Refactoring gestion Agent

README.md

@@ -157,7 +157,7 @@ curl -X POST "http://localhost:7860/completion" \
   -d '{
     "message": "Bonjour, comment vas-tu?",
     "model": "gpt-4o",
-    "agent_type": "simple",
+    "agent": "V2",
     "stream": false,
     "temperature": 0.7
   }'
@@ -333,13 +333,13 @@ def create_custom_graph(llm):
 from graphs.custom_graph import create_custom_graph
 
 agent_registry.register_agent(
-    AgentType.CUSTOM,
+    "my_agent",
     create_custom_graph,
     "Description de votre agent"
 )
 ```
 
-3. Utilisez-le via l'API sans changement de code!
+3. Utilisez-le via l'API avec `"agent": "my_agent"` dans la requête `POST /completion`.
 
 ## 🧪 Tests
 
@@ -370,6 +370,7 @@ LANGCHAIN_PROJECT=routeur-ia
 
 ## 📝 TODO / Roadmap
 
+- [ ] Harmoniser la nomenclature du pipeline voix: le champ `agent_type` est encore utilisé dans `services/voice/voice_pipeline.py` pour des événements internes et devra être renommé en `agent` lors d'un prochain refactoring.
 - [ ] Tests unitaires et d'intégration
 - [ ] Implémentation complète WebRTC avec aiortc
 - [ ] Agent RAG avec base vectorielle

api/routes/completion.py

@@ -61,7 +61,7 @@ async def complete(
     - Content-Type: `text/event-stream`
     
     Args:
-        request: Completion request with message, model, agent type, and streaming flag
+        request: Completion request with message, model, agent and streaming flag
         current_user: Authenticated user (JWT required)
         
     Returns:
@@ -86,7 +86,7 @@ async def complete(
             return await _complete(request)
             
     except ValueError as e:
-        # Agent type not available or validation error
+        # Agent not available or validation error
         raise HTTPException(
             status_code=status.HTTP_400_BAD_REQUEST,
             detail=str(e)
@@ -112,7 +112,7 @@ async def _complete(request: CompletionRequest) -> CompletionResponse:
     result = await agent_service.invoke(
         message=request.message,
         model_name=request.model,
-        agent_type=request.agent_type,
+        agent=request.agent,
         temperature=request.temperature,
         max_tokens=request.max_tokens,
         conversation_history=request.conversation_history,
@@ -139,7 +139,7 @@ async def _stream_completion(request: CompletionRequest) -> StreamingResponse:
             async for chunk in agent_service.stream(
                 message=request.message,
                 model_name=request.model,
-                agent_type=request.agent_type,
+                agent=request.agent,
                 temperature=request.temperature,
                 max_tokens=request.max_tokens,
                 conversation_history=request.conversation_history,

docs/API_EXAMPLES.md

@@ -9,6 +9,11 @@
 5. [WebSocket](#websocket)
 6. [Exemples avancés](#exemples-avancés)
 
+## Point d'attention
+
+- La completion HTTP utilise désormais `agent` (ex: `V1`, `V2`).
+- Le pipeline voix conserve temporairement un champ interne nommé `agent_type` dans les événements (`services/voice/voice_pipeline.py`), pour compatibilité. Ce point est prévu pour un alignement ultérieur vers `agent`.
+
 ## Authentification
 
 ### Obtenir un token JWT
@@ -60,7 +65,7 @@ curl -X POST http://localhost:7860/completion \
   -d '{
     "message": "Explique-moi la théorie de la relativité en 2 phrases",
     "model": "gpt-4o",
-    "agent_type": "simple",
+    "agent": "V2",
     "stream": false,
     "temperature": 0.7
   }'
@@ -71,7 +76,7 @@ curl -X POST http://localhost:7860/completion \
 {
   "response": "La théorie de la relativité d'Einstein comprend deux parties: la relativité restreinte (1905) qui établit que la vitesse de la lumière est constante et que le temps et l'espace sont relatifs, et la relativité générale (1915) qui décrit la gravitation comme une courbure de l'espace-temps causée par la masse et l'énergie. Ces théories ont révolutionné notre compréhension de l'univers et sont confirmées par de nombreuses expériences.",
   "model": "gpt-4o",
-  "agent_type": "simple",
+  "agent": "V2",
   "usage": {
     "prompt_tokens": 25,
     "completion_tokens": 98,
@@ -99,15 +104,15 @@ curl -N -X POST http://localhost:7860/completion \
 
 **Réponse (Server-Sent Events):**
 ```
-data: {"content": "Il", "done": false, "metadata": {"model": "gpt-3.5-turbo", "agent_type": "simple"}}
+data: {"content": "Il", "done": false, "metadata": {"model": "gpt-3.5-turbo", "agent": "V2"}}
 
-data: {"content": " était", "done": false, "metadata": {"model": "gpt-3.5-turbo", "agent_type": "simple"}}
+data: {"content": " était", "done": false, "metadata": {"model": "gpt-3.5-turbo", "agent": "V2"}}
 
-data: {"content": " une", "done": false, "metadata": {"model": "gpt-3.5-turbo", "agent_type": "simple"}}
+data: {"content": " une", "done": false, "metadata": {"model": "gpt-3.5-turbo", "agent": "V2"}}
 
 ...
 
-data: {"content": "", "done": true, "metadata": {"model": "gpt-3.5-turbo", "agent_type": "simple"}}
+data: {"content": "", "done": true, "metadata": {"model": "gpt-3.5-turbo", "agent": "V2"}}
 ```
 
 #### Champs d'empreinte carbone, latence, pricing et équivalences
@@ -146,7 +151,7 @@ Les réponses incluent désormais des métriques d'impact carbone calculées ave
   "done": true,
   "metadata": {
     "model": "mistral-large-latest",
-    "agent_type": "simple",
+    "agent": "V2",
     "usage": {"input_tokens":123, "output_tokens":456, "total_tokens":579},
     "usage_by_model": {
       "mistral-large-latest": {"input_tokens":123, "output_tokens":456, "total_tokens":579}
@@ -324,19 +329,19 @@ curl -X GET http://localhost:7860/agents \
 {
   "agents": [
     {
-      "type": "simple",
-      "name": "Simple",
-      "description": "Simple conversational agent without tools or memory",
+      "type": "V1",
+      "name": "V1",
+      "description": "Current production orchestrated workflow",
       "available": true
     },
     {
-      "type": "rag",
-      "name": "Rag",
-      "description": "Agent with Retrieval Augmented Generation (not yet implemented)",
-      "available": false
+      "type": "V2",
+      "name": "V2",
+      "description": "Isolated V2 workflow (default)",
+      "available": true
     }
   ],
-  "total": 4
+  "total": 2
 }
 ```
 

domain/models.py

@@ -2,7 +2,7 @@
 from pydantic import BaseModel, Field
 from typing import Optional, List, Dict, Any, Literal
 from datetime import datetime, timezone
-from .enums import ModelName, AgentType
+from .enums import ModelName
 
 
 # ============ Auth Models ============
@@ -27,7 +27,10 @@ class CompletionRequest(BaseModel):
     """Request for text completion."""
     message: str = Field(..., description="User message to complete")
     model: ModelName = Field(default=ModelName.MISTRAL_LARGE, description="LLM model to use")
-    agent_type: AgentType = Field(default=AgentType.SIMPLE, description="Agent type to use")
+    agent: Optional[str] = Field(
+        default=None,
+        description="Agent identifier to use (ex: 'V1' or 'V2'). If omitted, defaults to 'V2'."
+    )
     stream: bool = Field(default=False, description="Enable streaming response")
     temperature: float = Field(default=0.7, ge=0.0, le=2.0, description="Sampling temperature")
     max_tokens: Optional[int] = Field(default=None, description="Maximum tokens to generate")
@@ -48,7 +51,7 @@ class CompletionResponse(BaseModel):
     """Response for text completion (non-streaming)."""
     response: str
     model: str
-    agent_type: str
+    agent: Optional[str] = None
     usage: Optional[Dict[str, Any]] = None
     metadata: Optional[Dict[str, Any]] = None
     conversation_id: Optional[str] = Field(
@@ -114,7 +117,7 @@ class ModelsListResponse(BaseModel):
 
 class AgentInfo(BaseModel):
     """Information about an available agent."""
-    type: AgentType
+    type: str
     name: str
     description: str
     available: bool = True

graphs/README.md

@@ -1,63 +0,0 @@
-# LangGraph Graphs
-
-Ce dossier contient les différents graphes LangGraph utilisés par l'API.
-
-## Structure
-
-- `base_graph.py`: Graphe conversationnel simple par défaut
-- Vous pouvez ajouter d'autres graphes personnalisés ici
-
-## Comment créer un nouveau graphe
-
-1. Créez un nouveau fichier Python dans ce dossier (ex: `custom_graph.py`)
-2. Définissez votre `State` avec TypedDict
-3. Créez vos fonctions de nœuds
-4. Construisez le graphe avec `StateGraph`
-5. Compilez le graphe avec `.compile()`
-6. Enregistrez votre graphe dans `services/agent_registry.py`
-
-## Exemple de graphe personnalisé
-
-```python
-from typing import TypedDict, Annotated, Sequence
-from langchain_core.messages import BaseMessage
-from langgraph.graph import StateGraph, END
-from langgraph.graph.message import add_messages
-
-class CustomState(TypedDict):
-    messages: Annotated[Sequence[BaseMessage], add_messages]
-    custom_field: str
-
-def create_custom_graph(llm):
-    def custom_node(state: CustomState):
-        # Votre logique personnalisée
-        messages = state["messages"]
-        response = llm.invoke(messages)
-        return {"messages": [response]}
-    
-    workflow = StateGraph(CustomState)
-    workflow.add_node("custom", custom_node)
-    workflow.set_entry_point("custom")
-    workflow.add_edge("custom", END)
-    
-    return workflow.compile()
-```
-
-## Graphes disponibles
-
-### Simple Graph (`base_graph.py`)
-- Graphe conversationnel basique
-- Prend un message, l'envoie au LLM, retourne la réponse
-- Pas de mémoire persistante
-
-### Simple Graph with History (`base_graph.py`)
-- Graphe conversationnel avec support de l'historique
-- Utilise l'historique fourni dans la requête
-- Pas de mémoire persistante (stateless)
-
-## Notes
-
-- Tous les graphes sont stateless par défaut
-- L'historique de conversation doit être fourni par le client dans chaque requête
-- Pour ajouter des outils (RAG, recherche web, etc.), créez un nouveau graphe personnalisé
-

graphs/base_graph.py

@@ -1,193 +0,0 @@
-"""Simple base LangGraph for conversational agent."""
-from typing import TypedDict, Annotated, Sequence, List, Optional
-from langchain_core.messages import BaseMessage, HumanMessage, AIMessage, SystemMessage
-from langchain_core.language_models.chat_models import BaseChatModel
-from langchain_core.documents import Document
-from langgraph.graph import StateGraph, END
-from langgraph.graph.message import add_messages
-from .prompts import SYSTEM_PROMPT_TEMPLATE
-
-# RAG imports (reuse setup from knowledge/ocr.ipynb)
-import os
-from functools import lru_cache
-from supabase import create_client, Client
-from langchain_openai import OpenAIEmbeddings
-from langchain_community.vectorstores import SupabaseVectorStore
-
-
-class AgentState(TypedDict, total=False):
-    """State for the conversational agent with RAG."""
-    messages: Annotated[Sequence[BaseMessage], add_messages]
-    query: Optional[str]
-    formation_docs: List[Document]
-    prestation_docs: List[Document]
-    formation_context: str
-    prestation_context: str
-    project_docs: List[Document]
-    project_context: str
-
-
-def create_simple_graph(llm: BaseChatModel):
-    """
-    Create a simple conversational graph with LangGraph.
-    
-    This is a basic graph that takes a message, sends it to the LLM,
-    and returns the response. It can be easily replaced with more complex graphs.
-    
-    Args:
-        llm: Language model to use for generation
-        
-    Returns:
-        Compiled LangGraph
-    """
-
-    def call_model(state: AgentState) -> AgentState:
-        """Call the LLM with the current messages."""
-        print(f"Calling model with messages: {state['messages']}")
-        
-        messages = state["messages"]
-        response = llm.invoke(messages)
-        return {"messages": messages + [AIMessage(content=response.content)] }
-    
-    # Build the graph
-    workflow = StateGraph(AgentState)
-    
-    # Add nodes
-    workflow.add_node("agent", call_model)
-    
-    # Set entry point
-    workflow.set_entry_point("agent")
-    
-    # Add edge to end
-    workflow.add_edge("agent", END)
-    
-    # Compile and return
-    return workflow.compile()
-
-
-def create_simple_graph_with_history(llm: BaseChatModel):
-    """
-    Create a conversational graph with history + RAG retrieval from Supabase.
-    Entry -> retrieve (RAG) -> agent (generate) -> END
-    """
-
-    @lru_cache(maxsize=2)
-    def _get_retriever(doc_type: str, k: int = int(os.getenv("RAG_TOP_K", "5"))):
-        """Get retriever for specific document type (formation or prestation)."""
-        url = os.getenv("SUPABASE_URL")
-        key = (
-            os.getenv("SUPABASE_KEY")
-            or os.getenv("SUPABASE_SERVICE_ROLE_KEY")
-            or os.getenv("SUPABASE_ANON_KEY")
-            or os.getenv("NEXT_PUBLIC_SUPABASE_ANON_KEY")
-        )
-        if not url or not key:
-            raise ValueError("SUPABASE_URL and a SUPABASE_*KEY env var are required.")
-        client: Client = create_client(url, key)
-        vector_store = SupabaseVectorStore(
-            embedding=OpenAIEmbeddings(api_key=os.getenv("OPENAI_API_KEY")),
-            client=client,
-            table_name=os.getenv("SUPABASE_TABLE", "documents"),
-            query_name=os.getenv("SUPABASE_MATCH_FN", "match_documents"),
-        )
-        return vector_store.as_retriever(search_kwargs={"k": k, "filter": {"type": doc_type}})
-
-
-    def _format_docs(docs: List[Document], doc_type: str, max_chars_per_doc: int = 1200) -> str:
-        """Format documents with type-specific formatting."""
-        blocks = []
-        for i, doc in enumerate(docs, 1):
-            text = (doc.page_content or "")[:max_chars_per_doc]
-            meta = doc.metadata or {}
-            src = meta.get("source", "N/A")
-            page = meta.get("page_number", "N/A")
-            kind = meta.get("type", "N/A")
-            contact = meta.get("contact", None)
-            header = f"[{i}] source={src} page={page} type={kind}"
-            if contact:
-                header += f" contact={contact}"
-            blocks.append(f"<document>\n{header}\n{text}</document>".strip())
-        return "\n\n---\n\n".join(blocks)
-
-
-    def retrieve(state: AgentState) -> AgentState:
-        """Separate retriever node: builds query, fetches docs for both types, formats context."""
-        # Get query from state or last human message
-        q = state.get("query")
-        if not q:
-            q = ""
-            for msg in reversed(list(state.get("messages", []))):
-                if getattr(msg, "type", "") == "human":
-                    q = (msg.content or "").strip()
-                    break
-        
-        # Get retrievers for both types
-        formation_retriever = _get_retriever("formation", k=8)
-        prestation_retriever = _get_retriever("prestation", k=8)
-        
-        # Retrieve documents for both types
-        formation_docs = formation_retriever.invoke(q or "")
-        prestation_docs = prestation_retriever.invoke(q or "")
-        
-        # Format contexts for both types
-        formation_context = _format_docs(formation_docs, "formation")
-        prestation_context = _format_docs(prestation_docs, "prestation")
-        
-        return {
-            "formation_docs": formation_docs,
-            "prestation_docs": prestation_docs,
-            "formation_context": formation_context,
-            "prestation_context": prestation_context
-        }
-
-
-    def call_model_with_history(state: AgentState) -> AgentState:
-        """Generation node: SYSTEM + RAG context + conversation."""
-        messages = list(state.get("messages", []))
-        sys_msgs: List[BaseMessage] = [SystemMessage(content=SYSTEM_PROMPT_TEMPLATE)]
-        
-        # Get both contexts
-        formation_context = state.get("formation_context", "")
-        prestation_context = state.get("prestation_context", "")
-        
-        # Add formation context if available
-        if formation_context:
-            sys_msgs.append(SystemMessage(content=(
-                "CONTEXTE FORMATIONS (extraits du catalogue formations; n'utilise rien d'autre):\n\n"
-                f"{formation_context}\n\n"
-                "Consignes formations: Utilise exclusivement ce contexte pour recommander les formations. "
-                "Cite la page et la source pour chaque recommandation. "
-                "Une formation = un document."
-            )))
-        
-        # Add prestation context if available
-        if prestation_context:
-            sys_msgs.append(SystemMessage(content=(
-                "CONTEXTE PRESTATIONS (extraits du catalogue services; n'utilise rien d'autre):\n\n"
-                f"{prestation_context}\n\n"
-                "Consignes prestations: Utilise exclusivement ce contexte pour recommander les prestations. "
-                "Cite la page et la source pour chaque recommandation. "
-                "Un document peut contenir plusieurs prestations."
-            )))
-        
-        response = llm.invoke(sys_msgs + messages)
-        return {"messages": messages + [AIMessage(content=response.content)]}
-
-
-    # Build the graph
-    workflow = StateGraph(AgentState)
-
-    # Add nodes
-    workflow.add_node("retrieve", retrieve)
-    workflow.add_node("agent", call_model_with_history)
-
-    # Set entry point
-    workflow.set_entry_point("retrieve")
-
-    # Add edges
-    workflow.add_edge("retrieve", "agent")
-    workflow.add_edge("agent", END)
-
-    # Compile and return
-    return workflow.compile()
-

graphs/workflows/conversation.py

@@ -1,24 +0,0 @@
-"""Conversation workflow: retrieve -> chat_agent -> END."""
-from langgraph.graph import StateGraph, END
-from langchain_core.language_models.chat_models import BaseChatModel
-
-from graphs.state import AgentState
-from graphs.nodes.retrieval import retrieve_both_types
-from graphs.agents.chat_agent import chat_node
-
-
-def create_conversation_graph(llm: BaseChatModel):
-    workflow = StateGraph(AgentState)
-
-    # Nodes
-    workflow.add_node("retrieve", retrieve_both_types)
-    workflow.add_node("agent", chat_node(llm))
-
-    # Entry and edges
-    workflow.set_entry_point("retrieve")
-    workflow.add_edge("retrieve", "agent")
-    workflow.add_edge("agent", END)
-
-    return workflow.compile()
-
-

graphs/workflows/{conversation_with_summary.py → orchestrated_v2.py}

@@ -1,20 +1,32 @@
-"""Workflow: retrieve -> chat_agent -> summarizer -> END."""
+"""Orchestrated V2 workflow.
+
+V2 is intentionally isolated from V1 for safe incremental rollout.
+Current behavior mirrors V1 and can evolve independently.
+"""
 from langgraph.graph import StateGraph, END
 from langchain_core.language_models.chat_models import BaseChatModel
 
 from graphs.state import AgentState
-from graphs.nodes.retrieval import retrieve_both_types
+from graphs.agents.classifier_agent import classifier_node
+from graphs.nodes.retrieval import retrieve_catalogue, retrieve_projects
 from graphs.agents.chat_agent import chat_node
 from graphs.agents.summarizer_agent import summarizer_llm_node, summarizer_export_node
 from tools.pdf import markdown_to_pdf
 from tools.storage import upload_pdf_to_supabase
 
 
-def create_conversation_with_summary_graph(llm: BaseChatModel):
+def create_orchestrated_graph_v2(llm: BaseChatModel, checkpointer=None):
     workflow = StateGraph(AgentState)
 
-    # Nodes
-    workflow.add_node("retrieve", retrieve_both_types)
+    workflow.add_node("classify", classifier_node(llm))
+    workflow.add_node("retrieve", retrieve_catalogue)
+
+    def _router_passthrough(state: AgentState) -> AgentState:
+        q = state.get("query") or ""
+        return {"query": q}
+
+    workflow.add_node("retrieve_router", _router_passthrough)
+    workflow.add_node("retrieve_project", retrieve_projects)
     workflow.add_node("agent", chat_node(llm))
     workflow.add_node("summarizer_llm", summarizer_llm_node(llm))
     workflow.add_node(
@@ -25,13 +37,33 @@ def create_conversation_with_summary_graph(llm: BaseChatModel):
         ),
     )
 
-    # Entry and edges
-    workflow.set_entry_point("retrieve")
+    workflow.set_entry_point("classify")
+
+    workflow.add_conditional_edges(
+        "classify",
+        lambda s: getattr(s.get("classification"), "classification", "CLASSIC"),
+        {
+            "CLASSIC": "retrieve_router",
+            "SUMMARIZE": "summarizer_llm",
+            "UNKNOWN": "retrieve_router",
+        },
+    )
+
+    workflow.add_conditional_edges(
+        "retrieve_router",
+        lambda s: "PROJECT" if s.get("project_id") else "CLASSIC",
+        {
+            "PROJECT": "retrieve_project",
+            "CLASSIC": "retrieve",
+        },
+    )
+
+    workflow.add_edge("retrieve_project", "retrieve")
     workflow.add_edge("retrieve", "agent")
-    workflow.add_edge("agent", "summarizer_llm")
+    workflow.add_edge("agent", END)
+
     workflow.add_edge("summarizer_llm", "summarizer_export")
     workflow.add_edge("summarizer_export", END)
 
-    return workflow.compile()
-
+    return workflow.compile(checkpointer=checkpointer)
 

postman_collection.json

@@ -214,7 +214,7 @@
 						],
 						"body": {
 							"mode": "raw",
-							"raw": "{\n    \"message\": \"Bonjour, comment vas-tu?\",\n    \"model\": \"gpt-4o\",\n    \"agent_type\": \"simple\",\n    \"stream\": false,\n    \"temperature\": 0.7\n}"
+							"raw": "{\n    \"message\": \"Bonjour, comment vas-tu?\",\n    \"model\": \"mistral-large-latest\",\n    \"agent\": \"V2\",\n    \"stream\": false,\n    \"temperature\": 0.7\n}"
 						},
 						"url": {
 							"raw": "{{base_url}}/completion",
@@ -251,7 +251,7 @@
 						],
 						"body": {
 							"mode": "raw",
-							"raw": "{\n    \"message\": \"Explique-moi la théorie de la relativité en 2 phrases\",\n    \"model\": \"mistral-large-latest\",\n    \"agent_type\": \"simple\",\n    \"stream\": false,\n    \"temperature\": 0.7\n}"
+							"raw": "{\n    \"message\": \"Explique-moi la théorie de la relativité en 2 phrases\",\n    \"model\": \"mistral-large-latest\",\n    \"agent\": \"V2\",\n    \"stream\": false,\n    \"temperature\": 0.7\n}"
 						},
 						"url": {
 							"raw": "{{base_url}}/completion",
@@ -288,7 +288,7 @@
 						],
 						"body": {
 							"mode": "raw",
-							"raw": "{\n    \"message\": \"Raconte-moi une courte histoire\",\n    \"model\": \"gpt-3.5-turbo\",\n    \"agent_type\": \"simple\",\n    \"stream\": true,\n    \"temperature\": 0.9\n}"
+							"raw": "{\n    \"message\": \"Raconte-moi une courte histoire\",\n    \"model\": \"mistral-large-latest\",\n    \"agent\": \"V2\",\n    \"stream\": true,\n    \"temperature\": 0.9\n}"
 						},
 						"url": {
 							"raw": "{{base_url}}/completion",
@@ -325,7 +325,7 @@
 						],
 						"body": {
 							"mode": "raw",
-							"raw": "{\n    \"message\": \"Et en Python?\",\n    \"model\": \"gpt-4o\",\n    \"stream\": false,\n    \"conversation_history\": [\n        {\n            \"role\": \"user\",\n            \"content\": \"Comment faire une boucle en JavaScript?\"\n        },\n        {\n            \"role\": \"assistant\",\n            \"content\": \"En JavaScript, vous pouvez utiliser: for (let i = 0; i < 10; i++) { console.log(i); }\"\n        }\n    ]\n}"
+							"raw": "{\n    \"message\": \"Et en Python?\",\n    \"model\": \"mistral-large-latest\",\n    \"stream\": false,\n    \"conversation_history\": [\n        {\n            \"role\": \"user\",\n            \"content\": \"Comment faire une boucle en JavaScript?\"\n        },\n        {\n            \"role\": \"assistant\",\n            \"content\": \"En JavaScript, vous pouvez utiliser: for (let i = 0; i < 10; i++) { console.log(i); }\"\n        }\n    ]\n}"
 						},
 						"url": {
 							"raw": "{{base_url}}/completion",
@@ -362,7 +362,7 @@
 						],
 						"body": {
 							"mode": "raw",
-							"raw": "{\n    \"message\": \"Écris un poème court sur l'IA\",\n    \"model\": \"gpt-4o\",\n    \"agent_type\": \"simple\",\n    \"stream\": false,\n    \"temperature\": 1.2,\n    \"max_tokens\": 150\n}"
+							"raw": "{\n    \"message\": \"Écris un poème court sur l'IA\",\n    \"model\": \"mistral-large-latest\",\n    \"agent\": \"V2\",\n    \"stream\": false,\n    \"temperature\": 1.2,\n    \"max_tokens\": 150\n}"
 						},
 						"url": {
 							"raw": "{{base_url}}/completion",

services/agent_registry.py

@@ -1,37 +1,32 @@
 """Registry for managing multiple LangGraph agents."""
-from typing import Dict, Callable, Any
+from typing import Dict, Callable, Any, Optional
 from langchain_core.language_models.chat_models import BaseChatModel
-from domain.enums import AgentType
-from graphs.base_graph import create_simple_graph, create_simple_graph_with_history
 from graphs.workflows.orchestrated import create_orchestrated_graph
+from graphs.workflows.orchestrated_v2 import create_orchestrated_graph_v2
 
 
 class AgentRegistry:
     """
     Registry for managing multiple agent graph builders.
     
-    This allows for easy addition of new agent types without modifying
-    the API layer. Each agent type maps to a graph builder function.
+    This allows for easy addition of new agent ids without modifying
+    the API layer. Each agent id maps to a graph builder function.
     """
     
     def __init__(self):
         """Initialize the agent registry with default agents."""
-        self._builders: Dict[AgentType, Callable[[BaseChatModel], Any]] = {
-            AgentType.SIMPLE: create_orchestrated_graph,
-            # AgentType.RAG: create_rag_graph,  # À implémenter plus tard
-            # AgentType.TOOLS: create_tools_graph,  # À implémenter plus tard
+        self._agent_builders: Dict[str, Callable[[BaseChatModel], Any]] = {
+            "v1": create_orchestrated_graph,      # V1 (unchanged)
+            "v2": create_orchestrated_graph_v2,   # V2 (isolated)
         }
-        
-        self._descriptions = {
-            AgentType.SIMPLE: "Simple conversational agent without tools or memory",
-            AgentType.RAG: "Agent with Retrieval Augmented Generation (not yet implemented)",
-            AgentType.TOOLS: "Agent with tools like web search, calculator (not yet implemented)",
-            AgentType.CUSTOM: "Custom agent graph (not yet implemented)"
+        self._descriptions: Dict[str, str] = {
+            "v1": "Current production orchestrated workflow",
+            "v2": "Isolated V2 workflow (default)",
         }
     
     def register_agent(
         self,
-        agent_type: AgentType,
+        agent: str,
         builder: Callable[[BaseChatModel], Any],
         description: str = ""
     ) -> None:
@@ -39,45 +34,34 @@ class AgentRegistry:
         Register a new agent builder.
         
         Args:
-            agent_type: Type of agent
+            agent: Agent identifier (string)
             builder: Function that takes an LLM and returns a compiled graph
             description: Description of the agent
         """
-        self._builders[agent_type] = builder
+        agent_key = agent.strip().lower()
+        self._agent_builders[agent_key] = builder
         if description:
-            self._descriptions[agent_type] = description
-    
-    def get_builder(self, agent_type: AgentType) -> Callable[[BaseChatModel], Any]:
-        """
-        Get the builder function for an agent type.
-        
-        Args:
-            agent_type: Type of agent
-            
-        Returns:
-            Builder function
-            
-        Raises:
-            ValueError: If agent type is not registered
+            self._descriptions[agent_key] = description
+
+    def get_builder_for_request(
+        self,
+        agent: Optional[str],
+    ) -> Callable[[BaseChatModel], Any]:
+        """Resolve graph builder from request.
+
+        If agent is missing, defaults to V2.
         """
-        if agent_type not in self._builders:
+        agent_key = (agent or "V2").strip().lower()
+        if agent_key not in self._agent_builders:
             raise ValueError(
-                f"Agent type '{agent_type}' not implemented. "
-                f"Available types: {list(self._builders.keys())}"
+                f"Agent '{agent}' not implemented. "
+                f"Available agents: {list(self._agent_builders.keys())}"
             )
-        return self._builders[agent_type]
-    
-    def is_available(self, agent_type: AgentType) -> bool:
-        """
-        Check if an agent type is available.
-        
-        Args:
-            agent_type: Type of agent
-            
-        Returns:
-            True if agent is available, False otherwise
-        """
-        return agent_type in self._builders
+        return self._agent_builders[agent_key]
+
+    def resolve_agent_id(self, agent: Optional[str]) -> str:
+        """Return canonical agent id used in API metadata."""
+        return (agent or "V2").strip().upper()
     
     def list_agents(self) -> list[dict]:
         """
@@ -86,18 +70,15 @@ class AgentRegistry:
         Returns:
             List of agent information dictionaries
         """
-        agents = []
-        for agent_type in AgentType:
-            agents.append({
-                "type": agent_type.value,
-                "name": agent_type.value.capitalize(),
-                "description": self._descriptions.get(
-                    agent_type,
-                    "No description available"
-                ),
-                "available": self.is_available(agent_type)
-            })
-        return agents
+        return [
+            {
+                "type": key.upper(),
+                "name": key.upper(),
+                "description": self._descriptions.get(key, "No description available"),
+                "available": True,
+            }
+            for key in sorted(self._agent_builders.keys())
+        ]
 
 
 # Singleton instance

services/agent_service.py

@@ -4,7 +4,7 @@ import time
 from langchain_core.messages import AIMessageChunk, HumanMessage, AIMessage, BaseMessage, SystemMessage
 from langchain_core.language_models.chat_models import BaseChatModel
 from langgraph.checkpoint.memory import MemorySaver
-from domain.enums import ModelName, AgentType
+from domain.enums import ModelName
 from .llm_service import llm_service
 from .agent_registry import agent_registry
 from .usage_utils import normalize_usage
@@ -34,7 +34,7 @@ class AgentService:
         self,
         message: str,
         model_name: ModelName,
-        agent_type: AgentType = AgentType.SIMPLE,
+        agent: Optional[str] = None,
         temperature: float = 0.7,
         max_tokens: Optional[int] = None,
         conversation_history: Optional[List[Dict[str, str]]] = None,
@@ -47,7 +47,7 @@ class AgentService:
         Args:
             message: User message
             model_name: LLM model to use
-            agent_type: Type of agent graph
+            agent: Agent identifier. Defaults to "V2" when omitted.
             temperature: Sampling temperature
             max_tokens: Max tokens to generate
             conversation_history: Optional conversation history (ignored if conversation_id is set)
@@ -65,7 +65,8 @@ class AgentService:
             max_tokens=max_tokens
         )
         
-        builder = agent_registry.get_builder(agent_type)
+        resolved_agent = agent_registry.resolve_agent_id(agent)
+        builder = agent_registry.get_builder_for_request(agent=resolved_agent)
         use_memory = bool(conversation_id)
         graph = builder(llm, checkpointer=_text_checkpointer) if use_memory else builder(llm)
         
@@ -114,7 +115,7 @@ class AgentService:
         return {
             "response": response_content,
             "model": model_name.value,
-            "agent_type": agent_type.value,
+            "agent": resolved_agent,
             "usage": usage,
             "metadata": base_metadata,
         }
@@ -123,7 +124,7 @@ class AgentService:
         self,
         message: str,
         model_name: ModelName,
-        agent_type: AgentType = AgentType.SIMPLE,
+        agent: Optional[str] = None,
         temperature: float = 0.7,
         max_tokens: Optional[int] = None,
         conversation_history: Optional[List[Dict[str, str]]] = None,
@@ -136,7 +137,7 @@ class AgentService:
         Args:
             message: User message
             model_name: LLM model to use
-            agent_type: Type of agent graph
+            agent: Agent identifier. Defaults to "V2" when omitted.
             temperature: Sampling temperature
             max_tokens: Max tokens to generate
             conversation_history: Optional conversation history (ignored if conversation_id is set)
@@ -154,7 +155,8 @@ class AgentService:
             max_tokens=max_tokens
         )
         
-        builder = agent_registry.get_builder(agent_type)
+        resolved_agent = agent_registry.resolve_agent_id(agent)
+        builder = agent_registry.get_builder_for_request(agent=resolved_agent)
         use_memory = bool(conversation_id)
         graph = builder(llm, checkpointer=_text_checkpointer) if use_memory else builder(llm)
         
@@ -255,7 +257,7 @@ class AgentService:
                     "done": False,
                     "metadata": {
                         "model": model_name.value,
-                        "agent_type": agent_type.value,
+                        "agent": resolved_agent,
                         "usage": usage_totals
                     },
                     "documents": documents
@@ -278,7 +280,7 @@ class AgentService:
             "done": True,
             "metadata": {
                 "model": model_name.value,
-                "agent_type": agent_type.value,
+                "agent": resolved_agent,
                 "usage": usage_totals,
                 "usage_by_model": usage_by_model,
                 "latency_s": latency_s,

services/voice/voice_agent_service.py

@@ -5,7 +5,7 @@ from typing import Optional
 from langchain_core.messages import HumanMessage
 from langgraph.checkpoint.memory import MemorySaver
 
-from domain.enums import ModelName, AgentType
+from domain.enums import ModelName
 from services.llm_service import llm_service
 from services.agent_registry import agent_registry
 
@@ -29,7 +29,7 @@ class VoiceAgentService:
         self.checkpointer = MemorySaver()
 
         llm = llm_service.get_llm(model_name=model_name, streaming=False)
-        builder = agent_registry.get_builder(AgentType.SIMPLE)
+        builder = agent_registry.get_builder_for_request("V1")
         self.graph = builder(llm, checkpointer=self.checkpointer)
 
     async def process_message(

services/voice/voice_pipeline.py

@@ -6,7 +6,7 @@ from dataclasses import dataclass, field
 
 from langchain_core.runnables import RunnableGenerator
 
-from domain.enums import ModelName, AgentType
+from domain.enums import ModelName
 from services.agent_service import agent_service
 from .stt_service import STTService, stt_service
 from .tts_service import TTSService, tts_service, TTSVoice
@@ -28,7 +28,7 @@ logger = logging.getLogger(__name__)
 class VoiceSessionConfig:
     """Configuration for a voice session."""
     model: ModelName = ModelName.MISTRAL_LARGE
-    agent_type: AgentType = AgentType.SIMPLE  # Uses orchestrated graph by default
+    agent: Optional[str] = None  # Defaults to V2 when omitted
     voice: TTSVoice = "alloy"
     language: Optional[str] = None
     temperature: float = 0.7
@@ -192,7 +192,7 @@ class VoicePipeline:
             async for chunk in agent_service.stream(
                 message=text,
                 model_name=config.model,
-                agent_type=config.agent_type,
+                agent=config.agent,
                 temperature=config.temperature,
                 max_tokens=config.max_tokens,
                 conversation_history=config.conversation_history,
@@ -209,7 +209,7 @@ class VoicePipeline:
                     yield AgentOutputEvent.create(
                         response=full_response,
                         model=metadata.get("model"),
-                        agent_type=metadata.get("agent_type"),
+                        agent_type=metadata.get("agent"),
                         usage=metadata.get("usage")
                     )