Files
wangweiandCopilot 49ee50c104 fix: harden MCP endpoint after code review
Critical: the MCP SDK auto-enables DNS-rebinding protection when its host
parameter is left at the 127.0.0.1 default, hard-coding a loopback-only Host
allow-list. Every remote client (the only deployment this feature targets) was
refused with HTTP 421 before auth or the tool ran. Now driven by a new
MCP_ALLOWED_HOSTS setting, with '*' as an explicit, logged opt-out.

Also bounds query/top_k to match AskRequest (top_k is amplified 4x downstream,
so an unbounded value was a resource-exhaustion vector), decodes the
Authorization header as latin-1 per the ASGI spec instead of raising a 500 on
malformed bytes, and returns WWW-Authenticate on 401 per RFC 7235.

Moves the psycopg2 import guard into backend/tests/conftest.py: duplicated
across four test modules, it only worked because of alphabetical collection
order, and any earlier-sorting package would have reintroduced a live
connection attempt against the production database.

Registers the mcp module in the authoritative backend architecture doc.

84 backend tests pass. Verified against a live server: allowed remote Host
returns a valid initialize result, unknown Host returns 421, missing token
returns 401 with WWW-Authenticate.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-07-29 17:11:54 +08:00

28 lines
1.3 KiB
Python

"""Shared pytest fixtures and import-time guards for the backend test suite.
pytest imports this file before any test module beneath backend/tests/, which
makes it the only reliable place to install import-time guards: individual test
modules cannot guarantee they run first, because collection order follows
directory names.
"""
from __future__ import annotations
import sys
from unittest.mock import MagicMock
# app/shared/bootstrap.py (the composition root) eagerly imports the Postgres
# store modules, which do `import psycopg2` at their own module scope and later
# open a real connection pool. Any test that transitively imports bootstrap
# would therefore bind the real driver and attempt a live TCP connection to the
# configured production database, surfacing as a multi-second timeout rather
# than an obvious error. Binding mocks here — before the first test module is
# imported — makes that impossible regardless of collection order.
# setdefault (not assignment) keeps a real psycopg2 in place if something has
# already imported it deliberately.
_mock_psycopg2 = MagicMock()
_mock_psycopg2.extras = MagicMock()
sys.modules.setdefault("psycopg2", _mock_psycopg2)
sys.modules.setdefault("psycopg2.extras", _mock_psycopg2.extras)
sys.modules.setdefault("psycopg2.pool", MagicMock())