User Id","type":"string"},"card_id":{"description":"Identifier of the card to evaluate for upgrades.","title":"Card Id","type":"string"}},"required":["user_id","card_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"upgrade_card","description":"Kicks off an upgrade request that moves the card to a higher rewards tier.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the cardholder.","title":"User Id","type":"string"},"card_id":{"description":"Identifier of the card to upgrade.","title":"Card Id","type":"string"},"new_tier":{"description":"Target tier for the upgraded card.","title":"New Tier","type":"string"}},"required":["user_id","card_id","new_tier"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"withdraw_cash_on_atm_or_mobile_app","description":"Processes a cash withdrawal through ATM, branch, or mobile app channels with limit checks.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer.","title":"User Id","type":"string"},"account_id":{"description":"Account to debit for the withdrawal.","title":"Account Id","type":"string"},"amount":{"description":"Dollar amount to withdraw.","title":"Amount","type":"number"},"channel":{"description":"Channel used for the withdrawal ('atm', 'branch', or 'mobile_app').","enum":["atm","branch","mobile_app"],"title":"Channel","type":"string"}},"required":["user_id","account_id","amount","channel"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"verify_pin","description":"Confirms whether the provided PIN matches the customer's stored ATM credential.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer to verify.","title":"User Id","type":"string"},"pin":{"description":"Four-digit PIN supplied by the customer.","title":"Pin","type":"string"}},"required":["user_id","pin"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"check_atm_status","description":"Returns mock operational status information for a specified ATM.","parameters":{"properties":{"atm_id":{"description":"Identifier of the ATM to query.","title":"Atm Id","type":"string"}},"required":["atm_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_user_details","description":"Returns the full customer profile including accounts, loans, and payment instruments.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer to retrieve.","title":"User Id","type":"string"}},"required":["user_id"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"get_product_info","description":"Provides descriptive information and rates for the specified banking product category.","parameters":{"properties":{"product_type":{"description":"Product category to summarize (checking, savings, cd, credit_card, or loan).","enum":["checking","savings","cd","credit_card","loan"],"title":"Product Type","type":"string"},"specific_product":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Optional sub-product identifier for finer detail.","title":"Specific Product"}},"required":["product_type"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"generate_documents","description":"Generates official banking documents such as statements or confirmations for a customer.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer requesting the document.","title":"User Id","type":"string"},"document_type":{"description":"Document category to produce (statement, confirmation, disclosure, tax_document).","enum":["statement","confirmation","disclosure","tax_document"],"title":"Document Type","type":"string"},"account_id":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Optional account identifier to scope the document.","title":"Account Id"},"date_range":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Optional date range to include in the document.","title":"Date Range"}},"required":["user_id","document_type"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"send_certificate","description":"Issues an official banking certificate and delivers it through the requested channel.","parameters":{"properties":{"user_id":{"description":"Banking database identifier of the customer requesting the certificate.","title":"User Id","type":"string"},"certificate_type":{"description":"Type of certificate to generate (account_opening, loan_payoff, balance_confirmation).","enum":["account_opening","loan_payoff","balance_confirmation"],"title":"Certificate Type","type":"string"},"account_id":{"anyOf":[{"type":"string"},{"type":"null"}],"default":null,"description":"Optional account or loan identifier referenced by the certificate.","title":"Account Id"},"delivery_method":{"default":"email","description":"Delivery channel for the certificate ('email', 'mail', or 'branch_pickup').","enum":["email","mail","branch_pickup"],"title":"Delivery Method","type":"string"}},"required":["user_id","certificate_type"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"calculate","description":"Evaluates a safe mathematical expression to assist with quick financial calculations.","parameters":{"properties":{"expression":{"description":"Arithmetic expression containing digits and basic operators.","title":"Expression","type":"string"},"calculation_type":{"default":"basic","description":"Label describing the calculation context presented to the user.","title":"Calculation Type","type":"string"}},"required":["expression"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"think","description":"Captures intermediate reasoning steps that the agent wants to note without acting.","parameters":{"properties":{"thought":{"description":"Narrative describing the agent's internal reasoning.","title":"Thought","type":"string"}},"required":["thought"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} +{"name":"transfer_to_human_agents","description":"Escalates the conversation to a human banker when the request requires manual handling.","parameters":{"properties":{"reason":{"description":"Explanation to share with the human agent about the escalation.","title":"Reason","type":"string"},"priority":{"default":"medium","description":"Urgency level for the escalation queue ('low', 'medium', 'high', or 'urgent').","enum":["low","medium","high","urgent"],"title":"Priority","type":"string"},"context":{"default":"","description":"Additional context or transcript snippet helpful to the human banker.","title":"Context","type":"string"}},"required":["reason"],"title":"parameters","type":"object"},"return_type":{"properties":{"returns":{"title":"Returns","type":"string"}},"required":["returns"],"title":"returns","type":"object"}} diff --git a/conversational_tool_use_policy_tool_generation/golden_policies/tools_8.jsonl b/conversational_tool_use_policy_tool_generation/golden_policies/tools_8.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..6a1eb4c93fa49fee1195c5db0cf4b9261864bd15 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/golden_policies/tools_8.jsonl @@ -0,0 +1,15 @@ +{"name": "search_parking_availability", "doc": "Check real-time parking availability at a specific location.", "params": {"properties": {"location_id": {"description": "Unique identifier for parking location", "type": "string"}, "date": {"description": "Date in YYYY-MM-DD format", "type": "string"}, "start_time": {"description": "Requested entry time in HH:MM format", "type": "string"}, "end_time": {"description": "Requested exit time in HH:MM format", "type": "string"}, "parking_type": {"enum": ["standard", "EV", "accessible"], "description": "Type of parking space required", "type": "string"}}, "required": ["location_id", "date", "start_time", "end_time", "parking_type"], "title": "parameters", "type": "object"}, "returns": {"properties": {"available_spaces": {"type": "integer"}, "base_rate": {"type": "number"}, "peak_surcharge": {"type": "number"}, "total_price": {"type": "number"}}, "required": ["available_spaces", "total_price"], "title": "returns", "type": "object"}} +{"name": "book_reservation", "doc": "Create a new parking reservation.", "params": {"properties": {"user_id": {"type": "string"}, "location_id": {"type": "string"}, "entry_exit_times": {"properties": {"entry": {"type": "string"}, "exit": {"type": "string"}}, "required": ["entry", "exit"], "type": "object"}, "vehicle_details": {"properties": {"license_plate": {"type": "string"}, "vehicle_type": {"enum": ["standard", "EV", "accessible"]}}, "required": ["license_plate"], "type": "object"}, "payment_method_id": {"type": "string"}}, "required": ["user_id", "location_id", "entry_exit_times", "vehicle_details", "payment_method_id"], "title": "parameters", "type": "object"}, "returns": {"properties": {"reservation_id": {"type": "string"}, "confirmation_number": {"type": "string"}, "total_charged": {"type": "number"}}, "required": ["reservation_id"], "title": "returns", "type": "object"}} +{"name": "modify_reservation", "doc": "Modify an existing parking reservation with time validation", "params": {"type": "object", "properties": {"reservation_id": {"type": "string"}, "new_times": {"type": "object"}, "new_vehicle": {"type": "object"}, "payment_method_id": {"type": "string"}, "current_system_time": {"type": "string"}}, "required": ["reservation_id", "current_system_time"], "title": "parameters"}, "returns": {"type": "object", "properties": {"new_total": {"type": "number"}, "price_difference": {"type": "number"}}, "required": ["new_total"], "title": "returns"}} +{"name": "cancel_reservation", "doc": "Cancel an existing parking reservation with fee calculation", "params": {"type": "object", "properties": {"reservation_id": {"type": "string"}, "entry_time": {"type": "string"}, "current_time": {"type": "string"}}, "required": ["reservation_id", "entry_time", "current_time"], "title": "parameters"}, "returns": {"type": "object", "properties": {"refund_amount": {"type": "number"}, "cancellation_fee": {"type": "number"}}, "required": ["refund_amount"], "title": "returns"}} +{"name": "get_reservation_details", "doc": "Retrieve detailed information about a specific reservation.", "params": {"properties": {"reservation_id": {"type": "string"}}, "required": ["reservation_id"], "title": "parameters", "type": "object"}, "returns": {"properties": {"status": {"enum": ["active", "completed", "cancelled"]}, "payment_details": {"type": "object"}, "vehicle_info": {"type": "object"}}, "required": ["status"], "title": "returns", "type": "object"}} +{"name": "calculate_parking_cost", "doc": "Calculate estimated parking costs with dynamic pricing", "params": {"type": "object", "properties": {"entry_time": {"type": "string"}, "exit_time": {"type": "string"}, "membership_tier": {"enum": ["regular", "premium", "VIP"]}, "parking_type": {"enum": ["standard", "EV", "accessible"]}, "is_event_day": {"type": "boolean"}}, "required": ["entry_time", "exit_time"], "title": "parameters"}, "returns": {"type": "object", "properties": {"base_cost": {"type": "number"}, "peak_charges": {"type": "number"}, "event_surcharge": {"type": "number"}, "discounts": {"type": "number"}, "total_cost": {"type": "number"}}, "required": ["total_cost"], "title": "returns"}} +{"name": "handle_payment_dispute", "doc": "Initiate and manage payment disputes with evidence validation", "params": {"type": "object", "properties": {"reservation_id": {"type": "string"}, "dispute_type": {"enum": ["overcharge", "service_failure", "system_error"]}, "evidence": {"type": "object"}, "evidence_timestamp": {"type": "string"}}, "required": ["reservation_id", "dispute_type", "evidence_timestamp"], "title": "parameters"}, "returns": {"type": "object", "properties": {"dispute_id": {"type": "string"}, "resolution_timeframe": {"type": "string"}}, "required": ["dispute_id"], "title": "returns"}} +{"name": "check_overstay_fees", "doc": "Calculate potential overstay penalties with grace periods", "params": {"type": "object", "properties": {"reservation_id": {"type": "string"}, "proposed_exit_time": {"type": "string"}, "membership_tier": {"enum": ["regular", "premium", "VIP"]}}, "required": ["reservation_id", "proposed_exit_time"], "title": "parameters"}, "returns": {"type": "object", "properties": {"grace_period": {"type": "string"}, "hourly_fee": {"type": "number"}, "total_penalty": {"type": "number"}}, "required": ["total_penalty"], "title": "returns"}} +{"name": "transfer_to_human_agents", "doc": "Transfer complex issues to human support.", "params": {"properties": {"issue_summary": {"type": "string"}}, "required": ["issue_summary"], "title": "parameters", "type": "object"}, "returns": {"properties": {"confirmation": {"type": "string"}}, "required": ["confirmation"], "title": "returns", "type": "object"}} +{"name": "update_payment_method", "doc": "Update payment method for a reservation.", "params": {"properties": {"reservation_id": {"type": "string"}, "new_payment_id": {"type": "string"}}, "required": ["reservation_id", "new_payment_id"], "title": "parameters", "type": "object"}, "returns": {"properties": {"update_status": {"type": "string"}}, "required": ["update_status"], "title": "returns", "type": "object"}} +{"name": "get_user_details", "doc": "Retrieve user profile information with payment limits", "params": {"type": "object", "properties": {"user_id": {"type": "string"}}, "required": ["user_id"], "title": "parameters"}, "returns": {"type": "object", "properties": {"membership_tier": {"enum": ["regular", "premium", "VIP"]}, "payment_methods": {"type": "array"}, "reservation_history": {"type": "array"}, "payment_reversals": {"type": "integer"}, "prepaid_expiry": {"type": "string"}}, "required": ["membership_tier"], "title": "returns"}} +{"name": "send_compensation", "doc": "Issue compensation with approval requirements", "params": {"type": "object", "properties": {"user_id": {"type": "string"}, "amount": {"type": "number"}, "reason": {"type": "string"}, "dispute_type": {"type": "string"}, "membership_tier": {"enum": ["regular", "premium", "VIP"]}, "manager_approval": {"type": "boolean"}}, "required": ["user_id", "amount", "manager_approval"], "title": "parameters"}, "returns": {"type": "object", "properties": {"confirmation_code": {"type": "string"}}, "required": ["confirmation_code"], "title": "returns"}} +{"name": "verify_user_identity", "doc": "Verify user identity through supported authentication methods", "params": {"type": "object", "properties": {"auth_method": {"enum": ["license_plate", "email_phone", "corporate"], "type": "string"}, "credentials": {"type": "object"}}, "required": ["auth_method", "credentials"], "title": "parameters"}, "returns": {"type": "object", "properties": {"verification_status": {"type": "boolean"}, "auth_token": {"type": "string"}}, "required": ["verification_status"], "title": "returns"}} +{"name": "calculate_ev_charging_fees", "doc": "Calculate EV charging costs and idle fees", "params": {"type": "object", "properties": {"reservation_id": {"type": "string"}, "charging_duration": {"type": "number"}}, "required": ["reservation_id", "charging_duration"], "title": "parameters"}, "returns": {"type": "object", "properties": {"energy_used": {"type": "number"}, "charging_cost": {"type": "number"}, "idle_fees": {"type": "number"}}, "required": ["charging_cost"], "title": "returns"}} +{"name": "manage_accessibility_requests", "doc": "Handle accessibility accommodations and service animal registration", "params": {"type": "object", "properties": {"user_id": {"type": "string"}, "accommodation_type": {"enum": ["ADA_extension", "voice_navigation"], "type": "string"}, "animal_cert": {"type": "string"}}, "required": ["user_id"], "title": "parameters"}, "returns": {"type": "object", "properties": {"accommodation_status": {"type": "string"}}, "required": ["accommodation_status"], "title": "returns"}} \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_against_goldens.txt b/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_against_goldens.txt new file mode 100644 index 0000000000000000000000000000000000000000..0c5e2f0d29a54ec0dd9dd59a50f69ae0b4fc6ddc --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_against_goldens.txt @@ -0,0 +1,20 @@ +Your job is to judge the policy.md and tools.jsonl for an automated agent. +Specifically, given a list of (policy, tools) pairs that are intended for use by a chatbot agent, choose the worst set of policy and tools that you believe will not work in a real-life setting. + +Setting: +1. policy.md + - The policy is what the automated chatbot agent bases their actions upon. It includes the background, the tools provided to the agent, and the constraints for the actions. +2. tools.jsonl + - The list of tools that the agent can use to interact with the underlying system to address the customer's needs. Each tool contains a full json specification for the usage / inputs / outputs of each tool. + +Your task is as follows: +1. Carefully read the policy.md and tools.jsonl pairs for all the provided chatbots. +2. Check for styling: + - If the policy is well-written, the content is cohesive and the style is consistent. + - If the tools specifications are complete, make sense in the context of the policy and have no errors. +3. Check for realism and functional value: + - If the policy and tools have real-life use cases and functionalities rather than being toy problems. + - If the policy and tools are adopted by an automated chatbot agent, if the agent will bring real value to a business and their customers. +4. Write your judgment between tags. For the judgment, Write the index of the single worst policy/tools pair. + +Make sure to deliberate deeply and provide resoning before providing a judgment. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_cohesion.txt b/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_cohesion.txt new file mode 100644 index 0000000000000000000000000000000000000000..0748e6435893ad53cae5f58e1838b731fc05708d --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_cohesion.txt @@ -0,0 +1,28 @@ +Your job is to judge the policy.md and tools.jsonl for an automated agent for the {domain} domain. +Specifically, you must make a judgment about how consistent and cohesive the policy and tools are to each other. + +Setting: +1. policy.md + - The policy is what the automated chatbot agent bases their actions upon. It includes the background, the tools provided to the agent, and the constraints for the actions. +2. tools.jsonl + - The list of tools that the agent can use to interact with the underlying system to address the customer's needs. Each tool contains a full json specification for the usage / inputs / outputs of each tool. + +Your task is as follows: +1. Carefully read the policy.md and tools.jsonl and understand their usage. +2. Check for inconsistencies: + - If there are references to tools in the policy that are not in the tools list. + - If there are tools in the tools list that are not referred to in the policy. + - If the description of the tool is consistent with its counterpart in the policy. +3. Check for cohesion: + - Even if the policy and tools are consistent to each other, they must have synergy. If the contents of the tools and the policy may cause confusion for the automated agent, there is low cohesion and the policy / tools are not a good match for each other. +4. Write your judgment between tags. Write 'true' if the policy.md and tools.jsonl is consistent and cohesive to each other, or 'false' if not. + +Make sure to deliberate deeply and provide resoning before providing a judgment. + + +{policy} + + + +{tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_cohesion_gold_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_cohesion_gold_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..0748e6435893ad53cae5f58e1838b731fc05708d --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_cohesion_gold_v1.txt @@ -0,0 +1,28 @@ +Your job is to judge the policy.md and tools.jsonl for an automated agent for the {domain} domain. +Specifically, you must make a judgment about how consistent and cohesive the policy and tools are to each other. + +Setting: +1. policy.md + - The policy is what the automated chatbot agent bases their actions upon. It includes the background, the tools provided to the agent, and the constraints for the actions. +2. tools.jsonl + - The list of tools that the agent can use to interact with the underlying system to address the customer's needs. Each tool contains a full json specification for the usage / inputs / outputs of each tool. + +Your task is as follows: +1. Carefully read the policy.md and tools.jsonl and understand their usage. +2. Check for inconsistencies: + - If there are references to tools in the policy that are not in the tools list. + - If there are tools in the tools list that are not referred to in the policy. + - If the description of the tool is consistent with its counterpart in the policy. +3. Check for cohesion: + - Even if the policy and tools are consistent to each other, they must have synergy. If the contents of the tools and the policy may cause confusion for the automated agent, there is low cohesion and the policy / tools are not a good match for each other. +4. Write your judgment between tags. Write 'true' if the policy.md and tools.jsonl is consistent and cohesive to each other, or 'false' if not. + +Make sure to deliberate deeply and provide resoning before providing a judgment. + + +{policy} + + + +{tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_cohesion_user.txt b/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_cohesion_user.txt new file mode 100644 index 0000000000000000000000000000000000000000..f243349f0b0b62620b50987310fd3f33e1aecaa5 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/judge_policy_tools_cohesion_user.txt @@ -0,0 +1,39 @@ +Your job is to judge the policy.md and tools.jsonl for an automated agent for the {domain} domain. +Specifically, you must make a judgment about how consistent and cohesive the policy and tools are to each other. + +Setting: +1. policy.md + - The policy is what the automated chatbot agent bases their actions upon. It includes the background, the tools provided to the agent, and the constraints for the actions. +2. tools.jsonl + - The list of tools that the agent can use to interact with the underlying system to address the customer's needs. Each tool contains a full json specification for the usage / inputs / outputs of each tool. + +Definitions: +- Agent Tool: an action an agent can take, such as authenticating a user, looking up commodities, applying patches, sending an upload link, verifying a payment, etc. +- User Tool: an action a user can take, such as checking a status, uploading a document, sending a payment, etc. +- The user cannot call a user tool unless prompted by the agent. +- The agent cannot call a user tool, but can prompt the user to call it. + +Your task is as follows: +1. Carefully read the policy.md and tools.jsonl and understand their usage. +2. Check for inconsistencies: + - If there are references to tools in the policy that are not in the tools list. + - If there are tools in the tools list that are not referred to in the policy. + - If the description of the tool is consistent with its counterpart in the policy. + - If an agent tool or user tool is misclassified and inconsistent with the directions in the policy. +3. Check for cohesion: + - Even if the policy and tools are consistent to each other, they must have synergy. If the contents of the tools and the policy may cause confusion for the automated agent, there is low cohesion and the policy / tools are not a good match for each other. +4. Write your judgment between tags. Write 'true' if the policy.md and tools.jsonl is consistent and cohesive to each other, or 'false' if not. + +Make sure to deliberate deeply and provide resoning before providing a judgment. + + +{policy} + + + +{agent_tools} + + + +{user_tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1.txt new file mode 100644 index 0000000000000000000000000000000000000000..2f7d75ef2bc32bc0c2b6e7d685e0f60caac535b5 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1.txt @@ -0,0 +1,35 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. + +Your task is as follows: +1. Carefully read the given policy and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to maintain usability of the policy. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism of a real customer service agent. +4. Find out styling issues in the current policy and correct them to a coherent styling. +5. Ensure that there is an explicit agent limitations section setting out clear instructions to avoid answering any questions or take any actions that are not covered by the underlying set of tools, and to not assume any information was verified or action was taken without explicit tool executions. +6. Ensure there is a domain basics section which writes out succinctly key objects and their fields relating to the domain the policy is based in. +7. Edit out / remove an explicit section for a list of tools. +8. Write the new policy.md file for the chatbot agent between tags. + +Make sure the policy contains at least: +- A few sections for different functionalities of the customer service agent (i.e. enrollment changes, booking modifications, etc). +- At least one functionality which requires proactive asking from the agent to the user (i.e. can you confirm if you received the invoice? (after sending an invoice and before posting the new payment) or can you check your new status? (after making a status change request and before executing next action)). + +If there is a section for verifying the identity of the user through a code confirmation sent through SMS / email, make sure the policy explicitly instructs the agent to: +1. Ask the user for their email / phone number +2. Call a code-sending tool to send an authentication code to the user at their email / number +3. Receive the code back from the user +4. Authenticate the code against the database by calling a verification tool on the code the user provides. + +Make sure the agent limitations section contains in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_gold_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_gold_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..1af49591dbc2496b87dd76ec444e64283d7aa60e --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_gold_v1.txt @@ -0,0 +1,27 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +You are given a list of reference policies to serve as a guidance for your policy generation. + +Your task is as follows: +1. Carefully read the reference policies and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to mimic the level of detail in the reference policies. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism in the reference policies. +4. Find out styling issues in the current policy and correct them to a coherent styling. +5. Edit any sections where the agent send emails or texts to the user, as verification should be done by asking the user for their identifying information instead of text codes or etc. +6. Ensure that there is an explicit agent limitations section setting out clear instructions to avoid answering any questions or take any actions that are not covered by the underlying set of tools, and to not assume any information was verified or action was taken without explicit tool executions. +7. Write the new policy.md file for the chatbot agent between tags. + +Make sure the agent limitations section contains in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + +{reference_policies} + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..e57ac041dc01d02e71b2c6dfc04f72831997fe20 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v1.txt @@ -0,0 +1,24 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain given the tools.jsonl. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +The provided tools.jsonl is the list of tools the agent can actually interact with. You must re-generate the policy.md file to reflect specific tools the agent can use. + +Your task is as follows: +1. Carefully read the tools.jsonl and check whether the policy contains exact references to the tools provided to the agent. +2. If the tool is present in the policy, make sure that the policy makes it clear how to use the tool. +3. If the tool is not present in the policy but has a reference to its functionality, edit the policy such that it accurately reflects the existence and positioning of the tool. +4. If the tools is not present in the policy and has no reference to its functionality, edit the policy such that it now includes the tool and its functionality in a way that does not majorly disrupt or change the original policy. +5. Write the new policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- Do not change the original policy too much. Your job is to edit the policy, not re-write it from scratch. +- Keep all original keywords and ideas from the policy, such as the timestamp, constraints, limitations, escalation policy, authentication policy, etc. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + + +{policy} + + + +{tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v2.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v2.txt new file mode 100644 index 0000000000000000000000000000000000000000..b777722505234d87e089643dd8948c2cde0ea425 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v2.txt @@ -0,0 +1,21 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +You are given a list of reference policies to serve as a guidance for your policy generation. + +Your task is as follows: +1. Carefully read the reference policies and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to mimic the level of detail in the reference policies. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism in the reference policies. +4. Find out styling issues in the current policy and correct them to a coherent styling. +5. Write the new policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + +{reference_policies} + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v3_not_enough_tools_to_verify_what_has_been_verified.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v3_not_enough_tools_to_verify_what_has_been_verified.txt new file mode 100644 index 0000000000000000000000000000000000000000..2402de0b5d7dae189184ad94f7086de274b4e346 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v3_not_enough_tools_to_verify_what_has_been_verified.txt @@ -0,0 +1,22 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +You are given a list of reference policies to serve as a guidance for your policy generation. + +Your task is as follows: +1. Carefully read the reference policies and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to mimic the level of detail in the reference policies. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism in the reference policies. +4. Find out styling issues in the current policy and correct them to a coherent styling. +5. Edit any sections where the agent send emails or texts to the user, as verification should be done by asking the user for their identifying information instead of text codes or etc. +6. Write the new policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + +{reference_policies} + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v4_model_still_hallucinates_info_not_in_policy.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v4_model_still_hallucinates_info_not_in_policy.txt new file mode 100644 index 0000000000000000000000000000000000000000..1af49591dbc2496b87dd76ec444e64283d7aa60e --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v4_model_still_hallucinates_info_not_in_policy.txt @@ -0,0 +1,27 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +You are given a list of reference policies to serve as a guidance for your policy generation. + +Your task is as follows: +1. Carefully read the reference policies and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to mimic the level of detail in the reference policies. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism in the reference policies. +4. Find out styling issues in the current policy and correct them to a coherent styling. +5. Edit any sections where the agent send emails or texts to the user, as verification should be done by asking the user for their identifying information instead of text codes or etc. +6. Ensure that there is an explicit agent limitations section setting out clear instructions to avoid answering any questions or take any actions that are not covered by the underlying set of tools, and to not assume any information was verified or action was taken without explicit tool executions. +7. Write the new policy.md file for the chatbot agent between tags. + +Make sure the agent limitations section contains in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + +{reference_policies} + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v5.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v5.txt new file mode 100644 index 0000000000000000000000000000000000000000..686c6fd33bb26b82d010d61df8f436b7fe2cf048 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v5.txt @@ -0,0 +1,29 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +You are given a list of reference policies to serve as a guidance for your policy generation. + +Your task is as follows: +1. Carefully read the reference policies and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to mimic the level of detail in the reference policies. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism in the reference policies. +4. Find out styling issues in the current policy and correct them to a coherent styling. +5. Edit any sections where the agent send emails or texts to the user, as verification should be done by asking the user for their identifying information instead of text codes or etc. +6. Ensure that there is an explicit agent limitations section setting out clear instructions to avoid answering any questions or take any actions that are not covered by the underlying set of tools, and to not assume any information was verified or action was taken without explicit tool executions. +7. Write the new policy.md file for the chatbot agent between tags. + +Make sure the agent limitations section contains in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities. +- Never assume verification or action completion without explicit tool confirmation. +- Do not assume any information about the company, service, or any underlying systems and entities. +- Never give general advice or directions. You are not an expert of the service, but a tool to help resolve simple issues well-defined by the policy. + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + +{reference_policies} + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v6.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v6.txt new file mode 100644 index 0000000000000000000000000000000000000000..6fb8554c7477eb57e5b821c71ccbcb882710627f --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v6.txt @@ -0,0 +1,30 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +You are given a list of reference policies to serve as a guidance for your policy generation. + +Your task is as follows: +1. Carefully read the reference policies and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to mimic the level of detail in the reference policies. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism in the reference policies. +4. Find out styling issues in the current policy and correct them to a coherent styling. +5. Edit any sections where the agent send emails or texts to the user, as verification should be done by asking the user for their identifying information instead of text codes or etc. +6. Ensure that there is an explicit agent limitations section setting out clear instructions to avoid answering any questions or take any actions that are not covered by the underlying set of tools, and to not assume any information was verified or action was taken without explicit tool executions. +7. Write the new policy.md file for the chatbot agent between tags. + +Make sure the agent limitations section contains in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities. +- Never assume verification or action completion without explicit tool confirmation. +- Do not assume any information about the company, service, or any underlying systems and entities. +- Never give general advice or directions. You are not an expert of the service, but a tool to help resolve simple issues well-defined by the policy. +- If the task falls outside of your functionality as written in the policy, transfer to a human agent. + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + +{reference_policies} + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v7.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v7.txt new file mode 100644 index 0000000000000000000000000000000000000000..bb3c5a109614d983a517584bc63e3a54dbd1e4b7 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v7.txt @@ -0,0 +1,28 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +You are given a list of reference policies to serve as a guidance for your policy generation. + +Your task is as follows: +1. Carefully read the reference policies and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to mimic the level of detail in the reference policies. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism in the reference policies. +4. Ensure there is at least one part of the policy that outlines a functionality requiring user action and agent verification, and add it / edit it if necessary. +5. Make sure there is sufficient information for the user to take actions in the policy, such as a tool to send a link for upload or specific directions for a website. +6. Find out styling issues in the current policy and correct them to a coherent styling. +7. Ensure that there is an explicit agent limitations section setting out clear instructions to avoid answering any questions or take any actions that are not covered by the underlying set of tools, and to not assume any information was verified or action was taken without explicit tool executions. +8. Write the new policy.md file for the chatbot agent between tags. + +Make sure the agent limitations section contains in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + +{reference_policies} + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v8.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v8.txt new file mode 100644 index 0000000000000000000000000000000000000000..fcae45853d5db8ebdaaecf2375d12c3d04ac9dce --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_1_v8.txt @@ -0,0 +1,29 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +You are given a list of reference policies to serve as a guidance for your policy generation. + +Your task is as follows: +1. Carefully read the reference policies and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to mimic the level of detail in the reference policies. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism in the reference policies. +4. Ensure there is at least one part of the policy that outlines a functionality requiring user action and agent verification, and add it / edit it if necessary. +5. Make sure there are clear instructions for the agent to ask a user to call a named user tool to take an action (i.e. upload_document, send_payment), as well instructions for the agent to call the complementary named agent tool for verification (i.e. validate_document, verify_payment). +6. If the policy states that the agent wants the user to click a link or should receive a message, it should have a corresponding agent tool so that the agent can explicitly start the process (i.e. send_link, etc). +7. Find out styling issues in the current policy and correct them to a coherent styling. +8. Ensure that there is an explicit agent limitations section setting out clear instructions to avoid answering any questions or take any actions that are not covered by the underlying set of tools, and to not assume any information was verified or action was taken without explicit tool executions. +9. Write the new policy.md file for the chatbot agent between tags. + +Make sure the agent limitations section contains in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + +{reference_policies} + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_user_1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_user_1.txt new file mode 100644 index 0000000000000000000000000000000000000000..d002dce33a1f154313f20909af8323e76a3bb213 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_refine_user_1.txt @@ -0,0 +1,29 @@ +Your job is to edit the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +You are given a list of reference policies to serve as a guidance for your policy generation + +Your task is as follows: +1. Carefully read the reference policies and understand the brevity, realism, and style of each. +2. Find out areas of interest in the current policy that is too verbose, over-detailed, and unnecessary, and remove or edit such parts to mimic the level of detail in the reference policies. +3. Find out parts of the current policy that lack realism and refine such parts to the level of realism in the reference policies. +4. Ensure there is at least one part of the policy that outlines a functionality requiring user action and agent verification, and add it / edit it if necessary. +5. Make sure there are clear instructions for the agent to ask a user to call a named user tool to take an action (i.e. upload_document, send_payment), as well instructions for the agent to call the complementary named agent tool for verification (i.e. validate_document, verify_payment). +6. If the policy states that the agent wants the user to click a link or should receive a message, it should have a corresponding agent tool so that the agent can explicitly start the process (i.e. send_link, etc). +7. Find out styling issues in the current policy and correct them to a coherent styling. +8. Ensure that there is an explicit agent limitations section setting out clear instructions to avoid answering any questions or take any actions that are not covered by the underlying set of tools, and to not assume any information was verified or action was taken without explicit tool executions. +9. Write the new policy.md file for the chatbot agent between tags. + +Make sure the agent limitations section contains in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +When writing the policy, consider: +- Keep all original keywords and ideas from the policy. + +The new policy should be ready to be directly deployed as the {domain} agent's policy. + +{reference_policies} + + +{policy} + diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample.txt new file mode 100644 index 0000000000000000000000000000000000000000..6b11605c2a2e666f6177015aebe50daa6365969d --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample.txt @@ -0,0 +1,36 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. Think of at least 10 tools that can be used by the policy. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- If there is a need for the customer's identity to be verified, mention it in the policy. Make sure to include a way to help users verify their identity using various different identifiers to help customers who are missing key information. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. +- Including the current time in the policy the model can refer to, which is {timestamp}. +- Adding a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Any limitations to the agent regarding rate limits or performance. +- Do not have the agent send emails or texts to the user, verification can be done by asking the user for their identifying information. +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. +- You do not need to tell the agent to be friendly or helpful, the agent will by default be firendly and helpful. +- Information about how to log sensitive information does not need to be written as it is not handled by the autonomous agent. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference.txt new file mode 100644 index 0000000000000000000000000000000000000000..1340c18e58b3b9be491a3b7acb721ae93bd66117 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference.txt @@ -0,0 +1,42 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. Think of at least 10 tools that can be used by the policy. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- Include the current time in the policy the model can refer to, which is {timestamp}. +- Add a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. + +Make sure the policy contains at least: +- A few sections for different functionalities of the customer service agent (i.e. enrollment changes, booking modifications, etc). +- At least one functionality which requires proactive asking from the agent to the user (i.e. can you confirm if you received the invoice? (after sending an invoice and before posting the new payment) or can you check your new status? (after making a status change request and before executing next action)). + +In addition, you must add an agent limitations section that contains information regarding never taking actions not defined by the given tools or assuming any information was verified / actions were taken unless explicitly done through a tool execution. +It should contain in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. +Ensure that the style, brevity, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_gold_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_gold_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..0db8b318b01f77bdf88c5d0c363f3030823f2222 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_gold_v1.txt @@ -0,0 +1,38 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. Think of at least 10 tools that can be used by the policy. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- Include the current time in the policy the model can refer to, which is {timestamp}. +- Add a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. + +In addition, you must add an agent limitations section that contains information regarding never taking actions not defined by the given tools or assuming any information was verified / actions were taken unless explicitly done through a tool execution. +It should contain in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. +Ensure that the style, brevity, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_user.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_user.txt new file mode 100644 index 0000000000000000000000000000000000000000..b72dcf3a97b68244cd8c4180afe98636bbc536e0 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_user.txt @@ -0,0 +1,50 @@ +Your job is to generate the policy.md file for an automated customer service agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +A reference policy and agent tools is provided to help show an example output for the task. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying an automated customer service agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the automated customer service agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. +4. Additionally, think of tools that can interact with the user and back: e.g. send information to users or verify actions that the user has taken. + - Sending information to the user: (e.g. sending invoice, updating user profile, etc) + - Verify actions taken by the user: (e.g. verify payment, etc.) + - Do not make this the verification action. Initial verification should be done without the need for 2FA. +5. Think of functionalities to put in the policy requiring this flow and make agent directions for such functionalities: + a. User has a problem. + b. Agent asks the user to take an action (send payment, upload a document, etc) with explicit directions (please send it through our app / website) or by calling a tool (send_upload_link, send_confirmation_code, etc). + c. User takes the action (assume the user has a set of tools they can call such as send_payment, upload_document, etc.) + d. Agent verifies the action has been taken. +6. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- The user cannot call a user tool unless prompted by the agent. +- The agent cannot call a user tool, but can prompt the user to call it. +- Include the current time in the policy the model can refer to, which is {timestamp}. +- Add a statement to avoid answering questions that the automated agent isn’t qualified to handle—especially those that require advice not based on the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. + +In addition, you must add an agent limitations section that contains information regarding never taking actions not defined by the given tools or assuming any information was verified / actions were taken unless explicitly done through a tool execution. +It should contain in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for information, constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be information and constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. +- Tools that need pointers to objects that the agent cannot find through other tools. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot agent and business as if this were a real policy that could be used immediately. +Ensure that the style, brevity, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..c076c9ea16410ff29d1a32e6c141b02f26a14f8b --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v1.txt @@ -0,0 +1,46 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. Think of at least 10 tools that can be used by the policy. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- If there is a need for the customer's identity to be verified, mention it in the policy. Make sure to include a way to help users verify their identity using various different identifiers to help customers who are missing key information. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. +- Including the current time in the policy the model can refer to, which is {timestamp}. +- Adding a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Any limitations to the agent regarding rate limits or performance. +- Do not have the agent send emails or texts to the user, verification can be done by asking the user for their identifying information. +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. +- You do not need to tell the agent to be friendly or helpful, the agent will by default be firendly and helpful. +- Information about how to log sensitive information does not need to be written as it is not handled by the autonomous agent. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. + + + +{reference_policy} + + + +{reference_tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v2_too_many_invalid_verify_tool_results.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v2_too_many_invalid_verify_tool_results.txt new file mode 100644 index 0000000000000000000000000000000000000000..a9cea8d0c83bbe8deeffcb313b58ee5a176bca17 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v2_too_many_invalid_verify_tool_results.txt @@ -0,0 +1,38 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. Think of at least 10 tools that can be used by the policy. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- If there is a need for the customer's identity to be verified, mention it in the policy. Only include this tool if it is absolutely necessary for the specified domain. Include a way to help users verify their identity using various different identifiers to help customers who are missing key information. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. +- Including the current time in the policy the model can refer to, which is {timestamp}. +- Adding a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Any limitations to the agent regarding rate limits or performance. +- Do not have the agent send emails or texts to the user, verification can be done by asking the user for their identifying information. +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. +- You do not need to tell the agent to be friendly or helpful, the agent will by default be firendly and helpful. +- Information about how to log sensitive information does not need to be written as it is not handled by the autonomous agent. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. +Ensure that the style, brevity, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v3_not_enough_tools_to_verify_what_has_been_verified.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v3_not_enough_tools_to_verify_what_has_been_verified.txt new file mode 100644 index 0000000000000000000000000000000000000000..ff764e6dffd5b7b1d62179cd69591cbd0585eeea --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v3_not_enough_tools_to_verify_what_has_been_verified.txt @@ -0,0 +1,33 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. Think of at least 10 tools that can be used by the policy. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. +- Including the current time in the policy the model can refer to, which is {timestamp}. +- Adding a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. +Ensure that the style, brevity, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v4_model_still_hallucinates_info_not_in_policy.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v4_model_still_hallucinates_info_not_in_policy.txt new file mode 100644 index 0000000000000000000000000000000000000000..0db8b318b01f77bdf88c5d0c363f3030823f2222 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v4_model_still_hallucinates_info_not_in_policy.txt @@ -0,0 +1,38 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. Think of at least 10 tools that can be used by the policy. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- Include the current time in the policy the model can refer to, which is {timestamp}. +- Add a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. + +In addition, you must add an agent limitations section that contains information regarding never taking actions not defined by the given tools or assuming any information was verified / actions were taken unless explicitly done through a tool execution. +It should contain in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. +Ensure that the style, brevity, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v5.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v5.txt new file mode 100644 index 0000000000000000000000000000000000000000..0d0cb7bca9800beb32220ca9fb67eed757908bdf --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v5.txt @@ -0,0 +1,40 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. Think of at least 10 tools that can be used by the policy. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- Include the current time in the policy the model can refer to, which is {timestamp}. +- Add a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. + +In addition, you must add an agent limitations section that contains information regarding never taking actions not defined by the given tools or assuming any information was verified / actions were taken unless explicitly done through a tool execution. +It should contain in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities. +- Never assume verification or action completion without explicit tool confirmation. +- Do not assume any information about the company, service, or any underlying systems and entities. +- Never give general advice or directions. You are not an expert of the service, but a tool to help resolve simple issues well-defined by the policy. + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. +Ensure that the style, brevity, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v6.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v6.txt new file mode 100644 index 0000000000000000000000000000000000000000..bb251ef782c103817d83d8781d67e36f0037ec46 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_reference_v6.txt @@ -0,0 +1,48 @@ +Your job is to generate the policy.md file for an automated customer service agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying an automated customer service agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the automated customer service agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. +4. Additionally, think of tools that can interact with the user and back: e.g. send information to users or verify actions that the user has taken. + - Sending information to the user: (e.g. sending invoice, updating user profile, etc) + - Verify actions taken by the user: (e.g. verify payment, etc.) + - Do not make this the verification action. Initial verification should be done without the need for 2FA. +5. Think of functionalities to put in the policy requiring this flow and make agent directions for such functionalities: + a. User has a problem. + b. Agent asks the user to take an action (send payment, upload a document, etc) with explicit directions (please send it through our app / website) or by calling a tool (send_upload_link, send_confirmation_code, etc). + c. User takes the action. + d. Agent verifies the action has been taken. +6. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- Include the current time in the policy the model can refer to, which is {timestamp}. +- Add a statement to avoid answering questions that the automated agent isn’t qualified to handle—especially those that require advice not based on the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. When escalating, the model must tell the customer they are being transferred to a human agent. + +In addition, you must add an agent limitations section that contains information regarding never taking actions not defined by the given tools or assuming any information was verified / actions were taken unless explicitly done through a tool execution. +It should contain in the minimum these core values: +- Do not answer questions or perform actions outside defined tool capabilities +- Never assume verification or action completion without explicit tool confirmation + +You must also include the following sections: +1. A section for basic components and objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for information, constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be information and constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. +- Tools that need pointers to objects that the agent cannot find through other tools. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot agent and business as if this were a real policy that could be used immediately. +Ensure that the style, brevity, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_test.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_test.txt new file mode 100644 index 0000000000000000000000000000000000000000..489c9462e99736f477f33a4e5a3ed7acbf2402f7 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_test.txt @@ -0,0 +1,36 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- If there is a need for the customer's identity to be verified, mention it in the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. +- Including the current time in the policy the model can refer to, which is {timestamp}. +- Adding a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. + +You must also include the following sections: +1. A section for objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Any limitations to the agent regarding rate limits or performance. +- Do not have the agent send emails or texts to the user, verification can be done by asking the user for their identifying information. +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. +- You do not need to tell the agent to be friendly or helpful, the agent will by default be firendly and helpful. +- Information about how to log sensitive information does not need to be written as it is not handled by the autonomous agent. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..2f252fe0bc86249ce4e4b2bf7d5d6ef0e0938f2a --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v1.txt @@ -0,0 +1,23 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- If there is a need for the customer's identity to be verified, mention it in the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. +- Including a place for the datetime in the policy, where the current time is {timestamp}. + +Do not include the following in the policy: +- Two-factor authentication or any actions that ask for the user to provide a code or take an action. +- Any limitations to the agent regarding rate limits or performance. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v2.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v2.txt new file mode 100644 index 0000000000000000000000000000000000000000..0d31dfd9d90b0d5849d65a7b2990c7bc8ce00a7a --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v2.txt @@ -0,0 +1,35 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- If there is a need for the customer's identity to be verified, mention it in the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. +- Including the current time in the policy the model can refer to, which is {timestamp}. + +You must also include the following sections: +1. A section for objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Any limitations to the agent regarding rate limits or performance. +- Do not have the agent send emails or texts to the user, verification can be done by asking the user for their identifying information. +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. +- You do not need to tell the agent to be friendly or helpful, the agent will by default be firendly and helpful. +- Information about how to log sensitive information does not need to be written as it is not handled by the autonomous agent. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v3.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v3.txt new file mode 100644 index 0000000000000000000000000000000000000000..489c9462e99736f477f33a4e5a3ed7acbf2402f7 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v3.txt @@ -0,0 +1,36 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- If there is a need for the customer's identity to be verified, mention it in the policy. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. +- Including the current time in the policy the model can refer to, which is {timestamp}. +- Adding a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. + +You must also include the following sections: +1. A section for objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Any limitations to the agent regarding rate limits or performance. +- Do not have the agent send emails or texts to the user, verification can be done by asking the user for their identifying information. +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. +- You do not need to tell the agent to be friendly or helpful, the agent will by default be firendly and helpful. +- Information about how to log sensitive information does not need to be written as it is not handled by the autonomous agent. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v4.txt b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v4.txt new file mode 100644 index 0000000000000000000000000000000000000000..8d890b99e7d96a041ef498ce47c95f29141b6ec3 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/policy_sample_v4.txt @@ -0,0 +1,36 @@ +Your job is to generate the policy.md file for an automated agent for the {domain} domain. +Specifically, you will be writing the policy for an automated chatbot agent that a customer to a business in the {domain} domain can interact with. + +Your task is as follows: +1. Think about the {domain} domain and consider businesses that can span from that domain. +2. Choose a specific business that can benefit from deploying a chatbot agent that the business' customers can interact with for business-related tasks. +3. Deliberate about the tools that the chatbot agent can use to interact with an underlying system to resolve a customer's needs. Your tools can be in camel case or snake case. +4. Create a policy.md file for the chatbot agent between tags. + +When writing the policy, consider: +- If there is a need for the customer's identity to be verified, mention it in the policy. Make sure to include a way to help users verify their identity using various different identifiers to help customers who are missing key information. +- Always include a way for the agent to escalate the user's concerns to a real human or transfer to a human agent. +- Including the current time in the policy the model can refer to, which is {timestamp}. +- Adding a statement to avoid answering questions that the chatbot isn’t qualified to handle—especially those that require advice not based on the policy. + +You must also include the following sections: +1. A section for objects relevant to the policy: + - For example, the idea of a user (e.g. with name, email, etc) and business-specific objects (e.g. reservation for retail/airline/etc, tickets for service, etc.) + - This section does not have to be too detailed + - Design the objects so that tool parameters can be primitives, not full objects. +2. Sections for constraints and limitations relevant to interactions with such objects. + - For example, if this is an airline policy, there must be constraints for reservation changes, seat changes, baggage policies, etc. + - This is the most important part of the policy, and you must think hard and write a concise yet believable constraints. + +Do not include the following in the policy: +- Any limitations to the agent regarding rate limits or performance. +- Do not have the agent send emails or texts to the user, verification can be done by asking the user for their identifying information. +- Tools that call other generative models such as pdfs / images / videos should not be added nor considered. +- You do not need to tell the agent to be friendly or helpful, the agent will by default be firendly and helpful. +- Information about how to log sensitive information does not need to be written as it is not handled by the autonomous agent. + +The policy.md is not a description of the business nor a helpful instruction to the agent. +Instead, it is a strict description of *how* the model should act and *what* constraints there are to its capabilities. + +Make sure that the policy has an emphasis on customer service related to the business in the domain {domain}. +Do not make the generated policy sound like an example template. Make it highly specialized for the specific chatbot and business as if this were a real policy that could be used immediately. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1.txt new file mode 100644 index 0000000000000000000000000000000000000000..d7ae90db918b908df30daf231688821e66f1d58b --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1.txt @@ -0,0 +1,27 @@ +Your job is to edit the tools.jsonl file for the {domain} domain given the policy.md. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +You are given a list of reference tools to serve as a guidance for your tool generation. + +Your task is as follows: +1. Carefully read the provided policy.md and existing tools.jsonl and check if the original tools fit naturally for the use cases of the policy. +2. If a tool is referenced in the policy but omitted in the tools.jsonl, draft a new tool specification to be provided to the agent based on the description in the policy. +3. If an existing tool's execution result must be followed up with another tool, make sure that follow-up tool exists. If it does not exist, create the tool. +4. If the user can take any action based on the agent's directions as per the policy, make sure that agent's directions are grounded on a tool that exists. If it does not exist, create the tool. +5. Write the edited tools.jsonl file for the chatbot agent between tags. + +When writing the the tools list, consider: +1. Do not change the original tools list too much unless needed. +2. For existing tools, if they should not be edited, leave them as they are. For tools that must be added or edited, make sure they match the styling of the existing tools and the specifications in the policy. +3. Make sure to create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +4. Return types must be explicit and simple. Do not use 'returns' as a return type property name. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the policy. + + +{policy} + + + +{tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1_gold_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1_gold_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..d7ae90db918b908df30daf231688821e66f1d58b --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1_gold_v1.txt @@ -0,0 +1,27 @@ +Your job is to edit the tools.jsonl file for the {domain} domain given the policy.md. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +You are given a list of reference tools to serve as a guidance for your tool generation. + +Your task is as follows: +1. Carefully read the provided policy.md and existing tools.jsonl and check if the original tools fit naturally for the use cases of the policy. +2. If a tool is referenced in the policy but omitted in the tools.jsonl, draft a new tool specification to be provided to the agent based on the description in the policy. +3. If an existing tool's execution result must be followed up with another tool, make sure that follow-up tool exists. If it does not exist, create the tool. +4. If the user can take any action based on the agent's directions as per the policy, make sure that agent's directions are grounded on a tool that exists. If it does not exist, create the tool. +5. Write the edited tools.jsonl file for the chatbot agent between tags. + +When writing the the tools list, consider: +1. Do not change the original tools list too much unless needed. +2. For existing tools, if they should not be edited, leave them as they are. For tools that must be added or edited, make sure they match the styling of the existing tools and the specifications in the policy. +3. Make sure to create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +4. Return types must be explicit and simple. Do not use 'returns' as a return type property name. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the policy. + + +{policy} + + + +{tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..b9af7efe7b52551b1fb6c4e849b02c72827addec --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1_v1.txt @@ -0,0 +1,33 @@ +Your job is to edit the tools.jsonl file for the {domain} domain given the policy.md. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. + +Definitions: +- Agent Tool: an action an agent can take, such as authenticating a user, looking up commodities, applying patches, sending an upload link, verifying a payment, etc. +- User Tool: an action a user can take, such as checking a status, uploading a document, sending a payment, etc. + +Your task is as follows: +1. Carefully read the provided policy.md and existing tools.jsonl and check if there is a bijective mapping between tools referenced in the policy and tools currently available. +2. For each tool omitted in the tools.jsonl but referenced in the policy.md, draft a new tool specification to be provided to the agent based on the description in the policy. +3. For each tool existing in the tools.jsonl but unreferenced in the policy.md, remove that tool from the tools.jsonl. +4. Deliberate edits for the tools.jsonl based on the above criteria. +5. Write the edited tools.jsonl file for the chatbot agent between tags. + +When writing the the tools list, consider: +1. Do not change the orignal tools list too much. Your job is the edit the tools list, not re-write it from scratch. +2. For existing tools, if they should not be edited, leave them as they are. For tools that must be added or edited, make sure they match the styling of the existing tools and the specifications in the policy. +3. Make sure to create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the policy. + + +{policy} + + + +{agent_tools} + + + +{user_tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1_v2_not_enough_tools_to_verify_what_has_been_verified.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1_v2_not_enough_tools_to_verify_what_has_been_verified.txt new file mode 100644 index 0000000000000000000000000000000000000000..69d3d198e4e94c07aa3862ee38838daab505e101 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_1_v2_not_enough_tools_to_verify_what_has_been_verified.txt @@ -0,0 +1,25 @@ +Your job is to edit the tools.jsonl file for the {domain} domain given the policy.md. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +You are given a list of reference tools to serve as a guidance for your tool generation. + +Your task is as follows: +1. Carefully read the provided policy.md and existing tools.jsonl and check if the original tools fit naturally for the use cases of the policy. +2. If a tool is referenced in the policy but omitted in the tools.jsonl, draft a new tool specification to be provided to the agent based on the description in the policy. +3. Write the edited tools.jsonl file for the chatbot agent between tags. + +When writing the the tools list, consider: +1. Do not change the original tools list too much unless needed. +2. For existing tools, if they should not be edited, leave them as they are. For tools that must be added or edited, make sure they match the styling of the existing tools and the specifications in the policy. +3. Make sure to create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +4. Return types must be explicit and simple. Do not use 'returns' as a return type property name. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the policy. + + +{policy} + + + +{tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_user_1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_user_1.txt new file mode 100644 index 0000000000000000000000000000000000000000..e36859cd156404db803255a4c860e8a1807a2e6d --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_refine_user_1.txt @@ -0,0 +1,39 @@ +Your job is to edit the tools.jsonl file for the {domain} domain given the policy.md and the original tools.jsonl files. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +You are given the original agent and user tools to serve as a guidance for your tool editing. + +Definitions: +- Agent Tool: an action an agent can take, such as authenticating a user, looking up commodities, applying patches, sending an upload link, verifying a payment, etc. +- User Tool: an action a user can take, such as checking a status, uploading a document, sending a payment, etc. +- The user cannot call a user tool unless prompted by the agent. +- The agent cannot call a user tool, but can prompt the user to call it. + +Your task is as follows: +1. Carefully read the provided policy.md and existing tools.jsonl and check if the original tools fit naturally for the use cases of the policy. +2. If a tool is referenced in the policy but omitted in the tools.jsonl, draft a new tool specification to be provided to the agent based on the description in the policy. +3. If an existing tool's execution result must be followed up with another tool, make sure that follow-up tool exists. If it does not exist, create the tool. +4. If the user can take any action based on the agent's directions as per the policy, make sure that agent's directions are grounded on a tool that exists. If it does not exist, create the tool. +5. If the agent can ask the user to take an action (e.g. upload a file or send a payment), make sure there are appropriate tools to verify the user's actions. +6. For any user tool and agent tool that is misclassified, move them and edit them as needed. +7. Write the edited agent tools.jsonl file for the chatbot agent between tags and the user actions as user_tools.jsonl file between tags + +When writing the the tools list, consider: +1. Do not change the original tools list too much unless needed. +2. For existing tools, if they should not be edited, leave them as they are. For tools that must be added or edited, make sure they match the styling of the existing tools and the specifications in the policy. +3. Make sure to create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +4. Return types must be explicit and simple. Do not use 'returns' as a return type property name. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all agent tools listed in the policy. + + +{policy} + + + +{agent_tools} + + + +{user_tools} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample.txt new file mode 100644 index 0000000000000000000000000000000000000000..6753e2afa6bf6518275d12931504ee60406afe8c --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample.txt @@ -0,0 +1,20 @@ +Your job is to generate the tools.jsonl file for the {domain} domain. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. + +Your task is as follows: +1. Read and think about the provided policy and list out the tools that are mentioned in the policy. +2. Deliberate about the input properties and the return types of the tools. +4. Devise a description and docstring for each tool, as it will be provided to the agent to to be used for interaction with a downstream system. +5. Create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +6. Write the tools.jsonl file for the chatbot agent between tags. + +In addition to the tools in the policy, consider: +1. Always adding a verification / escalation tool if the policy mentions the need for one. +2. If you believe that there are implicit tasks described in the policy that do not have specific references to tool names, you can add additional tools with careful deliberation. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the original policy. + + +{policy} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference.txt new file mode 100644 index 0000000000000000000000000000000000000000..acab783eaa741a7a94883ba3d5f500893ae7dc0b --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference.txt @@ -0,0 +1,21 @@ +Your job is to generate the tools.jsonl file for the {domain} domain. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Read and think about the provided policy and list out the tools that are mentioned in the policy or should exist to support the agent's actions in the policy. +2. Deliberate about the input properties and the return types of the tools. +4. Devise a description and docstring for each tool, as it will be provided to the agent to to be used for interaction with a downstream system. +5. Create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +6. Write the tools.jsonl file for the chatbot agent between tags. + +In addition to the tools in the policy, consider: +1. The created tools must cover all use cases in the policy. The agent should not be able to perform any concrete action without the use of tools. +2. For instance, if verification is done by sending the user a verification code, make sure there is a tool to send the verification code to the user (after asking the user for their email / phone number) and there is a tool to verify the code the user mentions after. +3. Likewise, all actions that the policy can make the user take must be grounded on the agent's use of tools. +4. Always add an escalation tool, which should be simple and only include a parameter for a summary. +5. Return types must be explicit and simple. Do not use 'returns' as a return type property name. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the original policy. +Ensure that the style, detail, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_gold_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_gold_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..7b673c2c335d2d96fedb2902cf95978b01319af8 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_gold_v1.txt @@ -0,0 +1,20 @@ +Your job is to generate the tools.jsonl file for the {domain} domain. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Read and think about the provided policy and list out the tools that are mentioned in the policy or should exist to support the agent's actions in the policy. +2. Deliberate about the input properties and the return types of the tools. +4. Devise a description and docstring for each tool, as it will be provided to the agent to to be used for interaction with a downstream system. +5. Create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +6. Write the tools.jsonl file for the chatbot agent between tags. + +In addition to the tools in the policy, consider: +1. The created tools must cover all use cases in the policy. The agent should not be able to perform any concrete action without the use of tools. +2. Likewise, all actions that the policy can make the user take must be grounded on the agent's use of tools. +3. Always add an escalation tool, which should be simple and only include a parameter for a summary. +4. Return types must be explicit and simple. Do not use 'returns' as a return type property name. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the original policy. +Ensure that the style, detail, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_user.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_user.txt new file mode 100644 index 0000000000000000000000000000000000000000..c29530dfad929a67b50f88b3d435a79f7cc14a79 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_user.txt @@ -0,0 +1,27 @@ +Your job is to generate the tools.jsonl file for the {domain} domain. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +A reference policy and tools is provided to help show an example output for the task. + +Definitions: +- Agent Tool: an action an agent can take, such as authenticating a user, looking up commodities, applying patches, sending an upload link, verifying a payment, etc. +- User Tool: an action a user can take, such as checking a status, uploading a document, sending a payment, etc. +- The user cannot call a user tool unless prompted by the agent. +- The agent cannot call a user tool, but can prompt the user to call it. + +Your task is as follows: +1. Read and think about the provided policy and list out the tools that are mentioned in the policy or should exist to support the agent's actions in the policy. +2. Deliberate about the input properties and the return types of the tools. +3. If the agent can ask the user to take an action (e.g. upload a file or send a payment), make sure there are appropriate tools to verify the user's actions (e.g. check uploaded document, verify payment). +4. Devise a description and docstring for each tool, as it will be provided to the agent to to be used for interaction with a downstream system. +5. Create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +6. Write the agent tools.jsonl file for the chatbot agent between tags and the user actions as user_tools.jsonl file between tags + +In addition to the tools in the policy, consider: +1. The created tools must cover all use cases in the policy. The agent should not be able to perform any concrete action without the use of tools. +2. Likewise, all actions that the policy can make the user take must be grounded on the agent's use of tools. +3. Always add an escalation tool, which should be simple and only include a parameter for a summary. +4. Return types must be explicit and simple. Do not use 'returns' as a return type property name. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the original policy. +Ensure that the style, detail, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_v1.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_v1.txt new file mode 100644 index 0000000000000000000000000000000000000000..a620ff2627f591be1c6d5cc9df6b1488fe7cf29f --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_v1.txt @@ -0,0 +1,31 @@ +Your job is to generate the tools.jsonl file for the {domain} domain. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Read and think about the provided policy and list out the tools that are mentioned in the policy. +2. Deliberate about the input properties and the return types of the tools. +4. Devise a description and docstring for each tool, as it will be provided to the agent to to be used for interaction with a downstream system. +5. Create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +6. Write the tools.jsonl file for the chatbot agent between tags. + +In addition to the tools in the policy, consider: +1. Always adding a verification / escalation tool if the policy mentions the need for one. +2. If you believe that there are implicit tasks described in the policy that do not have specific references to tool names, you can add additional tools with careful deliberation. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the original policy. + + + +{reference_policy} + + + +{reference_tools} + + + + +{policy} + \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_v2_too_many_invalid_verify_tool_results.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_v2_too_many_invalid_verify_tool_results.txt new file mode 100644 index 0000000000000000000000000000000000000000..ec83e023a069802984ac086edfcf3a28a8f08880 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_v2_too_many_invalid_verify_tool_results.txt @@ -0,0 +1,18 @@ +Your job is to generate the tools.jsonl file for the {domain} domain. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Read and think about the provided policy and list out the tools that are mentioned in the policy. +2. Deliberate about the input properties and the return types of the tools. +4. Devise a description and docstring for each tool, as it will be provided to the agent to to be used for interaction with a downstream system. +5. Create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +6. Write the tools.jsonl file for the chatbot agent between tags. + +In addition to the tools in the policy, consider: +1. Always add an escalation tool, which should be simple and only include a parameter for a summary. +2. If adding a verification tool, make the return type a boolean only instead of an object. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the original policy. +Ensure that the style, detail, and realism of the generated policy be akin to the reference policies. \ No newline at end of file diff --git a/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_v3_not_enough_tools_to_verify_what_has_been_verified.txt b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_v3_not_enough_tools_to_verify_what_has_been_verified.txt new file mode 100644 index 0000000000000000000000000000000000000000..1ec2595a4f203726c6278bdfedcf6047d04d3476 --- /dev/null +++ b/conversational_tool_use_policy_tool_generation/prompt_history/tools_sample_reference_v3_not_enough_tools_to_verify_what_has_been_verified.txt @@ -0,0 +1,18 @@ +Your job is to generate the tools.jsonl file for the {domain} domain. +Specifically, you will be writing the tools specification for an automated chatbot agent that assists a customer based on the given policy. +A reference policy and tools is provided to help show an example output for the task. + +Your task is as follows: +1. Read and think about the provided policy and list out the tools that are mentioned in the policy. +2. Deliberate about the input properties and the return types of the tools. +4. Devise a description and docstring for each tool, as it will be provided to the agent to to be used for interaction with a downstream system. +5. Create a functional json specification of the tools with correct formatting. Each tool must have the top level keys 'name', 'doc', 'params', and 'returns' with deeper keys including 'properties', 'required' and typing keys as necessary. +6. Write the tools.jsonl file for the chatbot agent between tags. + +In addition to the tools in the policy, consider: +1. Always add an escalation tool, which should be simple and only include a parameter for a summary. +2. Return types must be explicit and simple. Do not use 'returns' as a return type property name. + +Do not write the actual code for tools. Only generate the tool specifications in the form of a jsonl. +You must include all tools listed in the original policy. +Ensure that the style, detail, and realism of the generated policy be akin to the reference policies. \ No newline at end of file