Manuel GanterandClaude Opus 5
1cb6369282
Ports two changes from the knowledge-base vault this skeleton is extracted from, keeping the shared scripts byte-identical across the pair: - `graph_summary` reports `by_tag` and `tag_count`, sized by `topTags`, and drops the `by_type`/`by_status` breakdowns. An MCP-only agent could not otherwise discover the tag spellings that `graph_search_nodes` matches verbatim. - `generate-graph.py --validate` now enforces a controlled vocabulary for a note's `type`, which is where the drift signal `by_type` used to give now lives. `AGENT_QUERY_GUIDE.md`, both skills and the-knowing-one quote the enforced enum. `tests/test_generate_graph.py` is new; `make test` is 50 tests. Regenerating the graph also drops a stale `GLOBAL_AGENTS` node for a file this skeleton does not carry. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
1033 lines
37 KiB
Python
Executable File
1033 lines
37 KiB
Python
Executable File
#!/usr/bin/env python3
|
|
"""MCP server for the knowledge-base repository.
|
|
|
|
This server is the single interface an agent needs to work the vault over stdio
|
|
with JSON-RPC 2.0. It exposes graph and note *querying* (graph_summary,
|
|
graph_search_nodes, graph_neighbors, read_note, refresh_graph), *writing*
|
|
(write_note for text, capture_raw for verbatim bytes from a local file or an
|
|
http(s) download) and *publishing* (edit_note, git_commit, git_push), so an
|
|
agent without shell or filesystem tools can ingest a resource end to end.
|
|
|
|
Every write is confined to the repository root, and staging is always explicit:
|
|
paths are named one by one, never `.` or `-A`.
|
|
|
|
It auto-detects the transport layer:
|
|
- Content-Length framing (MCP spec, used by most clients)
|
|
- Newline-delimited JSON (used by Claude Code ≥ 2.1)
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import hashlib
|
|
import json
|
|
import shutil
|
|
import subprocess
|
|
import sys
|
|
import urllib.error
|
|
import urllib.request
|
|
from collections import Counter
|
|
from dataclasses import dataclass
|
|
from pathlib import Path
|
|
from typing import Any, Dict, List
|
|
|
|
|
|
SERVER_NAME = "knowledge-base-mcp"
|
|
SERVER_VERSION = "0.1.0"
|
|
|
|
|
|
@dataclass
|
|
class AppState:
|
|
root: Path
|
|
graph_path: Path
|
|
|
|
|
|
class _StdinReader:
|
|
"""Buffered stdin reader that supports transport auto-detection."""
|
|
|
|
def __init__(self, raw: Any) -> None:
|
|
self._raw = raw
|
|
self._prefix: bytes = b""
|
|
|
|
def read(self, n: int) -> bytes:
|
|
if self._prefix:
|
|
chunk, self._prefix = self._prefix[:n], self._prefix[n:]
|
|
remaining = n - len(chunk)
|
|
return chunk + (self._raw.read(remaining) if remaining else b"")
|
|
return self._raw.read(n)
|
|
|
|
def readline(self) -> bytes:
|
|
if self._prefix:
|
|
nl = self._prefix.find(b"\n")
|
|
if nl != -1:
|
|
line, self._prefix = self._prefix[: nl + 1], self._prefix[nl + 1 :]
|
|
return line
|
|
prefix, self._prefix = self._prefix, b""
|
|
return prefix + self._raw.readline()
|
|
return self._raw.readline()
|
|
|
|
def peek_byte(self) -> bytes:
|
|
b = self._raw.read(1)
|
|
if b:
|
|
self._prefix = b + self._prefix
|
|
return b
|
|
|
|
|
|
_WHITESPACE_BYTES = {b"\r", b"\n", b"\t", b" "}
|
|
|
|
|
|
def _detect_transport(reader: _StdinReader) -> str:
|
|
"""Peek at the first non-whitespace byte to decide 'ndjson' vs 'framed'.
|
|
|
|
Leading ASCII whitespace bytes are discarded — they corrupt neither JSON
|
|
nor Content-Length headers, and some client wrappers emit stray newlines
|
|
before the first message. EOF before any non-whitespace defaults to
|
|
'ndjson' so the read loop exits cleanly on an empty stream.
|
|
"""
|
|
while True:
|
|
b = reader._raw.read(1)
|
|
if not b:
|
|
return "ndjson"
|
|
if b not in _WHITESPACE_BYTES:
|
|
reader._prefix = b + reader._prefix
|
|
return "ndjson" if b == b"{" else "framed"
|
|
|
|
|
|
def _read_message_framed(reader: _StdinReader) -> Dict[str, Any] | None:
|
|
headers: Dict[str, str] = {}
|
|
while True:
|
|
line = reader.readline()
|
|
if not line:
|
|
return None
|
|
if line in (b"\r\n", b"\n"):
|
|
break
|
|
key, _, value = line.decode("utf-8").partition(":")
|
|
headers[key.strip().lower()] = value.strip()
|
|
|
|
length_raw = headers.get("content-length")
|
|
if not length_raw:
|
|
raise ValueError("missing Content-Length header")
|
|
length = int(length_raw)
|
|
payload = reader.read(length)
|
|
if len(payload) != length:
|
|
raise ValueError("incomplete payload")
|
|
return json.loads(payload.decode("utf-8"))
|
|
|
|
|
|
def _read_message_ndjson(reader: _StdinReader) -> Dict[str, Any] | None:
|
|
while True:
|
|
line = reader.readline()
|
|
if not line:
|
|
return None
|
|
stripped = line.strip()
|
|
if not stripped:
|
|
continue
|
|
return json.loads(stripped.decode("utf-8"))
|
|
|
|
|
|
def _write_message_framed(stdout: Any, payload: Dict[str, Any]) -> None:
|
|
data = json.dumps(payload, ensure_ascii=True).encode("utf-8")
|
|
stdout.buffer.write(f"Content-Length: {len(data)}\r\n\r\n".encode("utf-8"))
|
|
stdout.buffer.write(data)
|
|
stdout.buffer.flush()
|
|
|
|
|
|
def _write_message_ndjson(stdout: Any, payload: Dict[str, Any]) -> None:
|
|
data = json.dumps(payload, ensure_ascii=True).encode("utf-8") + b"\n"
|
|
stdout.buffer.write(data)
|
|
stdout.buffer.flush()
|
|
|
|
|
|
|
|
def _response(message_id: Any, result: Dict[str, Any]) -> Dict[str, Any]:
|
|
return {"jsonrpc": "2.0", "id": message_id, "result": result}
|
|
|
|
|
|
def _error(message_id: Any, code: int, message: str) -> Dict[str, Any]:
|
|
return {
|
|
"jsonrpc": "2.0",
|
|
"id": message_id,
|
|
"error": {"code": code, "message": message},
|
|
}
|
|
|
|
|
|
def _load_graph(state: AppState) -> Dict[str, Any]:
|
|
if not state.graph_path.exists():
|
|
raise FileNotFoundError(
|
|
f"graph not found at {state.graph_path}. Run scripts/generate-graph.py first."
|
|
)
|
|
return json.loads(state.graph_path.read_text(encoding="utf-8"))
|
|
|
|
|
|
def _nodes_by_id(graph: Dict[str, Any]) -> Dict[str, Dict[str, Any]]:
|
|
nodes = graph.get("nodes", [])
|
|
return {str(node.get("id")): node for node in nodes if isinstance(node, dict)}
|
|
|
|
|
|
def _match_filters(node: Dict[str, Any], args: Dict[str, Any]) -> bool:
|
|
repo = args.get("repo")
|
|
note_type = args.get("type")
|
|
status = args.get("status")
|
|
tag = args.get("tag")
|
|
query = str(args.get("query", "")).strip().lower()
|
|
|
|
if repo and node.get("repo") != repo:
|
|
return False
|
|
if note_type and node.get("type") != note_type:
|
|
return False
|
|
if status and node.get("status") != status:
|
|
return False
|
|
if tag and tag not in (node.get("tags") or []):
|
|
return False
|
|
if query:
|
|
haystacks = [
|
|
str(node.get("id", "")).lower(),
|
|
str(node.get("title", "")).lower(),
|
|
str(node.get("path", "")).lower(),
|
|
" ".join(str(t).lower() for t in (node.get("tags") or [])),
|
|
]
|
|
if not any(query in h for h in haystacks):
|
|
return False
|
|
return True
|
|
|
|
|
|
_WILDCARD_PATHS = {".", "./", "-A", "--all", "*", "-a", ":/"}
|
|
|
|
# Ceiling for a single capture_raw call; papers and slide decks fit, a video
|
|
# download that would bloat the vault does not.
|
|
_DEFAULT_MAX_CAPTURE_BYTES = 100 * 1024 * 1024
|
|
_DOWNLOAD_TIMEOUT_SECONDS = 60
|
|
|
|
# graph_summary defaults: enough most-linked notes to spot the hubs, and enough
|
|
# tags to cover the vocabulary an agent actually filters on.
|
|
_DEFAULT_TOP_LINKED = 10
|
|
_DEFAULT_TOP_TAGS = 50
|
|
|
|
|
|
def _resolve_in_root(state: AppState, raw_path: Any, *, label: str = "path") -> Path:
|
|
"""Turn a vault-relative path into an absolute path inside the repo root.
|
|
|
|
Rejects empty values, absolute paths and anything escaping the root, so a
|
|
tool call can never touch a file outside the vault.
|
|
"""
|
|
value = str(raw_path or "").strip()
|
|
if not value:
|
|
raise ValueError(f"'{label}' is required")
|
|
candidate = Path(value)
|
|
if candidate.is_absolute():
|
|
raise ValueError(f"'{label}' must be vault-relative, got absolute '{value}'")
|
|
root = state.root.resolve()
|
|
target = (root / candidate).resolve()
|
|
if target != root and not target.is_relative_to(root):
|
|
raise ValueError(f"'{label}' resolves outside repository root: '{value}'")
|
|
return target
|
|
|
|
|
|
def _relative_to_root(state: AppState, target: Path) -> str:
|
|
return target.relative_to(state.root.resolve()).as_posix()
|
|
|
|
|
|
def _run_git(state: AppState, args: List[str]) -> subprocess.CompletedProcess:
|
|
return subprocess.run(
|
|
["git", *args],
|
|
cwd=state.root,
|
|
check=False,
|
|
capture_output=True,
|
|
text=True,
|
|
)
|
|
|
|
|
|
def _git_result(payload: Dict[str, Any], exit_code: int) -> Dict[str, Any]:
|
|
text = json.dumps(payload, indent=2)
|
|
if exit_code != 0:
|
|
return {
|
|
"isError": True,
|
|
"content": [{"type": "text", "text": text}],
|
|
"structuredContent": payload,
|
|
}
|
|
return _text_result(text, structured=payload)
|
|
|
|
|
|
def _tool_definitions() -> List[Dict[str, Any]]:
|
|
return [
|
|
{
|
|
"name": "graph_summary",
|
|
"description": (
|
|
"Orientation for a cold start: node and edge counts, the repo "
|
|
"scopes, the most-linked notes, and the tag vocabulary as "
|
|
"`by_tag` (most frequent first) plus `tag_count`, the number of "
|
|
"distinct tags. Read `by_tag` before filtering "
|
|
"`graph_search_nodes` by `tag` \u2014 that filter matches a tag "
|
|
"verbatim, so a guessed spelling returns nothing."
|
|
),
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"properties": {
|
|
"topLinked": {
|
|
"type": "integer",
|
|
"minimum": 1,
|
|
"maximum": 50,
|
|
"description": "Most-linked notes to return (default 10).",
|
|
},
|
|
"topTags": {
|
|
"type": "integer",
|
|
"minimum": 1,
|
|
"maximum": 500,
|
|
"description": "Tags to return, most frequent first (default 50).",
|
|
},
|
|
},
|
|
"additionalProperties": False,
|
|
},
|
|
},
|
|
{
|
|
"name": "graph_search_nodes",
|
|
"description": "Search graph nodes by metadata and free text.",
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"properties": {
|
|
"repo": {"type": "string"},
|
|
"type": {"type": "string"},
|
|
"status": {"type": "string"},
|
|
"tag": {"type": "string"},
|
|
"query": {"type": "string"},
|
|
"limit": {"type": "integer", "minimum": 1, "maximum": 200},
|
|
},
|
|
},
|
|
},
|
|
{
|
|
"name": "graph_neighbors",
|
|
"description": (
|
|
"Show inbound and outbound links for a node ID. Each neighbor "
|
|
"includes the edge `type` (relationship, e.g. supersedes, "
|
|
"depends-on), `anchor_text` (the link's display text), and "
|
|
"`annotation` (what to expect in that note)."
|
|
),
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"required": ["id"],
|
|
"properties": {
|
|
"id": {"type": "string"},
|
|
"limit": {"type": "integer", "minimum": 1, "maximum": 200},
|
|
},
|
|
},
|
|
},
|
|
{
|
|
"name": "read_note",
|
|
"description": (
|
|
"Read note content by graph node ID or vault-relative path. "
|
|
"Provide exactly one of `id` or `path`. "
|
|
"If both or neither are provided, the call returns an error."
|
|
),
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"properties": {
|
|
"id": {"type": "string", "description": "Graph node ID (e.g. 'AGENTS')."},
|
|
"path": {"type": "string", "description": "Vault-relative file path (e.g. 'README.md')."},
|
|
},
|
|
"additionalProperties": False,
|
|
},
|
|
},
|
|
{
|
|
"name": "refresh_graph",
|
|
"description": "Regenerate graph.json using scripts/generate-graph.py.",
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"properties": {
|
|
"validate": {"type": "boolean"},
|
|
},
|
|
},
|
|
},
|
|
{
|
|
"name": "write_note",
|
|
"description": (
|
|
"Create a note (or any text file) at a vault-relative path, "
|
|
"creating parent directories as needed. Fails if the file "
|
|
"exists unless `overwrite` is true. The generated "
|
|
"`index/graph.json` is off limits — use `refresh_graph`."
|
|
),
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"required": ["path", "content"],
|
|
"properties": {
|
|
"path": {
|
|
"type": "string",
|
|
"description": "Vault-relative path, e.g. '03_resources/demo/index.md'.",
|
|
},
|
|
"content": {"type": "string", "description": "Full file content."},
|
|
"overwrite": {
|
|
"type": "boolean",
|
|
"description": "Replace an existing file (default false).",
|
|
},
|
|
},
|
|
"additionalProperties": False,
|
|
},
|
|
},
|
|
{
|
|
"name": "edit_note",
|
|
"description": (
|
|
"Replace an exact string inside an existing note. `old_string` "
|
|
"must occur exactly once unless `replace_all` is true."
|
|
),
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"required": ["path", "old_string", "new_string"],
|
|
"properties": {
|
|
"path": {"type": "string", "description": "Vault-relative path of the note."},
|
|
"old_string": {"type": "string", "description": "Text to replace, matched verbatim."},
|
|
"new_string": {"type": "string", "description": "Replacement text."},
|
|
"replace_all": {
|
|
"type": "boolean",
|
|
"description": "Replace every occurrence (default false).",
|
|
},
|
|
},
|
|
"additionalProperties": False,
|
|
},
|
|
},
|
|
{
|
|
"name": "capture_raw",
|
|
"description": (
|
|
"Copy a local file or download an http(s) URL into the vault "
|
|
"byte-for-byte — the way an original PDF, image, or archive is "
|
|
"preserved under `03_resources/<slug>/raw/`. Provide exactly one "
|
|
"of `source_path` (absolute or CWD-relative path on this machine) "
|
|
"or `url`. Binary-safe; returns size and sha256. Read the "
|
|
"captured file back with `read_note`, which reports binary files "
|
|
"with an absolute path to open with a file-reading tool."
|
|
),
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"required": ["path"],
|
|
"properties": {
|
|
"path": {
|
|
"type": "string",
|
|
"description": "Vault-relative destination, e.g. '03_resources/demo/raw/paper.pdf'.",
|
|
},
|
|
"source_path": {
|
|
"type": "string",
|
|
"description": "Path of the local file to copy.",
|
|
},
|
|
"url": {
|
|
"type": "string",
|
|
"description": "http(s) URL to download verbatim.",
|
|
},
|
|
"overwrite": {
|
|
"type": "boolean",
|
|
"description": "Replace an existing capture (default false).",
|
|
},
|
|
"max_bytes": {
|
|
"type": "integer",
|
|
"minimum": 1,
|
|
"description": "Reject captures larger than this (default 100 MiB).",
|
|
},
|
|
},
|
|
"additionalProperties": False,
|
|
},
|
|
},
|
|
{
|
|
"name": "git_commit",
|
|
"description": (
|
|
"Stage the named vault-relative paths individually and commit "
|
|
"them. `paths` must list every path explicitly — wildcard "
|
|
"staging ('.', '-A', '*') is rejected. Returns an error result "
|
|
"if git refuses the commit (e.g. nothing staged)."
|
|
),
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"required": ["message", "paths"],
|
|
"properties": {
|
|
"message": {"type": "string", "description": "Commit message (Conventional Commits)."},
|
|
"paths": {
|
|
"type": "array",
|
|
"minItems": 1,
|
|
"items": {"type": "string"},
|
|
"description": "Vault-relative paths to stage, one entry per file or directory.",
|
|
},
|
|
},
|
|
"additionalProperties": False,
|
|
},
|
|
},
|
|
{
|
|
"name": "git_push",
|
|
"description": (
|
|
"Push commits to a remote. Defaults to 'origin' and the current "
|
|
"branch. Returns an error result if git refuses the push."
|
|
),
|
|
"inputSchema": {
|
|
"type": "object",
|
|
"properties": {
|
|
"remote": {"type": "string", "description": "Remote name (default 'origin')."},
|
|
"branch": {"type": "string", "description": "Branch to push (default: current branch)."},
|
|
"set_upstream": {
|
|
"type": "boolean",
|
|
"description": "Pass --set-upstream (default false).",
|
|
},
|
|
},
|
|
"additionalProperties": False,
|
|
},
|
|
},
|
|
]
|
|
|
|
|
|
def _text_result(text: str, structured: Dict[str, Any] | None = None) -> Dict[str, Any]:
|
|
result: Dict[str, Any] = {"content": [{"type": "text", "text": text}]}
|
|
if structured is not None:
|
|
result["structuredContent"] = structured
|
|
return result
|
|
|
|
|
|
def _tool_graph_summary(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
graph = _load_graph(state)
|
|
nodes = graph.get("nodes", [])
|
|
edges = graph.get("edges", [])
|
|
top_n = int(args.get("topLinked", _DEFAULT_TOP_LINKED))
|
|
top_tags = int(args.get("topTags", _DEFAULT_TOP_TAGS))
|
|
|
|
by_repo: Counter[str] = Counter(str(node.get("repo", "unknown")) for node in nodes)
|
|
tag_counts: Counter[str] = Counter(
|
|
str(tag) for node in nodes for tag in (node.get("tags") or [])
|
|
)
|
|
# Frequent tags first, alphabetical within a tie, so a truncated list is
|
|
# both the most useful slice and stable across calls.
|
|
ranked_tags = sorted(tag_counts.items(), key=lambda item: (-item[1], item[0]))
|
|
|
|
sorted_nodes = sorted(
|
|
nodes,
|
|
key=lambda n: int(n.get("inbound_count", 0)),
|
|
reverse=True,
|
|
)[:top_n]
|
|
payload = {
|
|
"version": graph.get("version"),
|
|
"generated_at": graph.get("generated_at"),
|
|
"node_count": len(nodes),
|
|
"edge_count": len(edges),
|
|
"by_repo": dict(by_repo),
|
|
"tag_count": len(tag_counts),
|
|
"by_tag": [
|
|
{"tag": tag, "count": count} for tag, count in ranked_tags[:top_tags]
|
|
],
|
|
"top_linked": [
|
|
{
|
|
"id": node.get("id"),
|
|
"title": node.get("title"),
|
|
"path": node.get("path"),
|
|
"inbound_count": node.get("inbound_count", 0),
|
|
}
|
|
for node in sorted_nodes
|
|
],
|
|
}
|
|
return _text_result(json.dumps(payload, indent=2), structured=payload)
|
|
|
|
|
|
def _tool_graph_search_nodes(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
graph = _load_graph(state)
|
|
nodes = graph.get("nodes", [])
|
|
limit = int(args.get("limit", 30))
|
|
filtered = [node for node in nodes if _match_filters(node, args)]
|
|
filtered.sort(key=lambda n: str(n.get("id", "")))
|
|
sliced = filtered[:limit]
|
|
payload = {
|
|
"total_matches": len(filtered),
|
|
"returned": len(sliced),
|
|
"items": sliced,
|
|
}
|
|
return _text_result(json.dumps(payload, indent=2), structured=payload)
|
|
|
|
|
|
def _tool_graph_neighbors(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
graph = _load_graph(state)
|
|
node_id = str(args["id"])
|
|
limit = int(args.get("limit", 50))
|
|
nodes = _nodes_by_id(graph)
|
|
if node_id not in nodes:
|
|
raise ValueError(f"node '{node_id}' not found")
|
|
|
|
edges = graph.get("edges", [])
|
|
outbound_edges = [e for e in edges if e.get("from") == node_id]
|
|
inbound_edges = [e for e in edges if e.get("to") == node_id]
|
|
|
|
def describe(edge_list: List[Dict[str, Any]], direction: str) -> List[Dict[str, Any]]:
|
|
# neighbor id is the "other end" relative to node_id
|
|
other_key = "to" if direction == "outbound" else "from"
|
|
result: List[Dict[str, Any]] = []
|
|
for edge in edge_list:
|
|
item_id = str(edge.get(other_key))
|
|
node = nodes.get(item_id)
|
|
entry = {
|
|
"id": node.get("id") if node else item_id,
|
|
"title": node.get("title") if node else "<missing>",
|
|
"path": node.get("path") if node else "",
|
|
"direction": direction,
|
|
"type": edge.get("type", "related"),
|
|
"anchor_text": edge.get("anchor_text", ""),
|
|
"annotation": edge.get("annotation", ""),
|
|
}
|
|
result.append(entry)
|
|
return result[:limit]
|
|
|
|
payload = {
|
|
"node": nodes[node_id],
|
|
"inbound_count": len(inbound_edges),
|
|
"outbound_count": len(outbound_edges),
|
|
"inbound": describe(inbound_edges, "inbound"),
|
|
"outbound": describe(outbound_edges, "outbound"),
|
|
}
|
|
return _text_result(json.dumps(payload, indent=2), structured=payload)
|
|
|
|
|
|
def _tool_read_note(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
has_id = bool(args.get("id"))
|
|
has_path = bool(args.get("path"))
|
|
|
|
if has_id and has_path:
|
|
raise ValueError("provide exactly one of 'id' or 'path', not both")
|
|
if not has_id and not has_path:
|
|
raise ValueError("exactly one of 'id' or 'path' is required")
|
|
|
|
path_value = args.get("path")
|
|
|
|
if has_id:
|
|
nodes = _nodes_by_id(_load_graph(state))
|
|
node = nodes.get(str(args["id"]))
|
|
if not node:
|
|
raise ValueError(f"node '{args['id']}' not found")
|
|
path_value = node.get("path")
|
|
|
|
target = (state.root / str(path_value)).resolve()
|
|
if not target.is_relative_to(state.root.resolve()):
|
|
raise ValueError("note path resolves outside repository root")
|
|
if not target.exists():
|
|
raise FileNotFoundError(f"note not found at {target}")
|
|
|
|
relative = str(target.relative_to(state.root).as_posix())
|
|
data = target.read_bytes()
|
|
try:
|
|
text = data.decode("utf-8")
|
|
except UnicodeDecodeError:
|
|
payload = {
|
|
"path": relative,
|
|
"binary": True,
|
|
"bytes": len(data),
|
|
"sha256": hashlib.sha256(data).hexdigest(),
|
|
"hint": (
|
|
f"{target} is not UTF-8 text (e.g. a PDF or image captured under "
|
|
"raw/). Open it with a file-reading tool that handles binaries; "
|
|
"this server returns text only."
|
|
),
|
|
}
|
|
return _text_result(json.dumps(payload, indent=2), structured=payload)
|
|
|
|
payload = {
|
|
"path": relative,
|
|
"binary": False,
|
|
"content": text,
|
|
}
|
|
return _text_result(json.dumps(payload, indent=2), structured=payload)
|
|
|
|
|
|
def _tool_refresh_graph(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
command = [sys.executable, "scripts/generate-graph.py"]
|
|
if bool(args.get("validate", False)):
|
|
command.append("--validate")
|
|
|
|
proc = subprocess.run(
|
|
command,
|
|
cwd=state.root,
|
|
check=False,
|
|
capture_output=True,
|
|
text=True,
|
|
)
|
|
|
|
payload = {
|
|
"command": command,
|
|
"exit_code": proc.returncode,
|
|
"stdout": proc.stdout,
|
|
"stderr": proc.stderr,
|
|
}
|
|
if proc.returncode != 0:
|
|
return {
|
|
"isError": True,
|
|
"content": [{"type": "text", "text": json.dumps(payload, indent=2)}],
|
|
"structuredContent": payload,
|
|
}
|
|
return _text_result(json.dumps(payload, indent=2), structured=payload)
|
|
|
|
|
|
def _tool_write_note(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
if "content" not in args:
|
|
raise ValueError("'content' is required")
|
|
content = str(args["content"])
|
|
target = _resolve_in_root(state, args.get("path"))
|
|
relative = _relative_to_root(state, target)
|
|
if relative == state.graph_path.relative_to(state.root.resolve()).as_posix():
|
|
raise ValueError(
|
|
"index/graph.json is generated \u2014 use refresh_graph instead of writing it"
|
|
)
|
|
|
|
existed = target.exists()
|
|
if existed and not bool(args.get("overwrite", False)):
|
|
raise FileExistsError(
|
|
f"note already exists at {relative}; pass overwrite=true to replace it"
|
|
)
|
|
if existed and target.is_dir():
|
|
raise ValueError(f"'{relative}' is a directory")
|
|
|
|
target.parent.mkdir(parents=True, exist_ok=True)
|
|
target.write_text(content, encoding="utf-8")
|
|
payload = {
|
|
"path": relative,
|
|
"created": not existed,
|
|
"bytes": len(content.encode("utf-8")),
|
|
}
|
|
return _text_result(json.dumps(payload, indent=2), structured=payload)
|
|
|
|
|
|
def _tool_edit_note(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
if "old_string" not in args or "new_string" not in args:
|
|
raise ValueError("'old_string' and 'new_string' are required")
|
|
old_string = str(args["old_string"])
|
|
new_string = str(args["new_string"])
|
|
if not old_string:
|
|
raise ValueError("'old_string' must not be empty")
|
|
if old_string == new_string:
|
|
raise ValueError("'old_string' and 'new_string' are identical")
|
|
|
|
target = _resolve_in_root(state, args.get("path"))
|
|
relative = _relative_to_root(state, target)
|
|
if not target.is_file():
|
|
raise FileNotFoundError(f"note not found at {relative}")
|
|
|
|
text = target.read_text(encoding="utf-8")
|
|
occurrences = text.count(old_string)
|
|
if occurrences == 0:
|
|
raise ValueError(f"'old_string' not found in {relative}")
|
|
replace_all = bool(args.get("replace_all", False))
|
|
if occurrences > 1 and not replace_all:
|
|
raise ValueError(
|
|
f"'old_string' occurs {occurrences} times in {relative}; "
|
|
"make it unique or pass replace_all=true"
|
|
)
|
|
|
|
updated = text.replace(old_string, new_string)
|
|
target.write_text(updated, encoding="utf-8")
|
|
payload = {
|
|
"path": relative,
|
|
"replacements": occurrences if replace_all else 1,
|
|
}
|
|
return _text_result(json.dumps(payload, indent=2), structured=payload)
|
|
|
|
|
|
def _download_bytes(url: str, max_bytes: int) -> bytes:
|
|
"""Fetch an http(s) URL into memory, refusing anything over `max_bytes`."""
|
|
request = urllib.request.Request(
|
|
url, headers={"User-Agent": f"{SERVER_NAME}/{SERVER_VERSION}"}
|
|
)
|
|
with urllib.request.urlopen(request, timeout=_DOWNLOAD_TIMEOUT_SECONDS) as response:
|
|
declared = response.headers.get("Content-Length")
|
|
if declared and declared.isdigit() and int(declared) > max_bytes:
|
|
raise ValueError(
|
|
f"remote file is {declared} bytes, over the {max_bytes} byte limit"
|
|
)
|
|
data = response.read(max_bytes + 1)
|
|
if len(data) > max_bytes:
|
|
raise ValueError(f"download exceeds the {max_bytes} byte limit")
|
|
return data
|
|
|
|
|
|
def _tool_capture_raw(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
source_path = str(args.get("source_path") or "").strip()
|
|
url = str(args.get("url") or "").strip()
|
|
if bool(source_path) == bool(url):
|
|
raise ValueError("provide exactly one of 'source_path' or 'url'")
|
|
|
|
max_bytes = int(args.get("max_bytes", _DEFAULT_MAX_CAPTURE_BYTES))
|
|
if max_bytes < 1:
|
|
raise ValueError("'max_bytes' must be positive")
|
|
|
|
target = _resolve_in_root(state, args.get("path"))
|
|
relative = _relative_to_root(state, target)
|
|
existed = target.exists()
|
|
if existed and not bool(args.get("overwrite", False)):
|
|
raise FileExistsError(
|
|
f"capture already exists at {relative}; pass overwrite=true to replace it"
|
|
)
|
|
if existed and target.is_dir():
|
|
raise ValueError(f"'{relative}' is a directory")
|
|
|
|
if source_path:
|
|
origin = Path(source_path).expanduser()
|
|
if not origin.exists():
|
|
raise FileNotFoundError(f"source file not found at {origin}")
|
|
if not origin.is_file():
|
|
raise ValueError(f"source is not a file: {origin}")
|
|
size = origin.stat().st_size
|
|
if size > max_bytes:
|
|
raise ValueError(
|
|
f"source file is {size} bytes, over the {max_bytes} byte limit"
|
|
)
|
|
target.parent.mkdir(parents=True, exist_ok=True)
|
|
shutil.copyfile(origin, target)
|
|
data = target.read_bytes()
|
|
source_label = str(origin)
|
|
else:
|
|
scheme = url.split(":", 1)[0].lower()
|
|
if scheme not in ("http", "https"):
|
|
raise ValueError(f"only http(s) URLs can be captured, got '{url}'")
|
|
try:
|
|
data = _download_bytes(url, max_bytes)
|
|
except (urllib.error.URLError, urllib.error.HTTPError, OSError) as exc:
|
|
payload = {
|
|
"url": url,
|
|
"path": relative,
|
|
"exit_code": 1,
|
|
"stderr": f"download failed: {exc}",
|
|
}
|
|
return _git_result(payload, 1)
|
|
target.parent.mkdir(parents=True, exist_ok=True)
|
|
target.write_bytes(data)
|
|
source_label = url
|
|
|
|
payload = {
|
|
"path": relative,
|
|
"absolute_path": str(target),
|
|
"source": source_label,
|
|
"created": not existed,
|
|
"bytes": len(data),
|
|
"sha256": hashlib.sha256(data).hexdigest(),
|
|
}
|
|
return _text_result(json.dumps(payload, indent=2), structured=payload)
|
|
|
|
|
|
def _tool_git_commit(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
message = str(args.get("message", "")).strip()
|
|
if not message:
|
|
raise ValueError("'message' is required")
|
|
|
|
raw_paths = args.get("paths")
|
|
if not isinstance(raw_paths, list) or not raw_paths:
|
|
raise ValueError("'paths' must list at least one vault-relative path")
|
|
|
|
relatives: List[str] = []
|
|
for raw in raw_paths:
|
|
value = str(raw or "").strip()
|
|
if value in _WILDCARD_PATHS or value.startswith("-"):
|
|
raise ValueError(
|
|
f"wildcard staging is not allowed: '{value}'. Name each path explicitly."
|
|
)
|
|
target = _resolve_in_root(state, value, label="paths[]")
|
|
relative = _relative_to_root(state, target)
|
|
if not target.exists():
|
|
raise FileNotFoundError(f"path not found at {relative}")
|
|
relatives.append(relative)
|
|
|
|
relatives = sorted(set(relatives))
|
|
for relative in relatives:
|
|
add = _run_git(state, ["add", "--", relative])
|
|
if add.returncode != 0:
|
|
payload = {
|
|
"command": ["git", "add", "--", relative],
|
|
"exit_code": add.returncode,
|
|
"stdout": add.stdout,
|
|
"stderr": add.stderr,
|
|
"committed_paths": [],
|
|
}
|
|
return _git_result(payload, add.returncode)
|
|
|
|
commit = _run_git(state, ["commit", "-m", message])
|
|
payload = {
|
|
"command": ["git", "commit", "-m", message],
|
|
"exit_code": commit.returncode,
|
|
"stdout": commit.stdout,
|
|
"stderr": commit.stderr,
|
|
"committed_paths": relatives if commit.returncode == 0 else [],
|
|
}
|
|
if commit.returncode == 0:
|
|
head = _run_git(state, ["rev-parse", "--short", "HEAD"])
|
|
payload["commit"] = head.stdout.strip()
|
|
return _git_result(payload, commit.returncode)
|
|
|
|
|
|
def _tool_git_push(state: AppState, args: Dict[str, Any]) -> Dict[str, Any]:
|
|
remote = str(args.get("remote") or "origin").strip()
|
|
branch = str(args.get("branch") or "").strip()
|
|
for label, value in (("remote", remote), ("branch", branch)):
|
|
if value.startswith("-"):
|
|
raise ValueError(f"'{label}' must not look like an option: '{value}'")
|
|
if not remote:
|
|
raise ValueError("'remote' must not be empty")
|
|
|
|
if not branch:
|
|
head = _run_git(state, ["rev-parse", "--abbrev-ref", "HEAD"])
|
|
if head.returncode != 0:
|
|
payload = {
|
|
"command": ["git", "rev-parse", "--abbrev-ref", "HEAD"],
|
|
"exit_code": head.returncode,
|
|
"stdout": head.stdout,
|
|
"stderr": head.stderr,
|
|
}
|
|
return _git_result(payload, head.returncode)
|
|
branch = head.stdout.strip()
|
|
|
|
command = ["push"]
|
|
if bool(args.get("set_upstream", False)):
|
|
command.append("--set-upstream")
|
|
command += [remote, branch]
|
|
|
|
push = _run_git(state, command)
|
|
payload = {
|
|
"command": ["git", *command],
|
|
"remote": remote,
|
|
"branch": branch,
|
|
"exit_code": push.returncode,
|
|
"stdout": push.stdout,
|
|
"stderr": push.stderr,
|
|
}
|
|
return _git_result(payload, push.returncode)
|
|
|
|
|
|
def _call_tool(state: AppState, name: str, arguments: Dict[str, Any]) -> Dict[str, Any]:
|
|
if name == "graph_summary":
|
|
return _tool_graph_summary(state, arguments)
|
|
if name == "graph_search_nodes":
|
|
return _tool_graph_search_nodes(state, arguments)
|
|
if name == "graph_neighbors":
|
|
return _tool_graph_neighbors(state, arguments)
|
|
if name == "read_note":
|
|
return _tool_read_note(state, arguments)
|
|
if name == "refresh_graph":
|
|
return _tool_refresh_graph(state, arguments)
|
|
if name == "write_note":
|
|
return _tool_write_note(state, arguments)
|
|
if name == "edit_note":
|
|
return _tool_edit_note(state, arguments)
|
|
if name == "capture_raw":
|
|
return _tool_capture_raw(state, arguments)
|
|
if name == "git_commit":
|
|
return _tool_git_commit(state, arguments)
|
|
if name == "git_push":
|
|
return _tool_git_push(state, arguments)
|
|
raise ValueError(f"unknown tool '{name}'")
|
|
|
|
|
|
def _list_resources(state: AppState) -> Dict[str, Any]:
|
|
return {
|
|
"resources": [
|
|
{
|
|
"uri": "repo://graph",
|
|
"name": "Repository Graph",
|
|
"description": "Raw index/graph.json",
|
|
"mimeType": "application/json",
|
|
},
|
|
{
|
|
"uri": "repo://readme",
|
|
"name": "Repository README",
|
|
"description": "Root README.md",
|
|
"mimeType": "text/markdown",
|
|
},
|
|
]
|
|
}
|
|
|
|
|
|
def _read_resource(state: AppState, uri: str) -> Dict[str, Any]:
|
|
if uri == "repo://graph":
|
|
text = state.graph_path.read_text(encoding="utf-8")
|
|
return {
|
|
"contents": [
|
|
{
|
|
"uri": uri,
|
|
"mimeType": "application/json",
|
|
"text": text,
|
|
}
|
|
]
|
|
}
|
|
if uri == "repo://readme":
|
|
readme = (state.root / "README.md").read_text(encoding="utf-8")
|
|
return {"contents": [{"uri": uri, "mimeType": "text/markdown", "text": readme}]}
|
|
raise ValueError(f"unknown resource '{uri}'")
|
|
|
|
|
|
def _handle_request(state: AppState, message: Dict[str, Any]) -> Dict[str, Any] | None:
|
|
method = message.get("method")
|
|
message_id = message.get("id")
|
|
params = message.get("params") or {}
|
|
|
|
if method == "notifications/initialized":
|
|
return None
|
|
|
|
if method == "initialize":
|
|
result = {
|
|
"protocolVersion": "2024-11-05",
|
|
"serverInfo": {"name": SERVER_NAME, "version": SERVER_VERSION},
|
|
"capabilities": {"tools": {}, "resources": {}},
|
|
}
|
|
return _response(message_id, result)
|
|
|
|
if method == "tools/list":
|
|
return _response(message_id, {"tools": _tool_definitions()})
|
|
|
|
if method == "tools/call":
|
|
try:
|
|
tool_name = str(params.get("name", ""))
|
|
arguments = params.get("arguments") or {}
|
|
if not isinstance(arguments, dict):
|
|
raise ValueError("tool arguments must be an object")
|
|
result = _call_tool(state, tool_name, arguments)
|
|
return _response(message_id, result)
|
|
except Exception as exc:
|
|
error_result = {
|
|
"isError": True,
|
|
"content": [{"type": "text", "text": str(exc)}],
|
|
}
|
|
return _response(message_id, error_result)
|
|
|
|
if method == "resources/list":
|
|
return _response(message_id, _list_resources(state))
|
|
|
|
if method == "resources/read":
|
|
try:
|
|
uri = str(params.get("uri", ""))
|
|
return _response(message_id, _read_resource(state, uri))
|
|
except Exception as exc:
|
|
return _error(message_id, -32602, str(exc))
|
|
|
|
if message_id is None:
|
|
return None
|
|
return _error(message_id, -32601, f"method not found: {method}")
|
|
|
|
|
|
def main() -> None:
|
|
root = Path(__file__).resolve().parents[1]
|
|
state = AppState(root=root, graph_path=root / "index" / "graph.json")
|
|
|
|
reader = _StdinReader(sys.stdin.buffer)
|
|
try:
|
|
transport = _detect_transport(reader)
|
|
except OSError as exc:
|
|
print(
|
|
json.dumps({"error": "transport detection failed", "detail": str(exc)}),
|
|
file=sys.stderr,
|
|
)
|
|
sys.exit(1)
|
|
if transport == "ndjson":
|
|
read_msg = _read_message_ndjson
|
|
write_msg = _write_message_ndjson
|
|
else:
|
|
read_msg = _read_message_framed
|
|
write_msg = _write_message_framed
|
|
|
|
while True:
|
|
try:
|
|
message = read_msg(reader)
|
|
except Exception as exc:
|
|
write_msg(sys.stdout, _error(None, -32700, f"parse error: {exc}"))
|
|
continue
|
|
|
|
if message is None:
|
|
break
|
|
|
|
response = _handle_request(state, message)
|
|
if response is not None:
|
|
write_msg(sys.stdout, response)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|