File size: 14,279 Bytes
21ff762 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 | #!/usr/bin/env python3
"""
mcp_entity.py -- Frozen ``McpRecord`` dataclass for the MCP-server catalog.
``ctx`` already indexes skills and agents as first-class entities in the
knowledge graph. MCP servers are the third subject type; this module is
the Phase-1 schema they are normalised into before they ever touch disk.
Design notes
------------
* The record is ``frozen=True`` with ``tuple`` collections so the same
instance can be safely shared across the fetcher, the dedup layer, and
the graph builder without defensive copies at every hop.
* ``from_dict`` is the sole entry point for raw fetcher payloads. It is
tolerant of garbage (unknown transports, messy slugs, short-ref GitHub
URLs) and fails loudly only when the slug is unrecoverable — that is
the one field downstream code cannot repair.
* No I/O. The module is pure so it can be unit-tested without a network,
a filesystem, or the rest of the ``ctx`` runtime wiring.
See ``src/wiki_utils.SAFE_NAME_RE`` for the legacy skill-name pattern;
MCP slugs enforce the stricter Tier-2 contract (lowercase + hyphens
only) used by ``skill_add_detector.validate_user_supplied_slug``.
"""
from __future__ import annotations
import copy
import re
from dataclasses import dataclass, field
from pathlib import Path
from typing import Any
__all__ = [
"ALLOWED_TRANSPORTS",
"MCP_SLUG_RE",
"McpRecord",
"canonicalize_github_url",
"normalize_slug",
]
# Tier-2 slug contract: lowercase, hyphens only, no leading/trailing
# hyphen, no consecutive hyphens. Mirrors the wiki's stricter hook-side
# validator so MCP entries are safe to use as filesystem paths and
# graph node ids without further escaping.
MCP_SLUG_RE = re.compile(r"^[a-z0-9]+(-[a-z0-9]+)*$")
# Subset of the MCP spec's transport tags we consider meaningful. Any
# fetcher input outside this set is silently dropped — see
# ``_normalize_transports``.
ALLOWED_TRANSPORTS: frozenset[str] = frozenset(
{"stdio", "http", "sse", "websocket"}
)
# GitHub URL matcher. Accepts http/https, optional ``www.``, trailing
# ``.git`` / slash, and captures org + repo for canonical reassembly.
# Host + scheme are matched case-insensitively (RFC 3986 §3.1 / §3.2.2);
# the org/repo path is preserved verbatim so display case round-trips.
_GITHUB_URL_RE = re.compile(
r"^(?:https?://)?(?:www\.)?github\.com/"
r"(?P<org>[A-Za-z0-9][A-Za-z0-9._-]*)/"
r"(?P<repo>[A-Za-z0-9][A-Za-z0-9._-]*?)"
r"(?:\.git)?/?$",
re.IGNORECASE,
)
# Short-ref matcher (``Org/Repo``) — only considered when the input
# contains no scheme and exactly one ``/``. Kept separate from the URL
# matcher so a stray ``/`` in free-text URLs doesn't falsely match.
_GITHUB_SHORT_REF_RE = re.compile(
r"^(?P<org>[A-Za-z0-9][A-Za-z0-9._-]*)/"
r"(?P<repo>[A-Za-z0-9][A-Za-z0-9._-]*?)"
r"(?:\.git)?/?$"
)
_DESCRIPTION_MAX = 300
_DESCRIPTION_TRUNCATE_AT = 297
_DESCRIPTION_FALLBACK = "No description available."
def normalize_slug(raw: str) -> str:
"""Normalize a free-text name/slug to ``[a-z0-9]+(-[a-z0-9]+)*``.
Lowercases, collapses any run of non-alphanumeric characters to a
single hyphen, and strips leading/trailing hyphens. Raises
``ValueError`` if nothing usable survives — downstream code cannot
invent a slug, so failing here is the correct behaviour.
"""
if not isinstance(raw, str):
raise ValueError(f"slug must be str, got {type(raw).__name__}")
lowered = raw.strip().lower()
collapsed = re.sub(r"[^a-z0-9]+", "-", lowered).strip("-")
if not collapsed:
raise ValueError(f"slug normalization produced empty string from {raw!r}")
# The collapse already guarantees the pattern, but re-assert so any
# future regex change fails loudly rather than silently accepting
# malformed output.
if not MCP_SLUG_RE.match(collapsed):
raise ValueError(f"slug {collapsed!r} does not match {MCP_SLUG_RE.pattern}")
return collapsed
def canonicalize_github_url(raw: str | None) -> str | None:
"""Canonicalize a GitHub URL to ``https://github.com/<org>/<repo>``.
Accepts full URLs (``https://github.com/Org/Repo``), scheme-less
variants (``github.com/Org/Repo.git``), and short refs
(``Org/Repo``). Returns ``None`` when the input is ``None``, empty,
or not recognisable as a GitHub reference.
Case in the org/repo path is preserved — GitHub is case-insensitive
for routing but users expect display case to round-trip. Dedup is
handled separately by :meth:`McpRecord.canonical_dedup_key`.
"""
if raw is None:
return None
candidate = raw.strip()
if not candidate:
return None
m = _GITHUB_URL_RE.match(candidate)
if m is None and "://" not in candidate and candidate.count("/") == 1:
m = _GITHUB_SHORT_REF_RE.match(candidate)
if m is None:
return None
return f"https://github.com/{m.group('org')}/{m.group('repo')}"
def _normalize_tags(raw: object) -> tuple[str, ...]:
"""Dedupe, lowercase, sort. Fall back to ``('uncategorized',)`` if empty."""
if not raw:
return ("uncategorized",)
if isinstance(raw, str):
# Tolerate a single comma-separated string from scruffier fetchers.
items: list[str] = [p for p in (s.strip() for s in raw.split(",")) if p]
elif isinstance(raw, (list, tuple, set, frozenset)):
items = [str(t).strip() for t in raw if str(t).strip()]
else:
return ("uncategorized",)
cleaned = {t.lower() for t in items if t}
if not cleaned:
return ("uncategorized",)
return tuple(sorted(cleaned))
def _normalize_transports(raw: object) -> tuple[str, ...]:
"""Filter to ``ALLOWED_TRANSPORTS``, lowercase, dedupe, sort."""
if not raw:
return ()
if isinstance(raw, str):
items: list[str] = [p for p in (s.strip() for s in raw.split(",")) if p]
elif isinstance(raw, (list, tuple, set, frozenset)):
items = [str(t).strip() for t in raw if str(t).strip()]
else:
return ()
kept = {t.lower() for t in items if t.lower() in ALLOWED_TRANSPORTS}
return tuple(sorted(kept))
def _normalize_description(raw: object) -> str:
"""Trim, fall back to placeholder when empty, truncate to 300 chars."""
if not isinstance(raw, str):
return _DESCRIPTION_FALLBACK
trimmed = raw.strip()
if not trimmed:
return _DESCRIPTION_FALLBACK
if len(trimmed) > _DESCRIPTION_MAX:
return trimmed[:_DESCRIPTION_TRUNCATE_AT] + "..."
return trimmed
def _normalize_sources(raw: object) -> tuple[str, ...]:
"""Dedupe + sort sources. Empty input yields an empty tuple."""
if not raw:
return ()
if isinstance(raw, str):
items: list[str] = [p for p in (s.strip() for s in raw.split(",")) if p]
elif isinstance(raw, (list, tuple, set, frozenset)):
items = [str(s).strip() for s in raw if str(s).strip()]
else:
return ()
cleaned = {s for s in items if s}
return tuple(sorted(cleaned))
def _optional_lower(raw: object) -> str | None:
if raw is None:
return None
if not isinstance(raw, str):
return None
trimmed = raw.strip().lower()
return trimmed or None
def _optional_str(raw: object) -> str | None:
if raw is None:
return None
if not isinstance(raw, str):
return None
trimmed = raw.strip()
return trimmed or None
def _optional_author_type(raw: object) -> str | None:
val = _optional_lower(raw)
if val in {"org", "user"}:
return val
return None
def _optional_int(raw: object) -> int | None:
if raw is None:
return None
if isinstance(raw, bool):
# ``bool`` is an ``int`` subclass; treat it as not-a-count.
return None
if isinstance(raw, int):
return raw
if isinstance(raw, str):
stripped = raw.strip()
if stripped.isdigit() or (stripped.startswith("-") and stripped[1:].isdigit()):
try:
return int(stripped)
except ValueError: # pragma: no cover — isdigit() guarded
return None
return None
@dataclass(frozen=True)
class McpRecord:
"""Normalized MCP-server record built from a fetcher payload.
All collection fields are ``tuple`` to preserve the ``frozen=True``
contract. ``raw`` is retained for debugging and is deep-copied on
construction so callers cannot mutate the stored payload after the
fact; it is never written to the entity page frontmatter.
"""
# Required
slug: str
name: str
description: str
sources: tuple[str, ...]
# Provenance
github_url: str | None
homepage_url: str | None
# Classification
tags: tuple[str, ...]
transports: tuple[str, ...]
language: str | None
license: str | None
# Authorship
author: str | None
author_type: str | None
# Metrics — enriched later; None at initial fetch.
stars: int | None
last_commit_at: str | None
# Raw payload — kept for debugging, never serialized to entity body.
raw: dict[str, Any] = field(default_factory=dict)
@classmethod
def from_dict(cls, data: dict[str, Any]) -> McpRecord:
"""Build an ``McpRecord`` from a fetcher's raw dict.
Normalizes slug (required), description, sources, tags, and
transports; canonicalizes the GitHub URL; coerces optional
string fields. Raises ``ValueError`` if the slug cannot be
recovered from either ``slug`` or ``name``.
"""
if not isinstance(data, dict):
raise ValueError(f"from_dict expected dict, got {type(data).__name__}")
# Prefer an explicit slug; fall back to name. Both are fed
# through the same normalizer so the canonical form is
# identical regardless of which field the fetcher populated.
raw_slug = data.get("slug") or data.get("name") or ""
if not isinstance(raw_slug, str) or not raw_slug.strip():
raise ValueError("McpRecord requires non-empty 'slug' or 'name'")
slug = normalize_slug(raw_slug)
raw_name = data.get("name")
if isinstance(raw_name, str) and raw_name.strip():
name = raw_name.strip()
else:
# Fetcher didn't supply a display name — use the slug so the
# record is still renderable.
name = slug
description = _normalize_description(data.get("description"))
sources = _normalize_sources(data.get("sources"))
github_url = canonicalize_github_url(
data.get("github_url") if isinstance(data.get("github_url"), str) else None
)
homepage_url = _optional_str(data.get("homepage_url"))
tags = _normalize_tags(data.get("tags"))
transports = _normalize_transports(data.get("transports"))
language = _optional_lower(data.get("language"))
license_ = _optional_str(data.get("license"))
author = _optional_str(data.get("author"))
author_type = _optional_author_type(data.get("author_type"))
stars = _optional_int(data.get("stars"))
last_commit_at = _optional_str(data.get("last_commit_at"))
raw_payload = data.get("raw")
if isinstance(raw_payload, dict):
raw_copy: dict[str, Any] = copy.deepcopy(raw_payload)
else:
# No explicit ``raw`` key: snapshot the entire input so
# debugging has the original fetcher output available.
raw_copy = copy.deepcopy(data)
return cls(
slug=slug,
name=name,
description=description,
sources=sources,
github_url=github_url,
homepage_url=homepage_url,
tags=tags,
transports=transports,
language=language,
license=license_,
author=author,
author_type=author_type,
stars=stars,
last_commit_at=last_commit_at,
raw=raw_copy,
)
def to_frontmatter(self) -> dict[str, Any]:
"""Return a dict ready for YAML frontmatter serialization.
Excludes ``raw``. Includes ``type: mcp-server`` and ``created`` /
``updated`` placeholders (both ``None``) so the caller — which
owns wall-clock time — can fill them without re-shaping the
dict. Lists are emitted as ``list`` (YAML-friendly) rather than
the internal ``tuple`` representation.
"""
return {
"type": "mcp-server",
"slug": self.slug,
"name": self.name,
"description": self.description,
"sources": list(self.sources),
"github_url": self.github_url,
"homepage_url": self.homepage_url,
"tags": list(self.tags),
"transports": list(self.transports),
"language": self.language,
"license": self.license,
"author": self.author,
"author_type": self.author_type,
"stars": self.stars,
"last_commit_at": self.last_commit_at,
"created": None,
"updated": None,
}
def entity_relpath(self) -> Path:
"""Return ``Path('<shard>/<slug>.md')``.
``shard`` is ``slug[0]`` for alphabetic slugs and the literal
``'0-9'`` when the slug starts with a digit. Sharding keeps the
entity directory from growing into a single multi-thousand-file
mess as the MCP catalog fills in.
"""
first = self.slug[0]
shard = "0-9" if first.isdigit() else first
return Path(shard) / f"{self.slug}.md"
def canonical_dedup_key(self) -> str:
"""Return the key used to detect duplicates across sources.
When a GitHub URL is present we strip trailing slashes and
lowercase the entire URL so ``https://github.com/Org/Repo`` and
``https://GitHub.com/org/repo/`` collapse to the same key.
Otherwise we fall back to ``'slug:' + self.slug`` so records
without a repo can still be deduplicated by their normalized
slug.
"""
if self.github_url:
return self.github_url.rstrip("/").lower()
return f"slug:{self.slug}"
|