mirror of
https://github.com/tiennm99/DocsGPT.git
synced 2026-10-04 20:13:04 +00:00
Expose a user's artifacts through the MCP server as readable resources: each is listed under an artifact:// URI with its mime type and read on demand as inline text or a base64 blob, bounded by a size cap and scoped strictly to the owning principal resolved from the request's API key, so no artifact is served across tenants. Ship the network-level egress controls the sandbox runner needs but cannot self-apply: a Kubernetes NetworkPolicy that allows public egress while denying RFC1918, link-local, and cloud-metadata ranges, an optional docker-compose egress overlay, and runner network-hardening docs.
131 lines
4.7 KiB
Python
131 lines
4.7 KiB
Python
"""FastMCP server exposing DocsGPT retrieval + artifacts over streamable HTTP.
|
|
|
|
Mounted at ``/mcp`` by ``application/asgi.py``. Bearer tokens are the
|
|
existing DocsGPT agent API keys — no new credential surface. The
|
|
``search_docs`` tool searches the caller's knowledge base; the artifact
|
|
resources middleware exposes the caller's own artifacts as MCP resources
|
|
(``resources/list`` / ``resources/read`` over ``artifact://`` URIs),
|
|
scoped strictly to the Bearer key's owning principal.
|
|
|
|
The tool reads the ``Authorization`` header directly via
|
|
``get_http_headers(include={"authorization"})``. The ``include`` kwarg
|
|
is required: by default ``get_http_headers`` strips ``authorization``
|
|
(and a handful of other hop-by-hop headers) so they aren't forwarded
|
|
to downstream services — since we deliberately want the caller's
|
|
token, we opt it back in.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import asyncio
|
|
import base64
|
|
import logging
|
|
from typing import Sequence
|
|
|
|
from fastmcp import FastMCP
|
|
from fastmcp.resources import ResourceContent, ResourceResult
|
|
from fastmcp.resources.base import Resource
|
|
from fastmcp.server.dependencies import get_http_headers
|
|
from fastmcp.server.middleware import CallNext, Middleware, MiddlewareContext
|
|
|
|
from application.services.artifact_resource_service import (
|
|
ArtifactReadResult,
|
|
ResourceDenied,
|
|
ResourceNotFound,
|
|
list_artifact_resources,
|
|
read_artifact_resource,
|
|
)
|
|
from application.services.search_service import (
|
|
InvalidAPIKey,
|
|
SearchFailed,
|
|
search,
|
|
)
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
def _extract_bearer_token() -> str | None:
|
|
auth = get_http_headers(include={"authorization"}).get("authorization", "")
|
|
parts = auth.split(None, 1)
|
|
if len(parts) != 2 or parts[0].lower() != "bearer" or not parts[1]:
|
|
return None
|
|
return parts[1]
|
|
|
|
|
|
def _read_result_to_mcp(result: ArtifactReadResult) -> ResourceResult:
|
|
"""Wrap a service read result as a FastMCP ``ResourceResult`` (text or blob)."""
|
|
if result.blob_b64 is not None:
|
|
raw = base64.b64decode(result.blob_b64)
|
|
return ResourceResult([ResourceContent(raw, mime_type=result.mime_type)])
|
|
return ResourceResult([ResourceContent(result.text or "", mime_type=result.mime_type)])
|
|
|
|
|
|
class ArtifactResourcesMiddleware(Middleware):
|
|
"""Expose the calling principal's artifacts as MCP resources (read/list).
|
|
|
|
The principal is the Bearer api_key's owner; resources are scoped to that
|
|
owner and a foreign/unauthenticated read is denied. Static resources
|
|
registered on the server (if any) are preserved by chaining ``call_next``.
|
|
"""
|
|
|
|
async def on_list_resources(
|
|
self,
|
|
context: MiddlewareContext,
|
|
call_next: CallNext,
|
|
) -> Sequence[Resource]:
|
|
"""Append the principal's artifact resources to the static resource list."""
|
|
existing = list(await call_next(context))
|
|
try:
|
|
artifacts = await asyncio.to_thread(
|
|
list_artifact_resources, _extract_bearer_token()
|
|
)
|
|
except Exception:
|
|
logger.exception("on_list_resources: artifact listing failed")
|
|
return existing
|
|
existing.extend(artifacts)
|
|
return existing
|
|
|
|
async def on_read_resource(
|
|
self,
|
|
context: MiddlewareContext,
|
|
call_next: CallNext,
|
|
) -> ResourceResult:
|
|
"""Serve ``artifact://`` reads from the artifact store; defer others."""
|
|
uri = str(getattr(context.message, "uri", ""))
|
|
if not uri.startswith("artifact://"):
|
|
return await call_next(context)
|
|
try:
|
|
result = await asyncio.to_thread(
|
|
read_artifact_resource, _extract_bearer_token(), uri
|
|
)
|
|
except ResourceDenied as exc:
|
|
raise PermissionError(str(exc) or "forbidden") from exc
|
|
except ResourceNotFound as exc:
|
|
raise ValueError(str(exc) or "resource not found") from exc
|
|
return _read_result_to_mcp(result)
|
|
|
|
|
|
mcp = FastMCP("docsgpt")
|
|
mcp.add_middleware(ArtifactResourcesMiddleware())
|
|
|
|
|
|
@mcp.tool
|
|
async def search_docs(query: str, chunks: int = 5) -> list[dict]:
|
|
"""Search the caller's DocsGPT knowledge base.
|
|
|
|
Authentication is via ``Authorization: Bearer <agent-api-key>`` on
|
|
the MCP request — the same opaque key that ``/api/search`` accepts
|
|
in its JSON body. Returns at most ``chunks`` hits, each a dict with
|
|
``text``, ``title``, ``source`` keys.
|
|
"""
|
|
api_key = _extract_bearer_token()
|
|
if not api_key:
|
|
raise PermissionError("Missing Bearer token")
|
|
try:
|
|
return await asyncio.to_thread(search, api_key, query, chunks)
|
|
except InvalidAPIKey as exc:
|
|
raise PermissionError("Invalid API key") from exc
|
|
except SearchFailed:
|
|
logger.exception("search_docs failed")
|
|
raise
|