Spaces:
Running on CPU Upgrade
Running on CPU Upgrade
File size: 14,877 Bytes
6c9a052 d98e7b3 6c9a052 d98e7b3 6c9a052 bb334c7 6c9a052 bb334c7 6c9a052 a4bb90e 6c9a052 a4bb90e 6c9a052 bb334c7 6c9a052 d98e7b3 6c9a052 570a57c 6c9a052 570a57c 6c9a052 5bb3b2d d98e7b3 5bb3b2d 6c9a052 5bb3b2d 6c9a052 5bb3b2d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 | """
mcp_server.py
─────────────
MCP 服务层
挂载方式(在 ui_nicegui.py 中):
from mcp_server import mcp
app.mount('/mcp', mcp.streamable_http_app())
接入地址:
https://sakizuki-danboorusearch.hf.space/mcp/mcp
支持的工具:
search_tags 自然语言搜索标签
get_related_tags 基于共现表查关联推荐
"""
import json
import asyncio
import logging
from anyio import BrokenResourceError, ClosedResourceError
from mcp.server.fastmcp import FastMCP
from mcp.server.transport_security import TransportSecuritySettings
from core.engine import DanbooruTagger
from core.models import SearchRequest
import core.counter as counter
import re
# ── 过滤客户端断连产生的无害报错噪音 ──────────────────────────────────
class _SuppressClientDisconnect(logging.Filter):
def filter(self, record: logging.LogRecord) -> bool:
exc = record.exc_info[1] if record.exc_info else None
if isinstance(exc, (BrokenResourceError, ClosedResourceError)):
return False # 丢弃该日志记录
return True
_disconnect_filter = _SuppressClientDisconnect()
logging.getLogger("mcp.server.streamable_http").addFilter(_disconnect_filter)
logging.getLogger("uvicorn.error").addFilter(_disconnect_filter)
mcp = FastMCP(
name="danbooru-searcher",
transport_security=TransportSecuritySettings(enable_dns_rebinding_protection=False),
)
@mcp.tool()
async def search_tags(
query: str,
use_segmentation: bool = True,
top_k: int = 5,
limit: int = 80,
popularity_weight: float = 0.15,
show_nsfw: bool = True,
include_wiki: bool = False,
category: str = "all",
group_mode: str = "off",
max_per_group: int = 2,
) -> str:
"""
Search Danbooru tags using natural language and return a ready-to-use prompt.
## Args
- query: Natural language description (Chinese recommended).
- use_segmentation: Split multi-concept input into segments for separate retrieval. True for scene descriptions, False for single-concept queries.
- top_k: Candidates recalled per segment. Semantics change with use_segmentation — see guide below.
- limit: Max tags returned.
- popularity_weight: Influence of tag post count on ranking (0.0–1.0). Default 0.15.
- show_nsfw: Include NSFW tags. Default True.
- include_wiki: Append wiki description to each result. Default False.
- category: Filter results to a specific tag category. Default "all".
"all" — All categories (通用 + 版权 + 人物 )
"general" — General: visual attributes, clothing, pose, background, etc.
"copyright" — Copyright: specific anime/game/franchise titles
"character" — Character: named characters from any series
Use this when you know what kind of tag you need — e.g. looking for a
character name vs. describing a scene visually.
- group_mode: Tag group processing mode. Default "off".
"off" — No group processing (backward compatible)
"expand" — Boost same-group tags for concept exploration
"diverse" — Limit tags per group for scene diversity
- max_per_group: Max tags per group in diverse mode. Default 2.
## Query writing guide
The `query` parameter supports explicit delimiter control for precise segmentation.
### Explicit delimiters
Use **spaces, newlines, Chinese commas (,), or Chinese dunhao (、)** to manually separate concepts.
Each delimiter-bounded segment ≤7 characters stays atomic — the engine respects your intent.
| Query style | Example | When to use |
|---|---|---|
| Concept list (spaces) | `运动社团 校队 比赛 运动会` | You know the exact concepts to search |
| Concept list (dun hao) | `反乌托邦、赛博朋克、蒸汽朋克` | Same, with Chinese list punctuation |
| Natural sentence | `一个穿着白色水手服的少女在雨中奔跑` | Scene description, let the engine auto-split |
| Mixed | `运动社团 一个穿水手服的少女` | Mix concepts with descriptive phrases |
Segments >7 characters are still auto-split by jieba, but the raw segment is kept as an additional query
to preserve clause-level semantics (multi-granularity retrieval).
### Recommendations
1. **Concept lists → use explicit delimiters:** Group independent concepts with spaces or dunhao.
`运动社团 校队 比赛 体育祭 田径部` is better than `运动社团校队比赛体育祭田径部`.
2. **Scene descriptions → write naturally:** Natural Chinese with Chinese commas works well for full scenes.
`一个穿着白色水手服,蓝色短裙的少女在雨中奔跑` — commas here are grammatical, not delimiters.
3. **Precise lookup → turn off segmentation:** For finding a specific character or copyright title,
set `use_segmentation=False` and combine with `category` filter.
e.g. `query="EVA中蓝发的零号机驾驶员"` with `category="character"` and `use_segmentation=False`.
4. **Category filtering:** Use `category` to narrow results. Looking for a character?
`category="character"`. Building a scene prompt? `category="general"`.
## Parameter guide
### Step 1 — Decide use_segmentation + top_k together
top_k means "candidates per segment"; its effect depends on whether segmentation is on.
Multi-concept input (scene description) → use_segmentation=True
| Sub-scenario | top_k | Reason |
|-----------------------|-------|----------------------------------------------------------|
| Full scene → prompt | 5 | Many segments; low top_k distributes result slots fairly |
| Vague concept explore | 80 | Few segments; high top_k needed for broad recall |
Single-concept input → use_segmentation=False
top_k acts as total candidate pool size. Use 20 for all single-concept cases.
| Sub-scenario | top_k |
|-------------------------------|-------|
| Describe subject / find tag | 20 |
| Precise lookup / spell fix | 20 |
### Step 2 — Decide limit independently
| Goal | limit |
|------------------------------|-------|
| Full prompt for image gen | 80 |
| Concept exploration | 20–80 |
| Precise lookup / role search | 10–20 |
### Auxiliary params
popularity_weight (default 0.15, rarely needs changing):
- Higher (0.3+): favor common, well-established tags
- Lower (0.0): surface niche/rare tags
include_wiki (default False):
- True: The meaning of the tag is important — disambiguation, explaining tags to users, exploring unfamiliar domains, or when you are unsure of the tag's meaning
- False: Prompt generation (Wiki is irrelevant to the downstream task), tags are known
### Quick reference
| Scenario | use_segmentation | top_k | limit |
|-------------------------------|------------------|-------|-------|
| Full scene → prompt (default) | True | 5 | 80 |
| Vague concept exploration | True | 80 | 80 |
| Describe subject / find tag | False | 20 | 20 |
| Precise lookup / spell fix | False | 20 | 10 |
### Workflow
After search_tags, pass selected tags to get_related_tags to discover complementary tags via co-occurrence (accessories, character features, scene atmosphere).
Supports chained exploration / iterative loops – take the interesting tags from the returned results as input to call get_related_tags again,
and use the results from get_related to feed back into a new round of search, enabling multi-hop deep traversal along the co-occurrence graph.
## Examples
Precise lookup / spell fix — e.g. "selafuku", "thighhigh", "twintail"
→ use_segmentation=False, top_k=20, limit=10
Vague concept exploration — e.g. "兔耳朵", "赛博朋克服装", "假肢"
→ use_segmentation=True, top_k=80, limit=80
Describe subject / find tag — e.g. "EVA中蓝发的零号机驾驶员", "命运石之门中的助手"
→ use_segmentation=False, top_k=20, limit=20
Full scene → prompt — e.g. "一个穿着白色水手服,蓝色短裙的少女在雨中的城市里奔跑"
→ use_segmentation=True, top_k=5, limit=80
## Returns
JSON with: prompt (comma-separated tags), keywords, results.
Each result: tag, cn_name, category, final_score, count[, wiki if include_wiki=True].
"""
_CATEGORY_MAP: dict[str, list[str]] = {
"all": ["General", "Character", "Copyright", "Artist", "Meta"],
"general": ["General"],
"character": ["Character"],
"copyright": ["Copyright"],
}
target_categories = _CATEGORY_MAP.get(
category,
_CATEGORY_MAP["all"], # unrecognized value → fall back to all
)
tagger = await DanbooruTagger.get_instance()
request = SearchRequest(
query=query,
top_k=top_k,
limit=limit,
popularity_weight=popularity_weight,
show_nsfw=show_nsfw,
use_segmentation=use_segmentation,
target_categories=target_categories,
group_mode=group_mode,
max_per_group=max_per_group,
)
response = await tagger.search_async(request)
# 计数:每次 MCP 搜索调用均计入搜索、成功、复制;访问不变
await counter.increment()
await counter.increment_success()
await counter.increment_copy()
await counter.increment_mcp()
results = []
for r in response.results:
if r.nsfw == '1' and not show_nsfw:
continue
item = {
"tag": r.tag,
"cn_name": r.cn_name,
"category": r.category,
"final_score": r.final_score,
"count": r.count,
}
if include_wiki:
item["wiki"] = r.wiki
results.append(item)
payload = {
"prompt": response.tags_sfw if not show_nsfw else response.tags_all,
"keywords": response.keywords,
"results": results,
}
han_chars = re.findall(r'[\u4e00-\u9fff]', query)
if len(query) > 0 and len(han_chars) / len(query) < 0.5:
payload["hint"] = (
"检测到英文查询,该搜索引擎对中文查询优化更好,如果搜索结果不合预期,推荐用中文重试"
)
return json.dumps(payload, ensure_ascii=False, indent=2)
@mcp.tool()
async def get_related_tags(
tags: list[str],
limit: int = 50,
show_nsfw: bool = True,
include_wiki: bool = False,
) -> str:
"""
Return co-occurrence-based tag recommendations for a given tag list (NPMI scoring).
This tool surfaces tags that frequently appear alongside the seeds in
Danbooru, mixing categories (General / Character / Copyright) by design.
## Typical use cases
- Attribute → characters who have it
e.g. ["fingerless_gloves"] → tifa_lockhart, cammy_white, bridget_(guilty_gear), ...
- Work → characters in it
e.g. ["overlord_(maruyama)"] → shalltear_bloodfallen, ainz_ooal_gown, albedo_(overlord), ...
- Character → their visual attributes
e.g. ["amiya_(arknights)"] → outfits, expressions, accessories
- Theme exploration
e.g. ["fighter_jet"] → aircraft types, actions, backgrounds
- Multi-tag intersection
e.g. ["maid", "twintails"] → tags specific to the combination, scored by summed NPMI
For within-category exploration (e.g. "more clothing tags like X"), use search_tags
with the `category` parameter instead.
## Workflow
Chain freely: search_tags → get_related_tags → get_related_tags → search_tags.
Each hop along the co-occurrence graph reveals tags unreachable by semantic search alone.
## Args
- tags: List of canonical Danbooru tag names (underscores, no spaces).
e.g. ["white_serafuku", "sailor_collar"]
- limit: Max recommendations returned. Default 50.
- show_nsfw: Include NSFW tags. Default True.
- include_wiki: Append wiki description to each result. Default False.
Set True when result tags are unfamiliar and need disambiguation.
## Returns
JSON array sorted by aggregated NPMI score (descending). Each result:
- tag, cn_name, category, count (post_count), cooc_score (normalized to [0,1])
- sources: seed tags that contributed to this score
- wiki: only if include_wiki=True
"""
tagger = await DanbooruTagger.get_instance()
# ── 检查标签是否存在,不存在则尝试 search_tags 纠错 ──────────────────
valid_tags = []
invalid_tags = []
for t in tags:
if t in tagger._name_to_idx:
valid_tags.append(t)
else:
invalid_tags.append(t)
corrections = {}
if invalid_tags:
for bad_tag in invalid_tags:
try:
req = SearchRequest(
query=bad_tag,
top_k=5,
limit=5,
popularity_weight=0.15,
use_segmentation=False,
target_layers=['英文']
)
resp = await tagger.search_async(req)
if resp.results:
corrections[bad_tag] = resp.results[0].tag
except Exception:
pass
if not valid_tags and not corrections:
return json.dumps({
"error": "所有传入的标签均不存在于标签表中",
"invalid_tags": invalid_tags,
}, ensure_ascii=False, indent=2)
# 用纠错后的标签替换无效标签
corrected_tags = []
for t in tags:
if t in valid_tags:
corrected_tags.append(t)
elif t in corrections:
corrected_tags.append(corrections[t])
results = await asyncio.to_thread(
tagger.get_related,
corrected_tags,
set(corrected_tags),
limit,
show_nsfw,
)
# 计数:每次 MCP related 调用均计入搜索、成功、复制;访问不变
await counter.increment()
await counter.increment_success()
await counter.increment_copy()
await counter.increment_mcp()
output = []
for r in results:
item = {
"tag": r.tag,
"cn_name": r.cn_name,
"category": r.category,
"count": r.post_count,
"cooc_score": r.cooc_score,
"sources": r.sources,
}
if include_wiki:
item["wiki"] = r.wiki
output.append(item)
payload = output
if corrections:
correction_notes = [
f"{bad} → {good}" for bad, good in corrections.items()
]
payload = {
"correction_note": "标签拼写错误,已经纠错: " + ", ".join(correction_notes),
"corrections": corrections,
"results": output,
}
return json.dumps(payload, ensure_ascii=False, indent=2) |