Jean-Luc Makiola
9643011e64
app/mcp_server.py: FastMCP (mcp SDK), streamable-http auf /mcp, statischer Bearer-Token (constant-time ASGI-Middleware), Fail-Fast ohne RAG_MCP_TOKEN. Tools rag_search (mit semester/fach/typ-Filter) + get_file_chunks. Läuft aus demselben Image wie der Ingestor und reused den Embed-Pfad → Vektoren sind garantiert kompatibel zum Ingest (der offizielle qdrant-MCP-Server kann nur fastembed → Dimension-/Schema-Mismatch). app/qdrant_store.py: search_chunks (query_points + optionaler Payload-Filter) und get_chunks_by_path (scroll, nach chunk_index sortiert). app/bulk.py: Amplification-Guard — /bulk-import lehnt mit 409 ab solange ein vorheriger Bulk noch BackgroundTasks abarbeitet. docker-compose.coolify.yml: rag-mcp-Service (nicht public, externes metamcp-net statt Stack-Coupling) + Traefik-Rate-Limit-Middleware am ingestor. tests/conftest.py: Settings-env_file in Tests neutralisieren (Dev-.env darf die Suite nicht kontaminieren). 68 passed, ruff clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
83 lines
2.8 KiB
Python
83 lines
2.8 KiB
Python
from unittest.mock import AsyncMock, call
|
|
|
|
from fastapi.testclient import TestClient
|
|
|
|
from app.webhook.models import EventType
|
|
|
|
|
|
def _make_app(monkeypatch):
|
|
monkeypatch.setenv("NEXTCLOUD_WEBDAV_URL", "http://nc")
|
|
monkeypatch.setenv("NEXTCLOUD_USER", "u")
|
|
monkeypatch.setenv("NEXTCLOUD_APP_PASSWORD", "p")
|
|
monkeypatch.setenv("OLLAMA_URL", "http://ollama")
|
|
monkeypatch.setenv("OLLAMA_EMBED_MODEL", "m")
|
|
monkeypatch.setenv("QDRANT_URL", "http://qdrant")
|
|
monkeypatch.setenv("QDRANT_COLLECTION", "rag_test")
|
|
monkeypatch.setenv("WEBHOOK_SECRET", "abc")
|
|
|
|
from app.config import get_settings
|
|
get_settings.cache_clear()
|
|
import app.ingest.pipeline as pipe
|
|
pipe._qdrant_client.cache_clear()
|
|
|
|
monkeypatch.setattr("app.main._startup_ensure_collection", AsyncMock())
|
|
from app.main import app
|
|
return app
|
|
|
|
|
|
def test_bulk_import_lists_and_dispatches(monkeypatch):
|
|
app = _make_app(monkeypatch)
|
|
|
|
listed = [
|
|
"Documents/THB/Studium/2.Semester/Databases/a.pdf",
|
|
"Documents/THB/Studium/2.Semester/Databases/b.docx",
|
|
"Documents/THB/Studium/2.Semester/Databases/.rag-meta.json", # ignored
|
|
]
|
|
monkeypatch.setattr("app.bulk.list_files_recursive", AsyncMock(return_value=listed))
|
|
|
|
process_mock = AsyncMock()
|
|
monkeypatch.setattr("app.bulk.process_file", process_mock)
|
|
|
|
with TestClient(app) as client:
|
|
r = client.post(
|
|
"/bulk-import",
|
|
json={"path": "Documents/THB/Studium/2.Semester/Databases"},
|
|
headers={"X-Webhook-Secret": "abc"},
|
|
)
|
|
|
|
assert r.status_code == 202
|
|
body = r.json()
|
|
assert body["dispatched"] == 2 # only .pdf and .docx, not the json sidecar
|
|
process_mock.assert_has_calls(
|
|
[
|
|
call("Documents/THB/Studium/2.Semester/Databases/a.pdf", EventType.CREATED),
|
|
call("Documents/THB/Studium/2.Semester/Databases/b.docx", EventType.CREATED),
|
|
]
|
|
)
|
|
assert process_mock.await_count == 2
|
|
|
|
|
|
def test_bulk_import_rejects_wrong_secret(monkeypatch):
|
|
app = _make_app(monkeypatch)
|
|
with TestClient(app) as client:
|
|
r = client.post("/bulk-import", json={"path": "x"}, headers={"X-Webhook-Secret": "nope"})
|
|
assert r.status_code == 401
|
|
|
|
|
|
def test_bulk_import_rejects_while_in_progress(monkeypatch):
|
|
app = _make_app(monkeypatch)
|
|
# Simulate an in-flight bulk: the amplification guard must reject.
|
|
monkeypatch.setattr("app.bulk._inflight", 4)
|
|
|
|
list_mock = AsyncMock()
|
|
monkeypatch.setattr("app.bulk.list_files_recursive", list_mock)
|
|
|
|
with TestClient(app) as client:
|
|
r = client.post(
|
|
"/bulk-import", json={"path": "x"}, headers={"X-Webhook-Secret": "abc"}
|
|
)
|
|
|
|
assert r.status_code == 409
|
|
# Guard fires before any WebDAV listing happens.
|
|
list_mock.assert_not_awaited()
|