@oriro/orirocli 0.1.7 → 0.1.9

package/skills/graphify/manifest.py

@@ -0,0 +1,4 @@
+# re-export manifest helpers from detect for backwards compatibility
+from graphify.detect import save_manifest, load_manifest, detect_incremental
+
+__all__ = ["save_manifest", "load_manifest", "detect_incremental"]

package/skills/graphify/mcp_ingest.py

@@ -0,0 +1,392 @@
+"""mcp_ingest.py — Extract MCP (Model Context Protocol) server configuration files.
+
+Reads `.mcp.json` / `claude_desktop_config.json` / `mcp.json` / `mcp_servers.json`
+and turns the `mcpServers` map into Graphify nodes and edges.
+
+Symmetry with `serve.py`: Graphify exposes itself AS an MCP server. This module
+indexes MCP servers AS a corpus type, completing the loop — an agent that runs
+graphify with `--mcp` can now query its own configured MCP layer.
+
+Entry point:
+  extract_mcp_config(path: Path) -> dict[str, list[dict]]
+
+  Returns `{"nodes": [...], "edges": [...]}` compatible with Graphify's
+  extraction-result format. Returns `{"nodes": [...], "edges": [...], "error": "..."}`
+  when the file is malformed, too large, or has no `mcpServers` map — the empty
+  result keeps it indistinguishable from "no MCP config here" for downstream
+  callers.
+
+Detected filenames (case-sensitive, matched on basename):
+  - .mcp.json                       (Claude Code project config)
+  - claude_desktop_config.json      (Claude Desktop)
+  - mcp.json                        (generic / per-tool)
+  - mcp_servers.json                (alternate naming)
+
+Schema emitted:
+  Node kinds:
+    - file              the config file itself (label = filename)
+    - mcp_server        one per entry under mcpServers
+    - mcp_command       executable (npx, uvx, node, python, ...) — global ID
+    - mcp_package       npm / pypi package id parsed from args — global ID
+    - env_var           env variable NAME only — global ID. VALUES ARE NEVER READ.
+
+  Edge relations:
+    - contains          file -> mcp_server
+    - references        mcp_server -> mcp_command
+    - references        mcp_server -> mcp_package
+    - requires_env      mcp_server -> env_var   (new relation; distinguishes
+                                                  env dependencies from generic refs)
+
+Security:
+  - Env var VALUES are never read, persisted, labelled, or surfaced. Only env
+    var NAMES become nodes. (`env: {"API_KEY": "sk-..."}` -> node "API_KEY" only.)
+  - File size capped at 1 MiB (matches extract_json).
+  - All labels go through `sanitize_label` (control characters stripped, length
+    capped) before emission.
+  - Args are NOT persisted as nodes/edges to avoid leaking paths or secrets that
+    some servers embed as positional args.
+
+Cross-config emergent edges:
+  Because `mcp_command`, `mcp_package`, and `env_var` nodes use global IDs (no
+  per-file stem prefix), the same package or env var across two MCP configs
+  produces shared nodes — naturally surfacing "what configs depend on this
+  thing?" via graph traversal. Server nodes ARE stem-scoped so two configs
+  declaring different servers under the same key (e.g., both have "filesystem")
+  do not collide.
+"""
+
+from __future__ import annotations
+
+import json
+import re
+import unicodedata
+from pathlib import Path
+from typing import Any
+
+from graphify.security import sanitize_label
+
+
+MCP_CONFIG_FILENAMES: frozenset[str] = frozenset({
+    ".mcp.json",
+    "claude_desktop_config.json",
+    "mcp.json",
+    "mcp_servers.json",
+})
+
+_MAX_BYTES = 1_048_576  # 1 MiB — same cap as extract_json
+_MAX_SERVERS_PER_FILE = 200  # generous; flags pathological configs
+
+
+def is_mcp_config_path(path: Path) -> bool:
+    """Return True when ``path`` is a recognised MCP config filename."""
+    return path.name in MCP_CONFIG_FILENAMES
+
+
+def extract_mcp_config(path: Path) -> dict[str, Any]:
+    """Parse an MCP config file into Graphify nodes and edges.
+
+    Behaviour matches other extractors in `extract.py`:
+      - returns ``{"nodes": [...], "edges": [...]}`` on success
+      - returns ``{"nodes": [], "edges": [], "error": "<reason>"}`` on parse
+        failure, oversize file, or missing ``mcpServers`` map
+    """
+    try:
+        with path.open("rb") as fh:
+            raw = fh.read(_MAX_BYTES + 1)
+    except OSError as exc:
+        return {"nodes": [], "edges": [], "error": f"mcp_ingest read error: {exc}"}
+
+    if len(raw) > _MAX_BYTES:
+        return {"nodes": [], "edges": [], "error": "mcp config too large to index"}
+
+    try:
+        text = raw.decode("utf-8")
+    except UnicodeDecodeError as exc:
+        return {"nodes": [], "edges": [], "error": f"mcp_ingest decode error: {exc}"}
+
+    try:
+        doc = json.loads(text)
+    except json.JSONDecodeError as exc:
+        return {"nodes": [], "edges": [], "error": f"mcp_ingest json error: {exc}"}
+
+    if not isinstance(doc, dict):
+        return {"nodes": [], "edges": [], "error": "mcp_ingest: root is not an object"}
+
+    servers = doc.get("mcpServers")
+    if not isinstance(servers, dict):
+        # Some tools nest the map (e.g., {"mcp": {"servers": {...}}}). Try one
+        # well-known alternate shape but do not search exhaustively.
+        nested = doc.get("mcp")
+        if isinstance(nested, dict):
+            servers = nested.get("servers")
+        if not isinstance(servers, dict):
+            return {"nodes": [], "edges": [], "error": "mcp_ingest: no mcpServers map"}
+
+    str_path = str(path)
+    file_nid = _make_id(str_path)
+    nodes: list[dict[str, Any]] = []
+    edges: list[dict[str, Any]] = []
+    seen_node_ids: set[str] = set()
+    seen_edge_keys: set[tuple[str, str, str]] = set()
+
+    _add_node(
+        nodes, seen_node_ids,
+        nid=file_nid,
+        label=path.name,
+        kind="mcp_config_file",
+        source_file=str_path,
+        line=1,
+    )
+
+    file_stem = _file_stem(path)
+    server_count = 0
+    for server_name, spec in servers.items():
+        if not isinstance(server_name, str) or not server_name:
+            continue
+        if not isinstance(spec, dict):
+            # Skip non-object server entries silently — the broken entry is
+            # the user's, not ours.
+            continue
+        if server_count >= _MAX_SERVERS_PER_FILE:
+            break
+        server_count += 1
+        _emit_server(
+            server_name=server_name,
+            spec=spec,
+            file_nid=file_nid,
+            file_stem=file_stem,
+            source_file=str_path,
+            nodes=nodes,
+            edges=edges,
+            seen_node_ids=seen_node_ids,
+            seen_edge_keys=seen_edge_keys,
+        )
+
+    return {"nodes": nodes, "edges": edges}
+
+
+def _emit_server(
+    *,
+    server_name: str,
+    spec: dict[str, Any],
+    file_nid: str,
+    file_stem: str,
+    source_file: str,
+    nodes: list[dict[str, Any]],
+    edges: list[dict[str, Any]],
+    seen_node_ids: set[str],
+    seen_edge_keys: set[tuple[str, str, str]],
+) -> None:
+    """Emit nodes/edges for one entry under ``mcpServers``."""
+    server_nid = _make_id(file_stem, "mcp_server", server_name)
+    _add_node(
+        nodes, seen_node_ids,
+        nid=server_nid,
+        label=server_name,
+        kind="mcp_server",
+        source_file=source_file,
+        line=1,  # JSON doesn't expose line numbers without a parser pass
+    )
+    _add_edge(
+        edges, seen_edge_keys,
+        source=file_nid,
+        target=server_nid,
+        relation="contains",
+        source_file=source_file,
+        line=1,
+    )
+
+    command = spec.get("command")
+    if isinstance(command, str) and command.strip():
+        cmd_label = command.strip()
+        cmd_nid = _make_id("mcp_command", cmd_label)
+        _add_node(
+            nodes, seen_node_ids,
+            nid=cmd_nid,
+            label=cmd_label,
+            kind="mcp_command",
+            source_file=source_file,
+            line=1,
+        )
+        _add_edge(
+            edges, seen_edge_keys,
+            source=server_nid,
+            target=cmd_nid,
+            relation="references",
+            source_file=source_file,
+            line=1,
+            context="command",
+        )
+
+    args = spec.get("args")
+    if isinstance(args, list):
+        package = _detect_package_from_args(args)
+        if package:
+            pkg_nid = _make_id("mcp_package", package)
+            _add_node(
+                nodes, seen_node_ids,
+                nid=pkg_nid,
+                label=package,
+                kind="mcp_package",
+                source_file=source_file,
+                line=1,
+            )
+            _add_edge(
+                edges, seen_edge_keys,
+                source=server_nid,
+                target=pkg_nid,
+                relation="references",
+                source_file=source_file,
+                line=1,
+                context="package",
+            )
+
+    env = spec.get("env")
+    if isinstance(env, dict):
+        # ONLY KEYS. Values may contain secrets and are never read here.
+        for env_name in env.keys():
+            if not isinstance(env_name, str) or not env_name:
+                continue
+            env_nid = _make_id("env_var", env_name)
+            _add_node(
+                nodes, seen_node_ids,
+                nid=env_nid,
+                label=env_name,
+                kind="env_var",
+                source_file=source_file,
+                line=1,
+            )
+            _add_edge(
+                edges, seen_edge_keys,
+                source=server_nid,
+                target=env_nid,
+                relation="requires_env",
+                source_file=source_file,
+                line=1,
+            )
+
+
+# ── Package detection from args ───────────────────────────────────────────────
+
+# Patterns observed in real MCP server configs:
+#   ["-y", "@modelcontextprotocol/server-filesystem", "/data"]   (npx)
+#   ["-y", "@org/pkg@1.2.3"]
+#   ["mcp-server-fetch"]                                          (uvx / python)
+#   ["mcp-server-time", "--local-timezone=UTC"]
+#   ["@scoped/some-mcp"]                                          (pnpx)
+#   ["mcp-server-fetch"]                                          (uvx direct)
+_NPM_PKG_RE = re.compile(r"^@[a-z0-9][a-z0-9._-]*/[a-z0-9][a-z0-9._-]*(?:@[\w.\-+]+)?$")
+_PY_MCP_PKG_RE = re.compile(r"^[a-z0-9][a-z0-9._-]*-mcp(?:-[a-z0-9._-]+)?$|^mcp-[a-z0-9][a-z0-9._-]*$")
+_ARG_FLAG_RE = re.compile(r"^-{1,2}\w")
+
+
+def _detect_package_from_args(args: list[Any]) -> str | None:
+    """Return the first arg that looks like an npm or pypi package id, else None.
+
+    Skips short flags (-y, --yes) and option arguments (--local-timezone=UTC).
+    """
+    for raw in args:
+        if not isinstance(raw, str):
+            continue
+        arg = raw.strip()
+        if not arg or _ARG_FLAG_RE.match(arg):
+            continue
+        if _NPM_PKG_RE.match(arg):
+            return _strip_version(arg)
+        if _PY_MCP_PKG_RE.match(arg):
+            return arg
+    return None
+
+
+def _strip_version(pkg: str) -> str:
+    """Drop the ``@version`` suffix from an npm package id, preserving the scope.
+
+    Scoped:   ``@scope/name`` or ``@scope/name@1.2.3`` — there are at most two
+              ``@`` chars; the second is the version separator.
+    Unscoped: ``name`` or ``name@1.2.3``.
+    """
+    if pkg.startswith("@"):
+        version_at = pkg.find("@", 1)
+        return pkg if version_at == -1 else pkg[:version_at]
+    version_at = pkg.find("@")
+    return pkg if version_at == -1 else pkg[:version_at]
+
+
+# ── Node / edge construction (Graphify schema) ────────────────────────────────
+
+
+def _add_node(
+    nodes: list[dict[str, Any]],
+    seen: set[str],
+    *,
+    nid: str,
+    label: str,
+    kind: str,
+    source_file: str,
+    line: int,
+) -> None:
+    """Append a node if not already present. ``kind`` is metadata, not file_type."""
+    if not nid or nid in seen:
+        return
+    seen.add(nid)
+    nodes.append({
+        "id": nid,
+        "label": sanitize_label(label),
+        "file_type": "code",
+        "source_file": source_file,
+        "source_location": f"L{line}",
+        "metadata": {"mcp_kind": kind},
+    })
+
+
+def _add_edge(
+    edges: list[dict[str, Any]],
+    seen: set[tuple[str, str, str]],
+    *,
+    source: str,
+    target: str,
+    relation: str,
+    source_file: str,
+    line: int,
+    context: str | None = None,
+) -> None:
+    """Append an edge if (source, target, relation) is not already present."""
+    if not source or not target or source == target:
+        return
+    key = (source, target, relation)
+    if key in seen:
+        return
+    seen.add(key)
+    edge: dict[str, Any] = {
+        "source": source,
+        "target": target,
+        "relation": relation,
+        "confidence": "EXTRACTED",
+        "confidence_score": 1.0,
+        "source_file": source_file,
+        "source_location": f"L{line}",
+        "weight": 1.0,
+    }
+    if context:
+        edge["context"] = context
+    edges.append(edge)
+
+
+# ── ID helpers (kept local; mirror extract.py shape) ──────────────────────────
+
+
+def _make_id(*parts: str) -> str:
+    """Build a stable node ID. Must match extract._make_id's normalisation rules."""
+    combined = "_".join(p.strip("_.") for p in parts if p)
+    combined = unicodedata.normalize("NFKC", combined)
+    cleaned = re.sub(r"[^\w]+", "_", combined, flags=re.UNICODE)
+    cleaned = re.sub(r"_+", "_", cleaned)
+    return cleaned.strip("_").casefold()
+
+
+def _file_stem(path: Path) -> str:
+    """Mirror extract._file_stem: include parent dir name to disambiguate."""
+    parent = path.parent.name
+    if parent and parent not in (".", ""):
+        return f"{parent}.{path.stem}"
+    return path.stem

package/skills/graphify/multigraph_compat.py

@@ -0,0 +1,212 @@
+"""Runtime compatibility probe for Graphify MultiDiGraph mode.
+
+Verifies that the current NetworkX runtime supports the behaviors a future
+opt-in --multigraph build will rely on. The probe is BEHAVIOR-based, not
+version-based — both NX 3.4.2 (Py 3.10 lane) and NX 3.6.1+ (Py 3.11+ lane)
+pass. The probe result is cached for the process lifetime via lru_cache.
+
+No call sites added yet; downstream multigraph PRs will gate on
+require_multigraph_capabilities() before enabling MDG mode.
+"""
+
+from __future__ import annotations
+
+from collections.abc import Callable
+from dataclasses import dataclass
+from functools import lru_cache
+import sys
+from typing import Any
+
+import networkx as nx
+from networkx.readwrite import json_graph
+
+
+@dataclass(frozen=True)
+class CapabilityCheck:
+    name: str
+    ok: bool
+    detail: str
+
+
+@dataclass(frozen=True)
+class MultigraphCapabilityResult:
+    python_version: str
+    networkx_version: str
+    checks: tuple[CapabilityCheck, ...]
+
+    @property
+    def ok(self) -> bool:
+        return all(check.ok for check in self.checks)
+
+    @property
+    def failed(self) -> tuple[CapabilityCheck, ...]:
+        return tuple(check for check in self.checks if not check.ok)
+
+    def error_message(self) -> str:
+        if self.ok:
+            return (
+                "Graphify MultiDiGraph capability probe passed "
+                f"(Python {self.python_version}, NetworkX {self.networkx_version})."
+            )
+        failed = "; ".join(f"{check.name}: {check.detail}" for check in self.failed)
+        return (
+            "error: --multigraph requires NetworkX keyed MultiDiGraph node-link "
+            "round-trip support. "
+            f"Detected Python {self.python_version}, NetworkX {self.networkx_version}. "
+            f"Failed capability check(s): {failed}. "
+            "Default simple graph mode remains available."
+        )
+
+
+def _check(name: str, func: Callable[[], bool | str]) -> CapabilityCheck:
+    try:
+        detail = func()
+    except Exception as exc:
+        return CapabilityCheck(name, False, f"{type(exc).__name__}: {exc}")
+    if detail is True:
+        return CapabilityCheck(name, True, "ok")
+    if isinstance(detail, str):
+        return CapabilityCheck(name, False, detail)
+    return CapabilityCheck(name, False, f"unexpected result {detail!r}")
+
+
+def _build_probe_graph() -> nx.MultiDiGraph:
+    graph = nx.MultiDiGraph()
+    graph.add_node("a", label="A")
+    graph.add_node("b", label="B")
+    graph.add_edge("a", "b", key="calls:a.py:L1", relation="calls", source_file="a.py")
+    graph.add_edge("a", "b", key="imports:a.py:L2", relation="imports", source_file="a.py")
+    return graph
+
+
+def _probe_keyed_parallel_edges() -> bool | str:
+    graph = _build_probe_graph()
+    if not graph.is_multigraph() or not graph.is_directed():
+        return f"probe graph type was {type(graph).__name__}"
+    if graph.number_of_edges("a", "b") != 2:
+        return f"expected 2 keyed parallel edges, got {graph.number_of_edges('a', 'b')}"
+    keys = set(graph["a"]["b"].keys())
+    expected = {"calls:a.py:L1", "imports:a.py:L2"}
+    if keys != expected:
+        return f"expected keys {sorted(expected)}, got {sorted(keys)}"
+    return True
+
+
+def _probe_node_link_round_trip() -> bool | str:
+    graph = _build_probe_graph()
+    data = json_graph.node_link_data(graph, edges="links")
+    if data.get("multigraph") is not True:
+        return f"serialized multigraph flag was {data.get('multigraph')!r}"
+    if data.get("directed") is not True:
+        return f"serialized directed flag was {data.get('directed')!r}"
+    links = data.get("links")
+    if not isinstance(links, list) or len(links) != 2:
+        length = 0 if not isinstance(links, list) else len(links)
+        return f"serialized links length was {length}"
+    serialized_keys: set[str] = set()
+    for edge in links:
+        if isinstance(edge, dict):
+            edge_key = edge.get("key")
+            if isinstance(edge_key, str):
+                serialized_keys.add(edge_key)
+    expected = {"calls:a.py:L1", "imports:a.py:L2"}
+    if serialized_keys != expected:
+        return f"serialized keys {sorted(serialized_keys)} did not match {sorted(expected)}"
+    loaded = json_graph.node_link_graph(data, edges="links")
+    if not isinstance(loaded, nx.MultiDiGraph):
+        return f"round-trip graph type was {type(loaded).__name__}"
+    if loaded.number_of_edges("a", "b") != 2:
+        return f"round-trip edge count was {loaded.number_of_edges('a', 'b')}"
+    loaded_keys = set(loaded["a"]["b"].keys())
+    if loaded_keys != expected:
+        return f"round-trip keys {sorted(loaded_keys)} did not match {sorted(expected)}"
+    return True
+
+
+def _probe_duplicate_key_overwrite_semantics() -> bool | str:
+    graph = nx.MultiDiGraph()
+    graph.add_edge("x", "y", key="same", marker="first")
+    graph.add_edge("x", "y", key="same", marker="second")
+    edges = list(graph.edges(keys=True, data=True))
+    if len(edges) != 1:
+        return f"expected one edge after duplicate-key add, got {len(edges)}"
+    if edges[0][3].get("marker") != "second":
+        return f"expected second attr overwrite, got {edges[0][3].get('marker')!r}"
+    return True
+
+
+def _probe_reserved_key_attr_rejected() -> bool | str:
+    """Verify the Python language guarantee that NetworkX add_edge inherits.
+
+    Python forbids passing the same keyword argument twice — once explicitly
+    and once via **kwargs. This probe confirms that protection still applies
+    to nx.MultiDiGraph.add_edge: a future loader that builds attrs from JSON
+    will be reliably protected from accidentally setting `key` via attrs while
+    also passing `key=` explicitly.
+
+    The probe always passes on any Python 3.x version. Its purpose is to
+    document the invariant explicitly in the probe suite so that if a future
+    Python version relaxes this rule (extremely unlikely), the probe surfaces
+    the regression.
+    """
+    graph = nx.MultiDiGraph()
+    attrs: dict[str, Any] = {"key": "attr-key", "relation": "calls"}
+    try:
+        graph.add_edge("a", "b", key="schema-key", **attrs)
+    except TypeError:
+        return True
+    return "add_edge accepted duplicate key keyword and attr; loader must not rely on this"
+
+
+def _probe_remove_edges_from_two_tuple_semantics() -> bool | str:
+    graph = nx.MultiDiGraph()
+    graph.add_edge("a", "b", key="one")
+    graph.add_edge("a", "b", key="two")
+    graph.remove_edges_from([("a", "b")])
+    remaining = graph.number_of_edges("a", "b")
+    if remaining != 1:
+        return f"expected one remaining edge after two-tuple removal, got {remaining}"
+    return True
+
+
+def _probe_to_undirected_preserves_multigraph_type() -> bool | str:
+    graph = _build_probe_graph()
+    undirected = graph.to_undirected()
+    undirected_view = graph.to_undirected(as_view=True)
+    if not isinstance(undirected, nx.MultiGraph):
+        return f"to_undirected() returned {type(undirected).__name__}"
+    if not isinstance(undirected_view, nx.MultiGraph):
+        return f"to_undirected(as_view=True) returned {type(undirected_view).__name__}"
+    return True
+
+
+@lru_cache(maxsize=1)
+def probe_multigraph_capabilities() -> MultigraphCapabilityResult:
+    checks = (
+        _check("keyed_parallel_edges", _probe_keyed_parallel_edges),
+        _check("node_link_edges_links_round_trip", _probe_node_link_round_trip),
+        _check("duplicate_key_overwrite_semantics", _probe_duplicate_key_overwrite_semantics),
+        _check("reserved_key_attr_rejected", _probe_reserved_key_attr_rejected),
+        _check(
+            "remove_edges_from_two_tuple_semantics",
+            _probe_remove_edges_from_two_tuple_semantics,
+        ),
+        _check(
+            "to_undirected_preserves_multigraph_type",
+            _probe_to_undirected_preserves_multigraph_type,
+        ),
+    )
+    return MultigraphCapabilityResult(
+        python_version=(
+            f"{sys.version_info.major}.{sys.version_info.minor}.{sys.version_info.micro}"
+        ),
+        networkx_version=nx.__version__,
+        checks=checks,
+    )
+
+
+def require_multigraph_capabilities() -> MultigraphCapabilityResult:
+    result = probe_multigraph_capabilities()
+    if not result.ok:
+        raise RuntimeError(result.error_message())
+    return result