Spaces:
Running on CPU Upgrade
Running on CPU Upgrade
Commit ·
9459bd9
1
Parent(s): 914fa31
file refactor, still works
Browse files- agent/tools/docs_tools.py +412 -615
- test_dataset_tools.py +0 -89
agent/tools/docs_tools.py
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
"""
|
| 2 |
-
Documentation search tools for
|
| 3 |
-
Tools for exploring and fetching HuggingFace documentation and API specifications
|
| 4 |
"""
|
| 5 |
|
| 6 |
import asyncio
|
|
|
|
| 7 |
import os
|
| 8 |
from typing import Any
|
| 9 |
|
|
@@ -14,23 +14,16 @@ from whoosh.fields import ID, TEXT, Schema
|
|
| 14 |
from whoosh.filedb.filestore import RamStorage
|
| 15 |
from whoosh.qparser import MultifieldParser, OrGroup
|
| 16 |
|
| 17 |
-
#
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
# Simple in-memory caches for docs and search indexes
|
| 21 |
-
_DOCS_CACHE: dict[str, list[dict[str, str]]] = {}
|
| 22 |
-
_INDEX_CACHE: dict[str, tuple[Any, MultifieldParser]] = {}
|
| 23 |
-
_CACHE_LOCK = asyncio.Lock()
|
| 24 |
|
| 25 |
-
# Result limiting defaults
|
| 26 |
DEFAULT_MAX_RESULTS = 20
|
| 27 |
MAX_RESULTS_CAP = 50
|
| 28 |
|
| 29 |
-
# Gradio documentation endpoints (hosted separately from HF docs)
|
| 30 |
GRADIO_LLMS_TXT_URL = "https://gradio.app/llms.txt"
|
| 31 |
-
|
| 32 |
|
| 33 |
-
# High-level endpoints that bundle related documentation sections
|
| 34 |
COMPOSITE_ENDPOINTS: dict[str, list[str]] = {
|
| 35 |
"optimum": [
|
| 36 |
"optimum",
|
|
@@ -56,32 +49,34 @@ COMPOSITE_ENDPOINTS: dict[str, list[str]] = {
|
|
| 56 |
],
|
| 57 |
}
|
| 58 |
|
| 59 |
-
|
| 60 |
-
def _expand_endpoint(endpoint: str) -> list[str]:
|
| 61 |
-
return COMPOSITE_ENDPOINTS.get(endpoint, [endpoint])
|
| 62 |
-
|
| 63 |
-
|
| 64 |
# ---------------------------------------------------------------------------
|
| 65 |
-
#
|
| 66 |
# ---------------------------------------------------------------------------
|
| 67 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
response = await client.get(GRADIO_LLMS_TXT_URL)
|
| 73 |
-
response.raise_for_status()
|
| 74 |
-
return response.text
|
| 75 |
|
| 76 |
|
| 77 |
-
async def
|
| 78 |
"""
|
| 79 |
-
|
| 80 |
-
|
|
|
|
| 81 |
"""
|
| 82 |
async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
|
| 83 |
-
|
| 84 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
headers={
|
| 86 |
"Content-Type": "application/json",
|
| 87 |
"Origin": "https://gradio-docs-mcp.up.railway.app",
|
|
@@ -92,143 +87,96 @@ async def _search_gradio_docs(query: str) -> str:
|
|
| 92 |
"FALLBACK_PROMPT": "No results found",
|
| 93 |
},
|
| 94 |
)
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
return result.get("SYS_PROMPT", "No results found")
|
| 98 |
|
| 99 |
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
if query:
|
| 104 |
-
header += f"Search query: '{query}'\n\n"
|
| 105 |
-
header += "Source: https://gradio.app/docs\n\n---\n\n"
|
| 106 |
-
return header + content
|
| 107 |
|
| 108 |
|
| 109 |
-
async def
|
| 110 |
-
"""Fetch
|
| 111 |
-
|
| 112 |
-
url = f"{base_url}/{endpoint}"
|
| 113 |
headers = {"Authorization": f"Bearer {hf_token}"}
|
| 114 |
|
| 115 |
async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 126 |
|
| 127 |
-
if not sidebar:
|
| 128 |
-
raise ValueError("Could not find navigation sidebar")
|
| 129 |
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
async def _fetch_single_glimpse(
|
| 145 |
-
client: httpx.AsyncClient, hf_token: str, item: dict[str, str]
|
| 146 |
-
) -> dict[str, str]:
|
| 147 |
-
"""Fetch a short glimpse for a single page"""
|
| 148 |
-
md_url = f"{item['url']}.md"
|
| 149 |
-
headers = {"Authorization": f"Bearer {hf_token}"}
|
| 150 |
-
|
| 151 |
-
try:
|
| 152 |
-
response = await client.get(md_url, headers=headers)
|
| 153 |
-
response.raise_for_status()
|
| 154 |
-
|
| 155 |
-
content = response.text.strip()
|
| 156 |
-
snippet_length = 200
|
| 157 |
-
glimpse = content[:snippet_length].strip()
|
| 158 |
-
if len(content) > snippet_length:
|
| 159 |
-
glimpse += "..."
|
| 160 |
-
|
| 161 |
-
return {
|
| 162 |
-
"title": item["title"],
|
| 163 |
-
"url": item["url"],
|
| 164 |
-
"md_url": md_url,
|
| 165 |
-
"glimpse": glimpse,
|
| 166 |
-
"content": content,
|
| 167 |
-
}
|
| 168 |
-
except Exception as e:
|
| 169 |
-
return {
|
| 170 |
-
"title": item["title"],
|
| 171 |
-
"url": item["url"],
|
| 172 |
-
"md_url": md_url,
|
| 173 |
-
"glimpse": f"[Could not fetch glimpse: {str(e)[:50]}]",
|
| 174 |
-
"content": "",
|
| 175 |
-
}
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
async def _fetch_all_glimpses(
|
| 179 |
-
hf_token: str, nav_data: list[dict[str, str]]
|
| 180 |
-
) -> list[dict[str, str]]:
|
| 181 |
-
"""Fetch glimpses for all pages in parallel"""
|
| 182 |
-
async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
|
| 183 |
-
result_items = await asyncio.gather(
|
| 184 |
-
*[_fetch_single_glimpse(client, hf_token, item) for item in nav_data]
|
| 185 |
-
)
|
| 186 |
|
| 187 |
-
|
|
|
|
|
|
|
|
|
|
| 188 |
|
|
|
|
|
|
|
|
|
|
| 189 |
|
| 190 |
-
async def _load_single_endpoint(hf_token: str, endpoint: str) -> list[dict[str, str]]:
|
| 191 |
-
"""Fetch docs for a single endpoint."""
|
| 192 |
-
html_content = await _fetch_html_page(hf_token, endpoint)
|
| 193 |
-
nav_data = _parse_sidebar_navigation(html_content)
|
| 194 |
-
if not nav_data:
|
| 195 |
-
raise ValueError(f"No navigation links found for endpoint '{endpoint}'")
|
| 196 |
|
| 197 |
-
|
| 198 |
-
|
| 199 |
-
|
| 200 |
-
return docs
|
| 201 |
|
| 202 |
|
| 203 |
-
async def
|
| 204 |
-
"""Return docs for a single endpoint or expanded composite."""
|
| 205 |
-
async with _CACHE_LOCK:
|
| 206 |
-
cached = _DOCS_CACHE.get(endpoint)
|
| 207 |
-
if cached is not None:
|
| 208 |
-
return cached
|
| 209 |
-
|
| 210 |
-
docs: list[dict[str, str]] = []
|
| 211 |
-
for member in _expand_endpoint(endpoint):
|
| 212 |
-
async with _CACHE_LOCK:
|
| 213 |
-
member_cached = _DOCS_CACHE.get(member)
|
| 214 |
-
if member_cached is None:
|
| 215 |
-
member_cached = await _load_single_endpoint(hf_token, member)
|
| 216 |
-
async with _CACHE_LOCK:
|
| 217 |
-
_DOCS_CACHE[member] = member_cached
|
| 218 |
-
docs.extend(member_cached)
|
| 219 |
-
|
| 220 |
-
async with _CACHE_LOCK:
|
| 221 |
-
_DOCS_CACHE[endpoint] = docs
|
| 222 |
-
return docs
|
| 223 |
-
|
| 224 |
-
|
| 225 |
-
async def _ensure_index(
|
| 226 |
endpoint: str, docs: list[dict[str, str]]
|
| 227 |
) -> tuple[Any, MultifieldParser]:
|
| 228 |
-
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
|
| 232 |
|
| 233 |
analyzer = StemmingAnalyzer()
|
| 234 |
schema = Schema(
|
|
@@ -260,226 +208,167 @@ async def _ensure_index(
|
|
| 260 |
group=OrGroup,
|
| 261 |
)
|
| 262 |
|
| 263 |
-
async with
|
| 264 |
-
|
| 265 |
return index, parser
|
| 266 |
|
| 267 |
|
| 268 |
async def _search_docs(
|
| 269 |
-
endpoint: str,
|
| 270 |
-
docs: list[dict[str, str]],
|
| 271 |
-
query: str,
|
| 272 |
-
limit: int | None,
|
| 273 |
) -> tuple[list[dict[str, Any]], str | None]:
|
| 274 |
-
"""
|
| 275 |
-
|
| 276 |
-
|
| 277 |
-
Returns (results, fallback_message). If fallback_message is not None, the caller
|
| 278 |
-
should surface fallback information to the user.
|
| 279 |
-
"""
|
| 280 |
-
index, parser = await _ensure_index(endpoint, docs)
|
| 281 |
|
| 282 |
try:
|
| 283 |
query_obj = parser.parse(query)
|
| 284 |
except Exception:
|
| 285 |
-
return
|
| 286 |
-
[],
|
| 287 |
-
"Query contained unsupported syntax; showing default ordering instead.",
|
| 288 |
-
)
|
| 289 |
|
| 290 |
with index.searcher() as searcher:
|
| 291 |
-
|
| 292 |
-
matches
|
| 293 |
-
|
| 294 |
-
|
| 295 |
-
|
| 296 |
-
|
| 297 |
-
|
| 298 |
-
|
| 299 |
-
|
| 300 |
-
|
| 301 |
-
|
| 302 |
-
|
| 303 |
-
)
|
| 304 |
|
| 305 |
if not matches:
|
| 306 |
-
return [], "No strong matches found; showing default ordering
|
| 307 |
-
|
| 308 |
return matches, None
|
| 309 |
|
| 310 |
|
| 311 |
-
|
| 312 |
-
|
| 313 |
-
|
| 314 |
-
total_items: int,
|
| 315 |
-
query: str | None = None,
|
| 316 |
-
fallback_message: str | None = None,
|
| 317 |
-
) -> str:
|
| 318 |
-
"""Format the exploration results as a readable string"""
|
| 319 |
-
base_url = "https://huggingface.co/docs"
|
| 320 |
-
url = f"{base_url}/{endpoint}"
|
| 321 |
-
result = f"Documentation structure for: {url}\n\n"
|
| 322 |
-
|
| 323 |
-
if query:
|
| 324 |
-
result += (
|
| 325 |
-
f"Query: '{query}' → showing {len(result_items)} result(s)"
|
| 326 |
-
f" out of {total_items} pages"
|
| 327 |
-
)
|
| 328 |
-
if fallback_message:
|
| 329 |
-
result += f" ({fallback_message})"
|
| 330 |
-
result += "\n\n"
|
| 331 |
-
else:
|
| 332 |
-
result += (
|
| 333 |
-
f"Found {len(result_items)} page(s) (total available: {total_items}).\n\n"
|
| 334 |
-
)
|
| 335 |
-
|
| 336 |
-
for i, item in enumerate(result_items, 1):
|
| 337 |
-
result += f"{i}. **{item['title']}**\n"
|
| 338 |
-
result += f" URL: {item['url']}\n"
|
| 339 |
-
result += f" Section: {item.get('section', endpoint)}\n"
|
| 340 |
-
if query and "score" in item:
|
| 341 |
-
result += f" Relevance score: {item['score']:.2f}\n"
|
| 342 |
-
result += f" Glimpse: {item['glimpse']}\n\n"
|
| 343 |
-
|
| 344 |
-
return result
|
| 345 |
|
| 346 |
|
| 347 |
-
|
| 348 |
-
hf_token: str,
|
| 349 |
endpoint: str,
|
|
|
|
|
|
|
| 350 |
query: str | None = None,
|
| 351 |
-
|
| 352 |
) -> str:
|
| 353 |
-
"""
|
| 354 |
-
|
| 355 |
-
|
| 356 |
-
total_count = len(cached_items)
|
| 357 |
-
if max_results is None:
|
| 358 |
-
limit = DEFAULT_MAX_RESULTS
|
| 359 |
-
limit_note = f"Showing top {DEFAULT_MAX_RESULTS} results (set max_results to adjust)."
|
| 360 |
-
else:
|
| 361 |
-
limit = max_results if max_results > 0 else None
|
| 362 |
-
limit_note = None
|
| 363 |
-
if limit is None:
|
| 364 |
-
return "Error: max_results must be greater than zero."
|
| 365 |
-
|
| 366 |
-
if limit > MAX_RESULTS_CAP:
|
| 367 |
-
limit_note = (
|
| 368 |
-
f"Requested {limit} results but showing top {MAX_RESULTS_CAP} (maximum allowed)."
|
| 369 |
-
)
|
| 370 |
-
limit = MAX_RESULTS_CAP
|
| 371 |
-
|
| 372 |
-
selected_items: list[dict[str, Any]]
|
| 373 |
-
fallback_message: str | None = None
|
| 374 |
|
| 375 |
if query:
|
| 376 |
-
|
| 377 |
-
|
| 378 |
-
|
| 379 |
-
|
| 380 |
-
limit,
|
| 381 |
-
)
|
| 382 |
-
|
| 383 |
-
if search_results:
|
| 384 |
-
selected_items = search_results
|
| 385 |
-
else:
|
| 386 |
-
selected_items = cached_items[:limit] if limit else cached_items
|
| 387 |
else:
|
| 388 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 389 |
|
| 390 |
-
|
| 391 |
-
return f"No documentation entries available for endpoint '{endpoint}'."
|
| 392 |
|
| 393 |
-
note = None
|
| 394 |
-
if fallback_message or limit_note:
|
| 395 |
-
pieces = []
|
| 396 |
-
if fallback_message:
|
| 397 |
-
pieces.append(fallback_message)
|
| 398 |
-
if limit_note:
|
| 399 |
-
pieces.append(limit_note)
|
| 400 |
-
note = "; ".join(pieces)
|
| 401 |
-
|
| 402 |
-
result = _format_exploration_results(
|
| 403 |
-
endpoint,
|
| 404 |
-
selected_items,
|
| 405 |
-
total_items=total_count,
|
| 406 |
-
query=query,
|
| 407 |
-
fallback_message=note,
|
| 408 |
-
)
|
| 409 |
|
| 410 |
-
|
|
|
|
|
|
|
| 411 |
|
| 412 |
|
| 413 |
async def explore_hf_docs_handler(arguments: dict[str, Any]) -> tuple[str, bool]:
|
| 414 |
-
"""
|
| 415 |
-
|
| 416 |
-
|
| 417 |
-
Args:
|
| 418 |
-
arguments: Dictionary with 'endpoint' parameter (e.g., 'trl', 'transformers', etc.)
|
| 419 |
-
|
| 420 |
-
Returns:
|
| 421 |
-
Tuple of (structured_navigation_with_glimpses, success)
|
| 422 |
-
"""
|
| 423 |
-
endpoint = arguments.get("endpoint", "")
|
| 424 |
query = arguments.get("query")
|
| 425 |
max_results = arguments.get("max_results")
|
| 426 |
|
| 427 |
if not endpoint:
|
| 428 |
return "Error: No endpoint provided", False
|
| 429 |
|
| 430 |
-
|
| 431 |
-
|
| 432 |
-
# Special handling for Gradio docs (hosted at gradio.app, not HF docs)
|
| 433 |
if endpoint.lower() == "gradio":
|
| 434 |
try:
|
| 435 |
clean_query = (
|
| 436 |
query.strip() if isinstance(query, str) and query.strip() else None
|
| 437 |
)
|
|
|
|
|
|
|
| 438 |
if clean_query:
|
| 439 |
-
|
| 440 |
-
|
| 441 |
-
|
| 442 |
-
# Fetch full docs when no query provided
|
| 443 |
-
content = await _fetch_gradio_full_docs()
|
| 444 |
-
return _format_gradio_results(content, query=clean_query), True
|
| 445 |
except httpx.HTTPStatusError as e:
|
| 446 |
-
return
|
| 447 |
-
f"HTTP error fetching Gradio docs: {e.response.status_code}",
|
| 448 |
-
False,
|
| 449 |
-
)
|
| 450 |
except httpx.RequestError as e:
|
| 451 |
return f"Request error fetching Gradio docs: {str(e)}", False
|
| 452 |
except Exception as e:
|
| 453 |
return f"Error fetching Gradio docs: {str(e)}", False
|
| 454 |
|
| 455 |
-
#
|
| 456 |
hf_token = os.environ.get("HF_TOKEN")
|
| 457 |
-
|
| 458 |
if not hf_token:
|
| 459 |
return "Error: HF_TOKEN environment variable not set", False
|
| 460 |
|
| 461 |
try:
|
| 462 |
-
|
| 463 |
-
|
| 464 |
-
|
| 465 |
-
|
| 466 |
-
|
| 467 |
-
|
| 468 |
-
|
| 469 |
-
|
| 470 |
-
|
| 471 |
-
|
| 472 |
-
|
| 473 |
-
|
| 474 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 475 |
)
|
| 476 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 477 |
|
| 478 |
except httpx.HTTPStatusError as e:
|
| 479 |
-
return
|
| 480 |
-
f"HTTP error: {e.response.status_code} - {e.response.text[:200]}",
|
| 481 |
-
False,
|
| 482 |
-
)
|
| 483 |
except httpx.RequestError as e:
|
| 484 |
return f"Request error: {str(e)}", False
|
| 485 |
except ValueError as e:
|
|
@@ -488,121 +377,98 @@ async def explore_hf_docs_handler(arguments: dict[str, Any]) -> tuple[str, bool]
|
|
| 488 |
return f"Unexpected error: {str(e)}", False
|
| 489 |
|
| 490 |
|
| 491 |
-
async def
|
| 492 |
-
"""Fetch
|
| 493 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 494 |
|
| 495 |
-
if
|
| 496 |
-
|
| 497 |
|
| 498 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 499 |
|
| 500 |
-
async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
|
| 501 |
-
response = await client.get(url)
|
| 502 |
-
response.raise_for_status()
|
| 503 |
|
| 504 |
-
|
| 505 |
-
|
|
|
|
|
|
|
|
|
|
| 506 |
|
| 507 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 508 |
|
| 509 |
|
| 510 |
def _extract_all_tags(spec: dict[str, Any]) -> list[str]:
|
| 511 |
-
"""Extract all unique tags from
|
| 512 |
tags = set()
|
| 513 |
-
|
| 514 |
-
# Get tags from the tags section
|
| 515 |
for tag_obj in spec.get("tags", []):
|
| 516 |
if "name" in tag_obj:
|
| 517 |
tags.add(tag_obj["name"])
|
| 518 |
-
|
| 519 |
-
|
| 520 |
-
for path, path_item in spec.get("paths", {}).items():
|
| 521 |
-
for method, operation in path_item.items():
|
| 522 |
if method in ["get", "post", "put", "delete", "patch", "head", "options"]:
|
| 523 |
-
for tag in
|
| 524 |
tags.add(tag)
|
| 525 |
-
|
| 526 |
-
return sorted(list(tags))
|
| 527 |
-
|
| 528 |
-
|
| 529 |
-
def _search_openapi_by_tag(spec: dict[str, Any], tag: str) -> list[dict[str, Any]]:
|
| 530 |
-
"""Search for API endpoints with a specific tag"""
|
| 531 |
-
results = []
|
| 532 |
-
paths = spec.get("paths", {})
|
| 533 |
-
servers = spec.get("servers", [])
|
| 534 |
-
base_url = (
|
| 535 |
-
servers[0].get("url", "https://huggingface.co")
|
| 536 |
-
if servers
|
| 537 |
-
else "https://huggingface.co"
|
| 538 |
-
)
|
| 539 |
-
|
| 540 |
-
for path, path_item in paths.items():
|
| 541 |
-
for method, operation in path_item.items():
|
| 542 |
-
if method not in [
|
| 543 |
-
"get",
|
| 544 |
-
"post",
|
| 545 |
-
"put",
|
| 546 |
-
"delete",
|
| 547 |
-
"patch",
|
| 548 |
-
"head",
|
| 549 |
-
"options",
|
| 550 |
-
]:
|
| 551 |
-
continue
|
| 552 |
-
|
| 553 |
-
operation_tags = operation.get("tags", [])
|
| 554 |
-
if tag in operation_tags:
|
| 555 |
-
# Extract parameters
|
| 556 |
-
parameters = operation.get("parameters", [])
|
| 557 |
-
request_body = operation.get("requestBody", {})
|
| 558 |
-
responses = operation.get("responses", {})
|
| 559 |
-
|
| 560 |
-
results.append(
|
| 561 |
-
{
|
| 562 |
-
"path": path,
|
| 563 |
-
"method": method.upper(),
|
| 564 |
-
"operationId": operation.get("operationId", ""),
|
| 565 |
-
"summary": operation.get("summary", ""),
|
| 566 |
-
"description": operation.get("description", ""),
|
| 567 |
-
"parameters": parameters,
|
| 568 |
-
"request_body": request_body,
|
| 569 |
-
"responses": responses,
|
| 570 |
-
"base_url": base_url,
|
| 571 |
-
}
|
| 572 |
-
)
|
| 573 |
-
|
| 574 |
-
return results
|
| 575 |
|
| 576 |
|
| 577 |
def _generate_curl_example(endpoint: dict[str, Any]) -> str:
|
| 578 |
-
"""Generate
|
| 579 |
method = endpoint["method"]
|
| 580 |
path = endpoint["path"]
|
| 581 |
base_url = endpoint["base_url"]
|
| 582 |
|
| 583 |
-
# Build
|
| 584 |
full_path = path
|
| 585 |
for param in endpoint.get("parameters", []):
|
| 586 |
if param.get("in") == "path" and param.get("required"):
|
| 587 |
-
|
| 588 |
example = param.get(
|
| 589 |
-
"example", param.get("schema", {}).get("example", f"<{
|
| 590 |
)
|
| 591 |
-
full_path = full_path.replace(f"{{{
|
| 592 |
|
| 593 |
curl = f"curl -X {method} \\\n '{base_url}{full_path}'"
|
| 594 |
|
| 595 |
-
# Add query parameters
|
| 596 |
query_params = [p for p in endpoint.get("parameters", []) if p.get("in") == "query"]
|
| 597 |
if query_params and query_params[0].get("required"):
|
| 598 |
param = query_params[0]
|
| 599 |
example = param.get("example", param.get("schema", {}).get("example", "value"))
|
| 600 |
curl += f"?{param['name']}={example}"
|
| 601 |
|
| 602 |
-
# Add headers
|
| 603 |
curl += " \\\n -H 'Authorization: Bearer $HF_TOKEN'"
|
| 604 |
|
| 605 |
-
# Add request body
|
| 606 |
if method in ["POST", "PUT", "PATCH"] and endpoint.get("request_body"):
|
| 607 |
content = endpoint["request_body"].get("content", {})
|
| 608 |
if "application/json" in content:
|
|
@@ -610,8 +476,6 @@ def _generate_curl_example(endpoint: dict[str, Any]) -> str:
|
|
| 610 |
schema = content["application/json"].get("schema", {})
|
| 611 |
example = schema.get("example", "{}")
|
| 612 |
if isinstance(example, dict):
|
| 613 |
-
import json
|
| 614 |
-
|
| 615 |
example = json.dumps(example, indent=2)
|
| 616 |
curl += f" \\\n -d '{example}'"
|
| 617 |
|
|
@@ -619,72 +483,50 @@ def _generate_curl_example(endpoint: dict[str, Any]) -> str:
|
|
| 619 |
|
| 620 |
|
| 621 |
def _format_parameters(parameters: list[dict[str, Any]]) -> str:
|
| 622 |
-
"""Format parameter information from OpenAPI spec"""
|
| 623 |
if not parameters:
|
| 624 |
return ""
|
| 625 |
|
| 626 |
-
# Group parameters by type
|
| 627 |
path_params = [p for p in parameters if p.get("in") == "path"]
|
| 628 |
query_params = [p for p in parameters if p.get("in") == "query"]
|
| 629 |
header_params = [p for p in parameters if p.get("in") == "header"]
|
| 630 |
|
| 631 |
output = []
|
| 632 |
|
| 633 |
-
|
| 634 |
-
|
| 635 |
-
|
| 636 |
-
|
| 637 |
-
|
| 638 |
-
|
| 639 |
-
|
| 640 |
-
example = param.get("example") or param.get("schema", {}).get("example", "")
|
| 641 |
-
|
| 642 |
-
output.append(f"- `{name}` ({param_type}){required}: {description}")
|
| 643 |
-
if example:
|
| 644 |
-
output.append(f" Example: `{example}`")
|
| 645 |
-
|
| 646 |
-
if query_params:
|
| 647 |
if output:
|
| 648 |
output.append("")
|
| 649 |
-
output.append("**
|
| 650 |
-
for
|
| 651 |
-
name =
|
| 652 |
-
required = " (required)" if
|
| 653 |
-
|
| 654 |
-
|
| 655 |
-
example =
|
| 656 |
-
|
| 657 |
-
output.append(f"- `{name}` ({
|
| 658 |
if example:
|
| 659 |
output.append(f" Example: `{example}`")
|
| 660 |
|
| 661 |
-
if header_params:
|
| 662 |
-
if output:
|
| 663 |
-
output.append("")
|
| 664 |
-
output.append("**Header Parameters:**")
|
| 665 |
-
for param in header_params:
|
| 666 |
-
name = param.get("name", "")
|
| 667 |
-
required = " (required)" if param.get("required") else " (optional)"
|
| 668 |
-
description = param.get("description", "")
|
| 669 |
-
|
| 670 |
-
output.append(f"- `{name}`{required}: {description}")
|
| 671 |
-
|
| 672 |
return "\n".join(output)
|
| 673 |
|
| 674 |
|
| 675 |
def _format_response_info(responses: dict[str, Any]) -> str:
|
| 676 |
-
"""Format response information from OpenAPI spec"""
|
| 677 |
if not responses:
|
| 678 |
return "No response information available"
|
| 679 |
|
| 680 |
output = []
|
| 681 |
-
for
|
| 682 |
-
|
| 683 |
-
|
| 684 |
-
|
| 685 |
-
output.append(f"- **{status_code}**: {desc}")
|
| 686 |
-
|
| 687 |
-
content = response_obj.get("content", {})
|
| 688 |
if "application/json" in content:
|
| 689 |
schema = content["application/json"].get("schema", {})
|
| 690 |
if "type" in schema:
|
|
@@ -694,72 +536,87 @@ def _format_response_info(responses: dict[str, Any]) -> str:
|
|
| 694 |
|
| 695 |
|
| 696 |
def _format_openapi_results(results: list[dict[str, Any]], tag: str) -> str:
|
| 697 |
-
"""Format OpenAPI search results
|
| 698 |
if not results:
|
| 699 |
return f"No API endpoints found with tag '{tag}'"
|
| 700 |
|
| 701 |
-
|
| 702 |
-
|
| 703 |
-
output += "---\n\n"
|
| 704 |
|
| 705 |
-
for i,
|
| 706 |
-
|
| 707 |
|
| 708 |
-
if
|
| 709 |
-
|
| 710 |
|
| 711 |
-
if
|
| 712 |
-
desc =
|
| 713 |
-
if len(
|
| 714 |
desc += "..."
|
| 715 |
-
|
| 716 |
|
| 717 |
-
|
| 718 |
-
params_info = _format_parameters(endpoint.get("parameters", []))
|
| 719 |
if params_info:
|
| 720 |
-
|
| 721 |
|
| 722 |
-
|
| 723 |
-
|
| 724 |
-
|
| 725 |
-
output += "\n```\n\n"
|
| 726 |
|
| 727 |
-
|
| 728 |
-
|
| 729 |
-
|
| 730 |
-
output += "\n\n"
|
| 731 |
|
| 732 |
-
|
| 733 |
-
|
| 734 |
-
return output
|
| 735 |
|
| 736 |
|
| 737 |
async def search_openapi_handler(arguments: dict[str, Any]) -> tuple[str, bool]:
|
| 738 |
-
"""
|
| 739 |
-
Search the HuggingFace OpenAPI specification by tag
|
| 740 |
-
|
| 741 |
-
Args:
|
| 742 |
-
arguments: Dictionary with 'tag' parameter
|
| 743 |
-
|
| 744 |
-
Returns:
|
| 745 |
-
Tuple of (search_results, success)
|
| 746 |
-
"""
|
| 747 |
tag = arguments.get("tag", "")
|
| 748 |
-
|
| 749 |
if not tag:
|
| 750 |
return "Error: No tag provided", False
|
| 751 |
|
| 752 |
try:
|
| 753 |
-
# Fetch OpenAPI spec (cached after first fetch)
|
| 754 |
spec = await _fetch_openapi_spec()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 755 |
|
| 756 |
-
|
| 757 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 758 |
|
| 759 |
-
|
| 760 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 761 |
|
| 762 |
-
return
|
| 763 |
|
| 764 |
except httpx.HTTPStatusError as e:
|
| 765 |
return f"HTTP error fetching OpenAPI spec: {e.response.status_code}", False
|
|
@@ -769,61 +626,81 @@ async def search_openapi_handler(arguments: dict[str, Any]) -> tuple[str, bool]:
|
|
| 769 |
return f"Error searching OpenAPI spec: {str(e)}", False
|
| 770 |
|
| 771 |
|
| 772 |
-
async def
|
| 773 |
-
"""
|
| 774 |
-
|
| 775 |
-
|
| 776 |
-
Args:
|
| 777 |
-
arguments: Dictionary with 'url' parameter (full URL to the doc page)
|
| 778 |
-
|
| 779 |
-
Returns:
|
| 780 |
-
Tuple of (full_markdown_content, success)
|
| 781 |
-
"""
|
| 782 |
-
url = arguments.get("url", "")
|
| 783 |
-
|
| 784 |
-
if not url:
|
| 785 |
-
return "Error: No URL provided", False
|
| 786 |
-
|
| 787 |
-
# Get HF token from environment
|
| 788 |
-
hf_token = os.environ.get("HF_TOKEN")
|
| 789 |
-
|
| 790 |
-
if not hf_token:
|
| 791 |
-
return (
|
| 792 |
-
"Error: HF_TOKEN environment variable not set",
|
| 793 |
-
False,
|
| 794 |
-
)
|
| 795 |
-
|
| 796 |
-
# Add .md extension if not already present
|
| 797 |
-
if not url.endswith(".md"):
|
| 798 |
-
url = f"{url}.md"
|
| 799 |
-
|
| 800 |
-
try:
|
| 801 |
-
# Make request with auth
|
| 802 |
-
headers = {"Authorization": f"Bearer {hf_token}"}
|
| 803 |
-
|
| 804 |
-
async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
|
| 805 |
-
response = await client.get(url, headers=headers)
|
| 806 |
-
response.raise_for_status()
|
| 807 |
-
|
| 808 |
-
content = response.text
|
| 809 |
-
|
| 810 |
-
# Return the markdown content directly
|
| 811 |
-
result = f"Documentation from: {url}\n\n{content}"
|
| 812 |
|
| 813 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 814 |
|
| 815 |
-
except httpx.HTTPStatusError as e:
|
| 816 |
-
return (
|
| 817 |
-
f"HTTP error fetching {url}: {e.response.status_code} - {e.response.text[:200]}",
|
| 818 |
-
False,
|
| 819 |
-
)
|
| 820 |
-
except httpx.RequestError as e:
|
| 821 |
-
return f"Request error fetching {url}: {str(e)}", False
|
| 822 |
-
except Exception as e:
|
| 823 |
-
return f"Error fetching documentation: {str(e)}", False
|
| 824 |
|
|
|
|
|
|
|
|
|
|
| 825 |
|
| 826 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 827 |
|
| 828 |
EXPLORE_HF_DOCS_TOOL_SPEC = {
|
| 829 |
"name": "explore_hf_docs",
|
|
@@ -845,48 +722,7 @@ EXPLORE_HF_DOCS_TOOL_SPEC = {
|
|
| 845 |
"properties": {
|
| 846 |
"endpoint": {
|
| 847 |
"type": "string",
|
| 848 |
-
"enum":
|
| 849 |
-
"hub",
|
| 850 |
-
"transformers",
|
| 851 |
-
"diffusers",
|
| 852 |
-
"datasets",
|
| 853 |
-
"gradio",
|
| 854 |
-
"trackio",
|
| 855 |
-
"smolagents",
|
| 856 |
-
"huggingface_hub",
|
| 857 |
-
"huggingface.js",
|
| 858 |
-
"transformers.js",
|
| 859 |
-
"inference-providers",
|
| 860 |
-
"inference-endpoints",
|
| 861 |
-
"peft",
|
| 862 |
-
"accelerate",
|
| 863 |
-
"optimum",
|
| 864 |
-
"tokenizers",
|
| 865 |
-
"courses",
|
| 866 |
-
"evaluate",
|
| 867 |
-
"tasks",
|
| 868 |
-
"dataset-viewer",
|
| 869 |
-
"trl",
|
| 870 |
-
"simulate",
|
| 871 |
-
"sagemaker",
|
| 872 |
-
"timm",
|
| 873 |
-
"safetensors",
|
| 874 |
-
"tgi",
|
| 875 |
-
"setfit",
|
| 876 |
-
"lerobot",
|
| 877 |
-
"autotrain",
|
| 878 |
-
"tei",
|
| 879 |
-
"bitsandbytes",
|
| 880 |
-
"sentence_transformers",
|
| 881 |
-
"chat-ui",
|
| 882 |
-
"leaderboards",
|
| 883 |
-
"lighteval",
|
| 884 |
-
"argilla",
|
| 885 |
-
"distilabel",
|
| 886 |
-
"microsoft-azure",
|
| 887 |
-
"kernels",
|
| 888 |
-
"google-cloud",
|
| 889 |
-
],
|
| 890 |
"description": (
|
| 891 |
"The documentation endpoint to explore. Each endpoint corresponds to a major section of the Hugging Face documentation:\n\n"
|
| 892 |
"• courses — All Hugging Face courses (LLM, robotics, MCP, smol (llm training), agents, deep RL, computer vision, games, diffusion, 3D, audio) and the cookbook recipes. Probably the best place for examples.\n"
|
|
@@ -934,15 +770,13 @@ EXPLORE_HF_DOCS_TOOL_SPEC = {
|
|
| 934 |
"query": {
|
| 935 |
"type": "string",
|
| 936 |
"description": (
|
| 937 |
-
"Optional keyword query to rank and filter documentation pages.
|
| 938 |
-
"
|
| 939 |
),
|
| 940 |
},
|
| 941 |
"max_results": {
|
| 942 |
"type": "integer",
|
| 943 |
-
"description": (
|
| 944 |
-
"Optional cap on number of results to return. Defaults to 20 when omitted and cannot exceed 50."
|
| 945 |
-
),
|
| 946 |
"minimum": 1,
|
| 947 |
"maximum": 50,
|
| 948 |
},
|
|
@@ -980,40 +814,3 @@ HF_DOCS_FETCH_TOOL_SPEC = {
|
|
| 980 |
"required": ["url"],
|
| 981 |
},
|
| 982 |
}
|
| 983 |
-
|
| 984 |
-
|
| 985 |
-
async def _get_api_search_tool_spec() -> dict[str, Any]:
|
| 986 |
-
"""
|
| 987 |
-
Dynamically generate the OpenAPI tool spec with tag enum populated at runtime
|
| 988 |
-
This must be called async to fetch the OpenAPI spec and extract tags
|
| 989 |
-
"""
|
| 990 |
-
spec = await _fetch_openapi_spec()
|
| 991 |
-
tags = _extract_all_tags(spec)
|
| 992 |
-
|
| 993 |
-
return {
|
| 994 |
-
"name": "search_hf_api_endpoints",
|
| 995 |
-
"description": (
|
| 996 |
-
"Search HuggingFace OpenAPI specification by tag to find API endpoints with curl examples. "
|
| 997 |
-
"**Use when:** (1) Need to interact with HF Hub API directly, (2) Building scripts for repo operations, "
|
| 998 |
-
"(3) Need authentication patterns, (4) Understanding API parameters and responses, "
|
| 999 |
-
"(5) Need curl examples for HTTP requests. "
|
| 1000 |
-
"Returns: Endpoint paths, methods, parameters, curl examples with authentication, and response schemas. "
|
| 1001 |
-
"**Pattern:** search_hf_api_endpoints (find endpoint) → use curl pattern in implementation. "
|
| 1002 |
-
"Tags group related operations: repos, models, datasets, inference, spaces, etc. "
|
| 1003 |
-
"**Note:** Each result includes curl example with $HF_TOKEN placeholder for authentication. "
|
| 1004 |
-
"**For tool building:** This provides the API foundation for creating Hub interaction scripts."
|
| 1005 |
-
),
|
| 1006 |
-
"parameters": {
|
| 1007 |
-
"type": "object",
|
| 1008 |
-
"properties": {
|
| 1009 |
-
"tag": {
|
| 1010 |
-
"type": "string",
|
| 1011 |
-
"enum": tags,
|
| 1012 |
-
"description": (
|
| 1013 |
-
"The API tag to search for. Each tag groups related API endpoints. "
|
| 1014 |
-
),
|
| 1015 |
-
},
|
| 1016 |
-
},
|
| 1017 |
-
"required": ["tag"],
|
| 1018 |
-
},
|
| 1019 |
-
}
|
|
|
|
| 1 |
"""
|
| 2 |
+
Documentation search tools for exploring HuggingFace and Gradio documentation.
|
|
|
|
| 3 |
"""
|
| 4 |
|
| 5 |
import asyncio
|
| 6 |
+
import json
|
| 7 |
import os
|
| 8 |
from typing import Any
|
| 9 |
|
|
|
|
| 14 |
from whoosh.filedb.filestore import RamStorage
|
| 15 |
from whoosh.qparser import MultifieldParser, OrGroup
|
| 16 |
|
| 17 |
+
# ---------------------------------------------------------------------------
|
| 18 |
+
# Configuration
|
| 19 |
+
# ---------------------------------------------------------------------------
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
|
|
|
|
| 21 |
DEFAULT_MAX_RESULTS = 20
|
| 22 |
MAX_RESULTS_CAP = 50
|
| 23 |
|
|
|
|
| 24 |
GRADIO_LLMS_TXT_URL = "https://gradio.app/llms.txt"
|
| 25 |
+
GRADIO_SEARCH_URL = "https://playground-worker.pages.dev/api/prompt"
|
| 26 |
|
|
|
|
| 27 |
COMPOSITE_ENDPOINTS: dict[str, list[str]] = {
|
| 28 |
"optimum": [
|
| 29 |
"optimum",
|
|
|
|
| 49 |
],
|
| 50 |
}
|
| 51 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
# ---------------------------------------------------------------------------
|
| 53 |
+
# Caches
|
| 54 |
# ---------------------------------------------------------------------------
|
| 55 |
|
| 56 |
+
_docs_cache: dict[str, list[dict[str, str]]] = {}
|
| 57 |
+
_index_cache: dict[str, tuple[Any, MultifieldParser]] = {}
|
| 58 |
+
_cache_lock = asyncio.Lock()
|
| 59 |
+
_openapi_cache: dict[str, Any] | None = None
|
| 60 |
|
| 61 |
+
# ---------------------------------------------------------------------------
|
| 62 |
+
# Gradio Documentation
|
| 63 |
+
# ---------------------------------------------------------------------------
|
|
|
|
|
|
|
|
|
|
| 64 |
|
| 65 |
|
| 66 |
+
async def _fetch_gradio_docs(query: str | None = None) -> str:
|
| 67 |
"""
|
| 68 |
+
Fetch Gradio documentation.
|
| 69 |
+
Without query: Get full documentation from llms.txt
|
| 70 |
+
With query: Run embedding search on guides/demos for relevant content
|
| 71 |
"""
|
| 72 |
async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
|
| 73 |
+
if not query:
|
| 74 |
+
resp = await client.get(GRADIO_LLMS_TXT_URL)
|
| 75 |
+
resp.raise_for_status()
|
| 76 |
+
return resp.text
|
| 77 |
+
|
| 78 |
+
resp = await client.post(
|
| 79 |
+
GRADIO_SEARCH_URL,
|
| 80 |
headers={
|
| 81 |
"Content-Type": "application/json",
|
| 82 |
"Origin": "https://gradio-docs-mcp.up.railway.app",
|
|
|
|
| 87 |
"FALLBACK_PROMPT": "No results found",
|
| 88 |
},
|
| 89 |
)
|
| 90 |
+
resp.raise_for_status()
|
| 91 |
+
return resp.json().get("SYS_PROMPT", "No results found")
|
|
|
|
| 92 |
|
| 93 |
|
| 94 |
+
# ---------------------------------------------------------------------------
|
| 95 |
+
# HF Documentation - Fetching
|
| 96 |
+
# ---------------------------------------------------------------------------
|
|
|
|
|
|
|
|
|
|
|
|
|
| 97 |
|
| 98 |
|
| 99 |
+
async def _fetch_endpoint_docs(hf_token: str, endpoint: str) -> list[dict[str, str]]:
|
| 100 |
+
"""Fetch all docs for an endpoint by parsing sidebar and fetching each page."""
|
| 101 |
+
url = f"https://huggingface.co/docs/{endpoint}"
|
|
|
|
| 102 |
headers = {"Authorization": f"Bearer {hf_token}"}
|
| 103 |
|
| 104 |
async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
|
| 105 |
+
resp = await client.get(url, headers=headers)
|
| 106 |
+
resp.raise_for_status()
|
| 107 |
+
|
| 108 |
+
soup = BeautifulSoup(resp.text, "html.parser")
|
| 109 |
+
sidebar = soup.find("nav", class_=lambda x: x and "flex-auto" in x)
|
| 110 |
+
if not sidebar:
|
| 111 |
+
raise ValueError(f"Could not find navigation sidebar for '{endpoint}'")
|
| 112 |
+
|
| 113 |
+
nav_items = []
|
| 114 |
+
for link in sidebar.find_all("a", href=True):
|
| 115 |
+
href = link["href"]
|
| 116 |
+
page_url = f"https://huggingface.co{href}" if href.startswith("/") else href
|
| 117 |
+
nav_items.append({"title": link.get_text(strip=True), "url": page_url})
|
| 118 |
+
|
| 119 |
+
if not nav_items:
|
| 120 |
+
raise ValueError(f"No navigation links found for '{endpoint}'")
|
| 121 |
+
|
| 122 |
+
async def fetch_page(item: dict[str, str]) -> dict[str, str]:
|
| 123 |
+
md_url = f"{item['url']}.md"
|
| 124 |
+
try:
|
| 125 |
+
r = await client.get(md_url, headers=headers)
|
| 126 |
+
r.raise_for_status()
|
| 127 |
+
content = r.text.strip()
|
| 128 |
+
glimpse = content[:200] + "..." if len(content) > 200 else content
|
| 129 |
+
except Exception as e:
|
| 130 |
+
content, glimpse = "", f"[Could not fetch: {str(e)[:50]}]"
|
| 131 |
+
return {
|
| 132 |
+
"title": item["title"],
|
| 133 |
+
"url": item["url"],
|
| 134 |
+
"md_url": md_url,
|
| 135 |
+
"glimpse": glimpse,
|
| 136 |
+
"content": content,
|
| 137 |
+
"section": endpoint,
|
| 138 |
+
}
|
| 139 |
+
|
| 140 |
+
return list(await asyncio.gather(*[fetch_page(item) for item in nav_items]))
|
| 141 |
|
|
|
|
|
|
|
| 142 |
|
| 143 |
+
async def _get_docs(hf_token: str, endpoint: str) -> list[dict[str, str]]:
|
| 144 |
+
"""Get docs for endpoint with caching. Expands composite endpoints."""
|
| 145 |
+
async with _cache_lock:
|
| 146 |
+
if endpoint in _docs_cache:
|
| 147 |
+
return _docs_cache[endpoint]
|
| 148 |
+
|
| 149 |
+
sub_endpoints = COMPOSITE_ENDPOINTS.get(endpoint, [endpoint])
|
| 150 |
+
all_docs: list[dict[str, str]] = []
|
| 151 |
+
|
| 152 |
+
for sub in sub_endpoints:
|
| 153 |
+
async with _cache_lock:
|
| 154 |
+
if sub in _docs_cache:
|
| 155 |
+
all_docs.extend(_docs_cache[sub])
|
| 156 |
+
continue
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 157 |
|
| 158 |
+
docs = await _fetch_endpoint_docs(hf_token, sub)
|
| 159 |
+
async with _cache_lock:
|
| 160 |
+
_docs_cache[sub] = docs
|
| 161 |
+
all_docs.extend(docs)
|
| 162 |
|
| 163 |
+
async with _cache_lock:
|
| 164 |
+
_docs_cache[endpoint] = all_docs
|
| 165 |
+
return all_docs
|
| 166 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 167 |
|
| 168 |
+
# ---------------------------------------------------------------------------
|
| 169 |
+
# HF Documentation - Search
|
| 170 |
+
# ---------------------------------------------------------------------------
|
|
|
|
| 171 |
|
| 172 |
|
| 173 |
+
async def _build_search_index(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 174 |
endpoint: str, docs: list[dict[str, str]]
|
| 175 |
) -> tuple[Any, MultifieldParser]:
|
| 176 |
+
"""Build or retrieve cached Whoosh search index."""
|
| 177 |
+
async with _cache_lock:
|
| 178 |
+
if endpoint in _index_cache:
|
| 179 |
+
return _index_cache[endpoint]
|
| 180 |
|
| 181 |
analyzer = StemmingAnalyzer()
|
| 182 |
schema = Schema(
|
|
|
|
| 208 |
group=OrGroup,
|
| 209 |
)
|
| 210 |
|
| 211 |
+
async with _cache_lock:
|
| 212 |
+
_index_cache[endpoint] = (index, parser)
|
| 213 |
return index, parser
|
| 214 |
|
| 215 |
|
| 216 |
async def _search_docs(
|
| 217 |
+
endpoint: str, docs: list[dict[str, str]], query: str, limit: int
|
|
|
|
|
|
|
|
|
|
| 218 |
) -> tuple[list[dict[str, Any]], str | None]:
|
| 219 |
+
"""Search docs using Whoosh. Returns (results, fallback_message)."""
|
| 220 |
+
index, parser = await _build_search_index(endpoint, docs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 221 |
|
| 222 |
try:
|
| 223 |
query_obj = parser.parse(query)
|
| 224 |
except Exception:
|
| 225 |
+
return [], "Query contained unsupported syntax; showing default ordering."
|
|
|
|
|
|
|
|
|
|
| 226 |
|
| 227 |
with index.searcher() as searcher:
|
| 228 |
+
results = searcher.search(query_obj, limit=limit)
|
| 229 |
+
matches = [
|
| 230 |
+
{
|
| 231 |
+
"title": hit["title"],
|
| 232 |
+
"url": hit["url"],
|
| 233 |
+
"md_url": hit.get("md_url", ""),
|
| 234 |
+
"section": hit.get("section", endpoint),
|
| 235 |
+
"glimpse": hit["glimpse"],
|
| 236 |
+
"score": round(hit.score, 2),
|
| 237 |
+
}
|
| 238 |
+
for hit in results
|
| 239 |
+
]
|
|
|
|
| 240 |
|
| 241 |
if not matches:
|
| 242 |
+
return [], "No strong matches found; showing default ordering."
|
|
|
|
| 243 |
return matches, None
|
| 244 |
|
| 245 |
|
| 246 |
+
# ---------------------------------------------------------------------------
|
| 247 |
+
# HF Documentation - Formatting
|
| 248 |
+
# ---------------------------------------------------------------------------
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 249 |
|
| 250 |
|
| 251 |
+
def _format_results(
|
|
|
|
| 252 |
endpoint: str,
|
| 253 |
+
items: list[dict[str, Any]],
|
| 254 |
+
total: int,
|
| 255 |
query: str | None = None,
|
| 256 |
+
note: str | None = None,
|
| 257 |
) -> str:
|
| 258 |
+
"""Format search results as readable text."""
|
| 259 |
+
base_url = f"https://huggingface.co/docs/{endpoint}"
|
| 260 |
+
out = f"Documentation structure for: {base_url}\n\n"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 261 |
|
| 262 |
if query:
|
| 263 |
+
out += f"Query: '{query}' → showing {len(items)} result(s) out of {total} pages"
|
| 264 |
+
if note:
|
| 265 |
+
out += f" ({note})"
|
| 266 |
+
out += "\n\n"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 267 |
else:
|
| 268 |
+
out += f"Found {len(items)} page(s) (total available: {total}).\n"
|
| 269 |
+
if note:
|
| 270 |
+
out += f"({note})\n"
|
| 271 |
+
out += "\n"
|
| 272 |
+
|
| 273 |
+
for i, item in enumerate(items, 1):
|
| 274 |
+
out += f"{i}. **{item['title']}**\n"
|
| 275 |
+
out += f" URL: {item['url']}\n"
|
| 276 |
+
out += f" Section: {item.get('section', endpoint)}\n"
|
| 277 |
+
if query and "score" in item:
|
| 278 |
+
out += f" Relevance score: {item['score']:.2f}\n"
|
| 279 |
+
out += f" Glimpse: {item['glimpse']}\n\n"
|
| 280 |
|
| 281 |
+
return out
|
|
|
|
| 282 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 283 |
|
| 284 |
+
# ---------------------------------------------------------------------------
|
| 285 |
+
# Handlers
|
| 286 |
+
# ---------------------------------------------------------------------------
|
| 287 |
|
| 288 |
|
| 289 |
async def explore_hf_docs_handler(arguments: dict[str, Any]) -> tuple[str, bool]:
|
| 290 |
+
"""Explore documentation structure with optional search query."""
|
| 291 |
+
endpoint = arguments.get("endpoint", "").lstrip("/")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 292 |
query = arguments.get("query")
|
| 293 |
max_results = arguments.get("max_results")
|
| 294 |
|
| 295 |
if not endpoint:
|
| 296 |
return "Error: No endpoint provided", False
|
| 297 |
|
| 298 |
+
# Gradio uses its own API
|
|
|
|
|
|
|
| 299 |
if endpoint.lower() == "gradio":
|
| 300 |
try:
|
| 301 |
clean_query = (
|
| 302 |
query.strip() if isinstance(query, str) and query.strip() else None
|
| 303 |
)
|
| 304 |
+
content = await _fetch_gradio_docs(clean_query)
|
| 305 |
+
header = "# Gradio Documentation\n\n"
|
| 306 |
if clean_query:
|
| 307 |
+
header += f"Query: '{clean_query}'\n\n"
|
| 308 |
+
header += "Source: https://gradio.app/docs\n\n---\n\n"
|
| 309 |
+
return header + content, True
|
|
|
|
|
|
|
|
|
|
| 310 |
except httpx.HTTPStatusError as e:
|
| 311 |
+
return f"HTTP error fetching Gradio docs: {e.response.status_code}", False
|
|
|
|
|
|
|
|
|
|
| 312 |
except httpx.RequestError as e:
|
| 313 |
return f"Request error fetching Gradio docs: {str(e)}", False
|
| 314 |
except Exception as e:
|
| 315 |
return f"Error fetching Gradio docs: {str(e)}", False
|
| 316 |
|
| 317 |
+
# HF docs
|
| 318 |
hf_token = os.environ.get("HF_TOKEN")
|
|
|
|
| 319 |
if not hf_token:
|
| 320 |
return "Error: HF_TOKEN environment variable not set", False
|
| 321 |
|
| 322 |
try:
|
| 323 |
+
max_results_int = int(max_results) if max_results is not None else None
|
| 324 |
+
except (TypeError, ValueError):
|
| 325 |
+
return "Error: max_results must be an integer", False
|
| 326 |
+
|
| 327 |
+
if max_results_int is not None and max_results_int <= 0:
|
| 328 |
+
return "Error: max_results must be greater than zero", False
|
| 329 |
+
|
| 330 |
+
try:
|
| 331 |
+
docs = await _get_docs(hf_token, endpoint)
|
| 332 |
+
total = len(docs)
|
| 333 |
+
|
| 334 |
+
# Determine limit
|
| 335 |
+
if max_results_int is None:
|
| 336 |
+
limit = DEFAULT_MAX_RESULTS
|
| 337 |
+
limit_note = f"Showing top {DEFAULT_MAX_RESULTS} results (set max_results to adjust)."
|
| 338 |
+
elif max_results_int > MAX_RESULTS_CAP:
|
| 339 |
+
limit = MAX_RESULTS_CAP
|
| 340 |
+
limit_note = f"Requested {max_results_int} but showing top {MAX_RESULTS_CAP} (maximum)."
|
| 341 |
+
else:
|
| 342 |
+
limit = max_results_int
|
| 343 |
+
limit_note = None
|
| 344 |
+
|
| 345 |
+
# Search or paginate
|
| 346 |
+
clean_query = (
|
| 347 |
+
query.strip() if isinstance(query, str) and query.strip() else None
|
| 348 |
)
|
| 349 |
+
fallback_msg = None
|
| 350 |
+
|
| 351 |
+
if clean_query:
|
| 352 |
+
results, fallback_msg = await _search_docs(
|
| 353 |
+
endpoint, docs, clean_query, limit
|
| 354 |
+
)
|
| 355 |
+
if not results:
|
| 356 |
+
results = docs[:limit]
|
| 357 |
+
else:
|
| 358 |
+
results = docs[:limit]
|
| 359 |
+
|
| 360 |
+
# Combine notes
|
| 361 |
+
notes = []
|
| 362 |
+
if fallback_msg:
|
| 363 |
+
notes.append(fallback_msg)
|
| 364 |
+
if limit_note:
|
| 365 |
+
notes.append(limit_note)
|
| 366 |
+
note = "; ".join(notes) if notes else None
|
| 367 |
+
|
| 368 |
+
return _format_results(endpoint, results, total, clean_query, note), True
|
| 369 |
|
| 370 |
except httpx.HTTPStatusError as e:
|
| 371 |
+
return f"HTTP error: {e.response.status_code} - {e.response.text[:200]}", False
|
|
|
|
|
|
|
|
|
|
| 372 |
except httpx.RequestError as e:
|
| 373 |
return f"Request error: {str(e)}", False
|
| 374 |
except ValueError as e:
|
|
|
|
| 377 |
return f"Unexpected error: {str(e)}", False
|
| 378 |
|
| 379 |
|
| 380 |
+
async def hf_docs_fetch_handler(arguments: dict[str, Any]) -> tuple[str, bool]:
|
| 381 |
+
"""Fetch full markdown content of a documentation page."""
|
| 382 |
+
url = arguments.get("url", "")
|
| 383 |
+
if not url:
|
| 384 |
+
return "Error: No URL provided", False
|
| 385 |
+
|
| 386 |
+
hf_token = os.environ.get("HF_TOKEN")
|
| 387 |
+
if not hf_token:
|
| 388 |
+
return "Error: HF_TOKEN environment variable not set", False
|
| 389 |
|
| 390 |
+
if not url.endswith(".md"):
|
| 391 |
+
url = f"{url}.md"
|
| 392 |
|
| 393 |
+
try:
|
| 394 |
+
async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
|
| 395 |
+
resp = await client.get(
|
| 396 |
+
url, headers={"Authorization": f"Bearer {hf_token}"}
|
| 397 |
+
)
|
| 398 |
+
resp.raise_for_status()
|
| 399 |
+
return f"Documentation from: {url}\n\n{resp.text}", True
|
| 400 |
+
except httpx.HTTPStatusError as e:
|
| 401 |
+
return (
|
| 402 |
+
f"HTTP error fetching {url}: {e.response.status_code} - {e.response.text[:200]}",
|
| 403 |
+
False,
|
| 404 |
+
)
|
| 405 |
+
except httpx.RequestError as e:
|
| 406 |
+
return f"Request error fetching {url}: {str(e)}", False
|
| 407 |
+
except Exception as e:
|
| 408 |
+
return f"Error fetching documentation: {str(e)}", False
|
| 409 |
+
|
| 410 |
+
|
| 411 |
+
# ---------------------------------------------------------------------------
|
| 412 |
+
# OpenAPI Search
|
| 413 |
+
# ---------------------------------------------------------------------------
|
| 414 |
|
|
|
|
|
|
|
|
|
|
| 415 |
|
| 416 |
+
async def _fetch_openapi_spec() -> dict[str, Any]:
|
| 417 |
+
"""Fetch and cache HuggingFace OpenAPI specification."""
|
| 418 |
+
global _openapi_cache
|
| 419 |
+
if _openapi_cache is not None:
|
| 420 |
+
return _openapi_cache
|
| 421 |
|
| 422 |
+
async with httpx.AsyncClient(timeout=30.0, follow_redirects=True) as client:
|
| 423 |
+
resp = await client.get("https://huggingface.co/.well-known/openapi.json")
|
| 424 |
+
resp.raise_for_status()
|
| 425 |
+
|
| 426 |
+
_openapi_cache = resp.json()
|
| 427 |
+
return _openapi_cache
|
| 428 |
|
| 429 |
|
| 430 |
def _extract_all_tags(spec: dict[str, Any]) -> list[str]:
|
| 431 |
+
"""Extract all unique tags from OpenAPI spec."""
|
| 432 |
tags = set()
|
|
|
|
|
|
|
| 433 |
for tag_obj in spec.get("tags", []):
|
| 434 |
if "name" in tag_obj:
|
| 435 |
tags.add(tag_obj["name"])
|
| 436 |
+
for path_item in spec.get("paths", {}).values():
|
| 437 |
+
for method, op in path_item.items():
|
|
|
|
|
|
|
| 438 |
if method in ["get", "post", "put", "delete", "patch", "head", "options"]:
|
| 439 |
+
for tag in op.get("tags", []):
|
| 440 |
tags.add(tag)
|
| 441 |
+
return sorted(tags)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 442 |
|
| 443 |
|
| 444 |
def _generate_curl_example(endpoint: dict[str, Any]) -> str:
|
| 445 |
+
"""Generate curl command example for an endpoint."""
|
| 446 |
method = endpoint["method"]
|
| 447 |
path = endpoint["path"]
|
| 448 |
base_url = endpoint["base_url"]
|
| 449 |
|
| 450 |
+
# Build URL with path parameters
|
| 451 |
full_path = path
|
| 452 |
for param in endpoint.get("parameters", []):
|
| 453 |
if param.get("in") == "path" and param.get("required"):
|
| 454 |
+
name = param["name"]
|
| 455 |
example = param.get(
|
| 456 |
+
"example", param.get("schema", {}).get("example", f"<{name}>")
|
| 457 |
)
|
| 458 |
+
full_path = full_path.replace(f"{{{name}}}", str(example))
|
| 459 |
|
| 460 |
curl = f"curl -X {method} \\\n '{base_url}{full_path}'"
|
| 461 |
|
| 462 |
+
# Add query parameters
|
| 463 |
query_params = [p for p in endpoint.get("parameters", []) if p.get("in") == "query"]
|
| 464 |
if query_params and query_params[0].get("required"):
|
| 465 |
param = query_params[0]
|
| 466 |
example = param.get("example", param.get("schema", {}).get("example", "value"))
|
| 467 |
curl += f"?{param['name']}={example}"
|
| 468 |
|
|
|
|
| 469 |
curl += " \\\n -H 'Authorization: Bearer $HF_TOKEN'"
|
| 470 |
|
| 471 |
+
# Add request body
|
| 472 |
if method in ["POST", "PUT", "PATCH"] and endpoint.get("request_body"):
|
| 473 |
content = endpoint["request_body"].get("content", {})
|
| 474 |
if "application/json" in content:
|
|
|
|
| 476 |
schema = content["application/json"].get("schema", {})
|
| 477 |
example = schema.get("example", "{}")
|
| 478 |
if isinstance(example, dict):
|
|
|
|
|
|
|
| 479 |
example = json.dumps(example, indent=2)
|
| 480 |
curl += f" \\\n -d '{example}'"
|
| 481 |
|
|
|
|
| 483 |
|
| 484 |
|
| 485 |
def _format_parameters(parameters: list[dict[str, Any]]) -> str:
|
| 486 |
+
"""Format parameter information from OpenAPI spec."""
|
| 487 |
if not parameters:
|
| 488 |
return ""
|
| 489 |
|
|
|
|
| 490 |
path_params = [p for p in parameters if p.get("in") == "path"]
|
| 491 |
query_params = [p for p in parameters if p.get("in") == "query"]
|
| 492 |
header_params = [p for p in parameters if p.get("in") == "header"]
|
| 493 |
|
| 494 |
output = []
|
| 495 |
|
| 496 |
+
for label, params in [
|
| 497 |
+
("Path Parameters", path_params),
|
| 498 |
+
("Query Parameters", query_params),
|
| 499 |
+
("Header Parameters", header_params),
|
| 500 |
+
]:
|
| 501 |
+
if not params:
|
| 502 |
+
continue
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 503 |
if output:
|
| 504 |
output.append("")
|
| 505 |
+
output.append(f"**{label}:**")
|
| 506 |
+
for p in params:
|
| 507 |
+
name = p.get("name", "")
|
| 508 |
+
required = " (required)" if p.get("required") else " (optional)"
|
| 509 |
+
desc = p.get("description", "")
|
| 510 |
+
ptype = p.get("schema", {}).get("type", "string")
|
| 511 |
+
example = p.get("example") or p.get("schema", {}).get("example", "")
|
| 512 |
+
|
| 513 |
+
output.append(f"- `{name}` ({ptype}){required}: {desc}")
|
| 514 |
if example:
|
| 515 |
output.append(f" Example: `{example}`")
|
| 516 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 517 |
return "\n".join(output)
|
| 518 |
|
| 519 |
|
| 520 |
def _format_response_info(responses: dict[str, Any]) -> str:
|
| 521 |
+
"""Format response information from OpenAPI spec."""
|
| 522 |
if not responses:
|
| 523 |
return "No response information available"
|
| 524 |
|
| 525 |
output = []
|
| 526 |
+
for status, resp_obj in list(responses.items())[:3]:
|
| 527 |
+
desc = resp_obj.get("description", "")
|
| 528 |
+
output.append(f"- **{status}**: {desc}")
|
| 529 |
+
content = resp_obj.get("content", {})
|
|
|
|
|
|
|
|
|
|
| 530 |
if "application/json" in content:
|
| 531 |
schema = content["application/json"].get("schema", {})
|
| 532 |
if "type" in schema:
|
|
|
|
| 536 |
|
| 537 |
|
| 538 |
def _format_openapi_results(results: list[dict[str, Any]], tag: str) -> str:
|
| 539 |
+
"""Format OpenAPI search results with curl examples."""
|
| 540 |
if not results:
|
| 541 |
return f"No API endpoints found with tag '{tag}'"
|
| 542 |
|
| 543 |
+
out = f"# API Endpoints for tag: `{tag}`\n\n"
|
| 544 |
+
out += f"Found {len(results)} endpoint(s)\n\n---\n\n"
|
|
|
|
| 545 |
|
| 546 |
+
for i, ep in enumerate(results, 1):
|
| 547 |
+
out += f"## {i}. {ep['method']} {ep['path']}\n\n"
|
| 548 |
|
| 549 |
+
if ep["summary"]:
|
| 550 |
+
out += f"**Summary:** {ep['summary']}\n\n"
|
| 551 |
|
| 552 |
+
if ep["description"]:
|
| 553 |
+
desc = ep["description"][:300]
|
| 554 |
+
if len(ep["description"]) > 300:
|
| 555 |
desc += "..."
|
| 556 |
+
out += f"**Description:** {desc}\n\n"
|
| 557 |
|
| 558 |
+
params_info = _format_parameters(ep.get("parameters", []))
|
|
|
|
| 559 |
if params_info:
|
| 560 |
+
out += params_info + "\n\n"
|
| 561 |
|
| 562 |
+
out += "**Usage:**\n```bash\n"
|
| 563 |
+
out += _generate_curl_example(ep)
|
| 564 |
+
out += "\n```\n\n"
|
|
|
|
| 565 |
|
| 566 |
+
out += "**Returns:**\n"
|
| 567 |
+
out += _format_response_info(ep["responses"])
|
| 568 |
+
out += "\n\n---\n\n"
|
|
|
|
| 569 |
|
| 570 |
+
return out
|
|
|
|
|
|
|
| 571 |
|
| 572 |
|
| 573 |
async def search_openapi_handler(arguments: dict[str, Any]) -> tuple[str, bool]:
|
| 574 |
+
"""Search HuggingFace OpenAPI specification by tag."""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 575 |
tag = arguments.get("tag", "")
|
|
|
|
| 576 |
if not tag:
|
| 577 |
return "Error: No tag provided", False
|
| 578 |
|
| 579 |
try:
|
|
|
|
| 580 |
spec = await _fetch_openapi_spec()
|
| 581 |
+
paths = spec.get("paths", {})
|
| 582 |
+
servers = spec.get("servers", [])
|
| 583 |
+
base_url = (
|
| 584 |
+
servers[0].get("url", "https://huggingface.co")
|
| 585 |
+
if servers
|
| 586 |
+
else "https://huggingface.co"
|
| 587 |
+
)
|
| 588 |
|
| 589 |
+
results = []
|
| 590 |
+
for path, path_item in paths.items():
|
| 591 |
+
for method, op in path_item.items():
|
| 592 |
+
if method not in [
|
| 593 |
+
"get",
|
| 594 |
+
"post",
|
| 595 |
+
"put",
|
| 596 |
+
"delete",
|
| 597 |
+
"patch",
|
| 598 |
+
"head",
|
| 599 |
+
"options",
|
| 600 |
+
]:
|
| 601 |
+
continue
|
| 602 |
+
if tag not in op.get("tags", []):
|
| 603 |
+
continue
|
| 604 |
|
| 605 |
+
results.append(
|
| 606 |
+
{
|
| 607 |
+
"path": path,
|
| 608 |
+
"method": method.upper(),
|
| 609 |
+
"operationId": op.get("operationId", ""),
|
| 610 |
+
"summary": op.get("summary", ""),
|
| 611 |
+
"description": op.get("description", ""),
|
| 612 |
+
"parameters": op.get("parameters", []),
|
| 613 |
+
"request_body": op.get("requestBody", {}),
|
| 614 |
+
"responses": op.get("responses", {}),
|
| 615 |
+
"base_url": base_url,
|
| 616 |
+
}
|
| 617 |
+
)
|
| 618 |
|
| 619 |
+
return _format_openapi_results(results, tag), True
|
| 620 |
|
| 621 |
except httpx.HTTPStatusError as e:
|
| 622 |
return f"HTTP error fetching OpenAPI spec: {e.response.status_code}", False
|
|
|
|
| 626 |
return f"Error searching OpenAPI spec: {str(e)}", False
|
| 627 |
|
| 628 |
|
| 629 |
+
async def _get_api_search_tool_spec() -> dict[str, Any]:
|
| 630 |
+
"""Generate OpenAPI tool spec with tags populated at runtime."""
|
| 631 |
+
spec = await _fetch_openapi_spec()
|
| 632 |
+
tags = _extract_all_tags(spec)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 633 |
|
| 634 |
+
return {
|
| 635 |
+
"name": "search_hf_api_endpoints",
|
| 636 |
+
"description": (
|
| 637 |
+
"Search HuggingFace OpenAPI specification by tag to find API endpoints with curl examples. "
|
| 638 |
+
"**Use when:** (1) Need to interact with HF Hub API directly, (2) Building scripts for repo operations, "
|
| 639 |
+
"(3) Need authentication patterns, (4) Understanding API parameters and responses, "
|
| 640 |
+
"(5) Need curl examples for HTTP requests. "
|
| 641 |
+
"Returns: Endpoint paths, methods, parameters, curl examples with authentication, and response schemas. "
|
| 642 |
+
"Tags group related operations: repos, models, datasets, inference, spaces, etc."
|
| 643 |
+
),
|
| 644 |
+
"parameters": {
|
| 645 |
+
"type": "object",
|
| 646 |
+
"properties": {
|
| 647 |
+
"tag": {
|
| 648 |
+
"type": "string",
|
| 649 |
+
"enum": tags,
|
| 650 |
+
"description": "The API tag to search for. Each tag groups related API endpoints.",
|
| 651 |
+
},
|
| 652 |
+
},
|
| 653 |
+
"required": ["tag"],
|
| 654 |
+
},
|
| 655 |
+
}
|
| 656 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 657 |
|
| 658 |
+
# ---------------------------------------------------------------------------
|
| 659 |
+
# Tool Specifications
|
| 660 |
+
# ---------------------------------------------------------------------------
|
| 661 |
|
| 662 |
+
DOC_ENDPOINTS = [
|
| 663 |
+
"hub",
|
| 664 |
+
"transformers",
|
| 665 |
+
"diffusers",
|
| 666 |
+
"datasets",
|
| 667 |
+
"gradio",
|
| 668 |
+
"trackio",
|
| 669 |
+
"smolagents",
|
| 670 |
+
"huggingface_hub",
|
| 671 |
+
"huggingface.js",
|
| 672 |
+
"transformers.js",
|
| 673 |
+
"inference-providers",
|
| 674 |
+
"inference-endpoints",
|
| 675 |
+
"peft",
|
| 676 |
+
"accelerate",
|
| 677 |
+
"optimum",
|
| 678 |
+
"tokenizers",
|
| 679 |
+
"courses",
|
| 680 |
+
"evaluate",
|
| 681 |
+
"tasks",
|
| 682 |
+
"dataset-viewer",
|
| 683 |
+
"trl",
|
| 684 |
+
"simulate",
|
| 685 |
+
"sagemaker",
|
| 686 |
+
"timm",
|
| 687 |
+
"safetensors",
|
| 688 |
+
"tgi",
|
| 689 |
+
"setfit",
|
| 690 |
+
"lerobot",
|
| 691 |
+
"autotrain",
|
| 692 |
+
"tei",
|
| 693 |
+
"bitsandbytes",
|
| 694 |
+
"sentence_transformers",
|
| 695 |
+
"chat-ui",
|
| 696 |
+
"leaderboards",
|
| 697 |
+
"lighteval",
|
| 698 |
+
"argilla",
|
| 699 |
+
"distilabel",
|
| 700 |
+
"microsoft-azure",
|
| 701 |
+
"kernels",
|
| 702 |
+
"google-cloud",
|
| 703 |
+
]
|
| 704 |
|
| 705 |
EXPLORE_HF_DOCS_TOOL_SPEC = {
|
| 706 |
"name": "explore_hf_docs",
|
|
|
|
| 722 |
"properties": {
|
| 723 |
"endpoint": {
|
| 724 |
"type": "string",
|
| 725 |
+
"enum": DOC_ENDPOINTS,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 726 |
"description": (
|
| 727 |
"The documentation endpoint to explore. Each endpoint corresponds to a major section of the Hugging Face documentation:\n\n"
|
| 728 |
"• courses — All Hugging Face courses (LLM, robotics, MCP, smol (llm training), agents, deep RL, computer vision, games, diffusion, 3D, audio) and the cookbook recipes. Probably the best place for examples.\n"
|
|
|
|
| 770 |
"query": {
|
| 771 |
"type": "string",
|
| 772 |
"description": (
|
| 773 |
+
"Optional keyword query to rank and filter documentation pages. "
|
| 774 |
+
"For Gradio, use concise queries like 'how to use the image component' or 'audio component demo'."
|
| 775 |
),
|
| 776 |
},
|
| 777 |
"max_results": {
|
| 778 |
"type": "integer",
|
| 779 |
+
"description": "Max results (default 20, max 50). Ignored for Gradio.",
|
|
|
|
|
|
|
| 780 |
"minimum": 1,
|
| 781 |
"maximum": 50,
|
| 782 |
},
|
|
|
|
| 814 |
"required": ["url"],
|
| 815 |
},
|
| 816 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
test_dataset_tools.py
DELETED
|
@@ -1,89 +0,0 @@
|
|
| 1 |
-
"""
|
| 2 |
-
Test script for hf_repo_files and hf_repo_git tools
|
| 3 |
-
"""
|
| 4 |
-
|
| 5 |
-
import asyncio
|
| 6 |
-
import sys
|
| 7 |
-
from typing import TypedDict
|
| 8 |
-
from unittest.mock import MagicMock
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
# Mock the types module before importing
|
| 12 |
-
class ToolResult(TypedDict, total=False):
|
| 13 |
-
formatted: str
|
| 14 |
-
totalResults: int
|
| 15 |
-
resultsShared: int
|
| 16 |
-
isError: bool
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
mock_types = MagicMock()
|
| 20 |
-
mock_types.ToolResult = ToolResult
|
| 21 |
-
sys.modules["agent.tools.types"] = mock_types
|
| 22 |
-
|
| 23 |
-
from agent.tools.hf_repo_files_tool import HfRepoFilesTool
|
| 24 |
-
from agent.tools.hf_repo_git_tool import HfRepoGitTool
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
async def test_hf_repo_files():
|
| 28 |
-
"""Test hf_repo_files tool"""
|
| 29 |
-
print("=" * 60)
|
| 30 |
-
print("Testing hf_repo_files")
|
| 31 |
-
print("=" * 60)
|
| 32 |
-
|
| 33 |
-
tool = HfRepoFilesTool()
|
| 34 |
-
|
| 35 |
-
# Test list
|
| 36 |
-
print("\n→ list files in gpt2:")
|
| 37 |
-
result = await tool.execute(
|
| 38 |
-
{"operation": "list", "repo_id": "openai-community/gpt2"}
|
| 39 |
-
)
|
| 40 |
-
print(f" isError: {result.get('isError', False)}")
|
| 41 |
-
print(f" totalResults: {result['totalResults']}")
|
| 42 |
-
# Just show first few lines
|
| 43 |
-
lines = result["formatted"].split("\n")
|
| 44 |
-
print(" Output (first 5 lines):\n" + "\n".join(f" {line}" for line in lines))
|
| 45 |
-
|
| 46 |
-
# Test read
|
| 47 |
-
print("\n→ read config.json from gpt2:")
|
| 48 |
-
result = await tool.execute(
|
| 49 |
-
{"operation": "read", "repo_id": "openai-community/gpt2", "path": "config.json"}
|
| 50 |
-
)
|
| 51 |
-
print(f" isError: {result.get('isError', False)}")
|
| 52 |
-
lines = result["formatted"].split("\n")
|
| 53 |
-
print(" Output (first 10 lines):\n" + "\n".join(f" {line}" for line in lines))
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
async def test_hf_repo_git():
|
| 57 |
-
"""Test hf_repo_git tool"""
|
| 58 |
-
print("\n" + "=" * 60)
|
| 59 |
-
print("Testing hf_repo_git")
|
| 60 |
-
print("=" * 60)
|
| 61 |
-
|
| 62 |
-
tool = HfRepoGitTool()
|
| 63 |
-
|
| 64 |
-
# Test list_refs
|
| 65 |
-
print("\n→ list_refs for gpt2:")
|
| 66 |
-
result = await tool.execute(
|
| 67 |
-
{"operation": "list_refs", "repo_id": "openai-community/gpt2"}
|
| 68 |
-
)
|
| 69 |
-
print(f" isError: {result.get('isError', False)}")
|
| 70 |
-
print(
|
| 71 |
-
" Output:\n"
|
| 72 |
-
+ "\n".join(f" {line}" for line in result["formatted"].split("\n"))
|
| 73 |
-
)
|
| 74 |
-
|
| 75 |
-
# Test help (no operation)
|
| 76 |
-
print("\n→ help (no operation):")
|
| 77 |
-
result = await tool.execute({})
|
| 78 |
-
print(f" isError: {result.get('isError', False)}")
|
| 79 |
-
lines = result["formatted"].split("\n")[:6]
|
| 80 |
-
print(" Output (first 6 lines):\n" + "\n".join(f" {line}" for line in lines))
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
if __name__ == "__main__":
|
| 84 |
-
print("\nHF Repo Tools Test\n")
|
| 85 |
-
asyncio.run(test_hf_repo_files())
|
| 86 |
-
asyncio.run(test_hf_repo_git())
|
| 87 |
-
print("\n" + "=" * 60)
|
| 88 |
-
print("Tests complete!")
|
| 89 |
-
print("=" * 60)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|