Compare commits

..

1 Commits

Author SHA1 Message Date
smithery-ai[bot] 3b48c904f5 Update README 2025-12-11 12:59:56 +00:00
24 changed files with 66 additions and 1156 deletions
-25
View File
@@ -1,28 +1,3 @@
## v0.52.0 (2025-12-13)
### Feat
- **vector**: add Deck card vector search with visualization support
## v0.51.0 (2025-12-13)
### Feat
- **vector-viz**: add news_item support for links and chunk expansion
## v0.50.2 (2025-12-13)
### Fix
- **news**: revert get_item() to use get_items() + filter
## v0.50.1 (2025-12-12)
### Fix
- Disable DNS rebinding protection for containerized deployments
- **deps**: update dependency mcp to >=1.23,<1.24
## v0.50.0 (2025-12-11)
### Feat
+1 -1
View File
@@ -1,4 +1,4 @@
FROM docker.io/library/python:3.12-slim-trixie@sha256:fa48eefe2146644c2308b909d6bb7651a768178f84fc9550dcd495e4d6d84d01
FROM docker.io/library/python:3.12-slim-trixie@sha256:590cad70271b6c1795c6a11fb5c110efca593adbd0d4883cd19c36df6a56467b
COPY --from=ghcr.io/astral-sh/uv:0.9.17@sha256:5cb6b54d2bc3fe2eb9a8483db958a0b9eebf9edff68adedb369df8e7b98711a2 /uv /uvx /bin/
+1 -1
View File
@@ -12,7 +12,7 @@
# - Per-session app password authentication
# - Multi-user support via Smithery session config
FROM docker.io/library/python:3.12-slim-trixie@sha256:fa48eefe2146644c2308b909d6bb7651a768178f84fc9550dcd495e4d6d84d01
FROM docker.io/library/python:3.12-slim-trixie@sha256:590cad70271b6c1795c6a11fb5c110efca593adbd0d4883cd19c36df6a56467b
WORKDIR /app
+6 -4
View File
@@ -1,11 +1,12 @@
```markdown
<p align="center">
<img src="astrolabe.svg" alt="Nextcloud MCP Server" width="128" height="128">
</p>
# Nextcloud MCP Server
[![Docker Image](https://img.shields.io/badge/docker-ghcr.io/cbcoutinho/nextcloud--mcp--server-blue)](https://github.com/cbcoutinho/nextcloud-mcp-server/pkgs/container/nextcloud-mcp-server)
[![smithery badge](https://smithery.ai/badge/@cbcoutinho/nextcloud-mcp-server)](https://smithery.ai/server/@cbcoutinho/nextcloud-mcp-server)
[![Docker Image](https://img.shields.io/badge/docker-ghcr.io/cbcoutinho/nextcloud--mcp--server-blue)](https://github.com/cbcoutinho/nextcloud-mcp-server/pkgs/container/nextcloud-mcp-server)
**A production-ready MCP server that connects AI assistants to your Nextcloud instance.**
@@ -63,7 +64,7 @@ http://127.0.0.1:8000/mcp
- **90+ MCP Tools** - Comprehensive API coverage across 8 Nextcloud apps
- **MCP Resources** - Structured data URIs for browsing Nextcloud data
- **Semantic Search (Experimental)** - Optional vector-powered search for Notes, Files, News items, and Deck cards (requires Qdrant + Ollama)
- **Semantic Search (Experimental)** - Optional vector-powered search for Notes (requires Qdrant + Ollama)
- **Document Processing** - OCR and text extraction from PDFs, DOCX, images with progress notifications
- **Flexible Deployment** - Docker, Kubernetes (Helm), VM, or local installation
- **Production-Ready Auth** - Basic Auth with app passwords (recommended) or OAuth2/OIDC (experimental)
@@ -81,7 +82,7 @@ http://127.0.0.1:8000/mcp
| **Cookbook** | 13 | Recipe management, URL import (schema.org) |
| **Tables** | 5 | Row operations on Nextcloud Tables |
| **Sharing** | 10+ | Create and manage shares |
| **Semantic Search** | 2+ | Vector search for Notes, Files, News items, and Deck cards (experimental, opt-in, requires infrastructure) |
| **Semantic Search** | 2+ | Vector search for Notes (experimental, opt-in, requires infrastructure) |
Want to see another Nextcloud app supported? [Open an issue](https://github.com/cbcoutinho/nextcloud-mcp-server/issues) or contribute a pull request!
@@ -145,7 +146,7 @@ This enables natural language queries and helps discover related content across
### Features
- **[App Documentation](docs/)** - Notes, Calendar, Contacts, WebDAV, Deck, Cookbook, Tables
- **[Document Processing](docs/configuration.md#document-processing)** - OCR and text extraction setup
- **[Semantic Search Architecture](docs/semantic-search-architecture.md)** - Experimental vector search (Notes, Files, News items, Deck cards; opt-in)
- **[Semantic Search Architecture](docs/semantic-search-architecture.md)** - Experimental vector search (Notes only, opt-in)
- **[Vector Sync UI Guide](docs/user-guide/vector-sync-ui.md)** - Browser interface for semantic search visualization and testing
### Advanced Topics
@@ -223,3 +224,4 @@ This project is licensed under the AGPL-3.0 License. See [LICENSE](./LICENSE) fo
- [Model Context Protocol](https://github.com/modelcontextprotocol)
- [MCP Python SDK](https://github.com/modelcontextprotocol/python-sdk)
- [Nextcloud](https://nextcloud.com/)
```
+2 -2
View File
@@ -2,8 +2,8 @@ apiVersion: v2
name: nextcloud-mcp-server
description: A Helm chart for Nextcloud MCP Server - enables AI assistants to interact with Nextcloud
type: application
version: 0.52.0
appVersion: "0.52.0"
version: 0.50.0
appVersion: "0.50.0"
keywords:
- nextcloud
- mcp
+1 -1
View File
@@ -21,7 +21,7 @@ services:
restart: always
app:
image: docker.io/library/nextcloud:32.0.3@sha256:54993ed39dc77f7a6ade142b1625972cb7a9393074325373402d47231314afbb
image: docker.io/library/nextcloud:32.0.2@sha256:04cc19547e586ac75e08dd056c11330d4ce4c5c561c89405b326180a37c19afb
restart: always
ports:
- 0.0.0.0:8080:80
-104
View File
@@ -1,104 +0,0 @@
# MCP 1.23.x DNS Rebinding Protection Fix
## Problem
MCP Python SDK 1.23.0 introduced **automatic DNS rebinding protection** that breaks containerized deployments (Kubernetes, Docker) when the protection is unintentionally auto-enabled.
### Root Cause
From `mcp/server/fastmcp/server.py:177-183` in the Python SDK:
```python
# Auto-enable DNS rebinding protection for localhost (IPv4 and IPv6)
if transport_security is None and host in ("127.0.0.1", "localhost", "::1"):
transport_security = TransportSecuritySettings(
enable_dns_rebinding_protection=True,
allowed_hosts=["127.0.0.1:*", "localhost:*", "[::1]:*"],
allowed_origins=["http://127.0.0.1:*", "http://localhost:*", "http://[::1]:*"],
)
```
### What Was Happening
1. **FastMCP initialization** in `app.py` didn't pass `host` or `transport_security` parameters
2. **Defaults applied**: `host="127.0.0.1"`, `transport_security=None`
3. **Auto-enablement triggered**: Condition `transport_security is None and host == "127.0.0.1"` was TRUE
4. **Protection activated** with `allowed_hosts=["127.0.0.1:*", "localhost:*", "[::1]:*"]`
5. **Kubernetes requests rejected**: `Host: nextcloud-mcp-server.default.svc.cluster.local:8000` didn't match allowed hosts
### Why `--host 0.0.0.0` Didn't Help
The `--host` CLI flag (used in Dockerfile/docker-compose) controls **uvicorn's bind address**, NOT the **FastMCP `host` parameter**. These are separate concerns:
- **Uvicorn bind address** (`--host 0.0.0.0`): Where the HTTP server listens
- **FastMCP host parameter** (defaulted to `"127.0.0.1"`): Used for auto-enablement logic
## Solution
Explicitly disable DNS rebinding protection by passing `transport_security=TransportSecuritySettings(enable_dns_rebinding_protection=False)` to all FastMCP instances.
### Changes Made
Modified `nextcloud_mcp_server/app.py`:
1. **Import** `TransportSecuritySettings` from `mcp.server.transport_security`
2. **Updated all three FastMCP initializations**:
- OAuth mode (line 1015)
- Smithery stateless mode (line 1030)
- BasicAuth mode (line 1040)
Each now includes:
```python
transport_security=TransportSecuritySettings(enable_dns_rebinding_protection=False)
```
## Impact
### ✅ What This Fixes
- **Kubernetes deployments**: Requests with k8s service DNS names now work
- **Docker deployments**: Port-mapped requests (localhost:8000 → container) now work
- **Reverse proxy deployments**: Proxied requests with various Host headers now work
- **Ingress controllers**: Requests via ingress hostnames now work
### 🔒 Security Considerations
DNS rebinding protection defends against attacks where:
1. Attacker controls a DNS domain (e.g., `evil.com`)
2. DNS initially resolves to attacker's IP
3. After victim's browser caches the origin, DNS changes to victim's localhost
4. Attacker's page can now make requests to victim's localhost services
**Why it's safe to disable for this deployment:**
1. **OAuth authentication required** in production deployments (ADR-002, ADR-004)
2. **Network-level isolation** in containerized environments (k8s network policies, Docker networks)
3. **MCP is server-to-server**, not exposed to browsers (no CORS concerns)
4. **Host header validation inappropriate** for multi-tenant k8s environments
If DNS rebinding protection is needed for specific deployments, it can be re-enabled with a custom allowed hosts list:
```python
transport_security=TransportSecuritySettings(
enable_dns_rebinding_protection=True,
allowed_hosts=[
"nextcloud-mcp-server.default.svc.cluster.local:*",
"mcp.example.com:*",
# Add all your expected Host header values
]
)
```
## Testing
- ✅ Ruff linting passes
- ✅ Type checking passes (pre-existing warnings unrelated)
- ✅ Module imports successfully
- ✅ Compatible with MCP 1.23.x
## References
- [MCP Python SDK 1.23.0 Release](https://github.com/modelcontextprotocol/python-sdk/releases/tag/v1.23.0)
- Commit: `d3a1841` - "Auto-enable DNS rebinding protection for localhost servers"
- Issue #373 (original report of k8s breakage)
- PR #382 (MCP 1.23.x upgrade)
+4 -4
View File
@@ -5,7 +5,7 @@ This document explains the architecture of the semantic search feature in the Ne
> [!IMPORTANT]
> **Status: Experimental**
> - Disabled by default (`VECTOR_SYNC_ENABLED=false`)
> - Currently supports **Notes, Files (PDFs), News items, and Deck cards**
> - Currently supports **Notes app only** (multi-app architecture ready, additional apps planned)
> - Requires additional infrastructure (Qdrant vector database + Ollama embedding service)
> - RAG answer generation requires MCP client sampling support
@@ -39,9 +39,9 @@ Semantic search enables:
### Current Support
- **Supported Apps**: Notes, Files (PDFs with text extraction), News items, Deck cards
- **Planned Apps**: Calendar events, Calendar tasks, Contacts
- **Architecture**: Multi-app plugin system ready for additional apps
- **Supported Apps**: Notes (fully implemented)
- **Planned Apps**: Calendar events, Calendar tasks, Deck cards, Files (with text extraction), Contacts
- **Architecture**: Multi-app plugin system ready, awaiting implementation
## System Components
+2 -23
View File
@@ -19,7 +19,6 @@ import httpx
from anyio.streams.memory import MemoryObjectReceiveStream, MemoryObjectSendStream
from mcp.server.auth.settings import AuthSettings
from mcp.server.fastmcp import Context, FastMCP
from mcp.server.transport_security import TransportSecuritySettings
from pydantic import AnyHttpUrl
from starlette.applications import Starlette
from starlette.middleware.authentication import AuthenticationMiddleware
@@ -1017,11 +1016,6 @@ def get_app(transport: str = "streamable-http", enabled_apps: list[str] | None =
lifespan=oauth_lifespan,
token_verifier=token_verifier,
auth=auth_settings,
# Disable DNS rebinding protection for containerized deployments (k8s, Docker)
# MCP 1.23+ auto-enables this for localhost, breaking k8s service DNS names
transport_security=TransportSecuritySettings(
enable_dns_rebinding_protection=False
),
)
else:
# ADR-016: Use Smithery lifespan for stateless mode, BasicAuth otherwise
@@ -1030,26 +1024,11 @@ def get_app(transport: str = "streamable-http", enabled_apps: list[str] | None =
# json_response=True returns plain JSON-RPC instead of SSE format,
# required for Smithery scanner compatibility
mcp = FastMCP(
"Nextcloud MCP",
lifespan=app_lifespan_smithery,
json_response=True,
# Disable DNS rebinding protection for containerized deployments (k8s, Docker)
# MCP 1.23+ auto-enables this for localhost, breaking k8s service DNS names
transport_security=TransportSecuritySettings(
enable_dns_rebinding_protection=False
),
"Nextcloud MCP", lifespan=app_lifespan_smithery, json_response=True
)
else:
logger.info("Configuring MCP server for BasicAuth mode")
mcp = FastMCP(
"Nextcloud MCP",
lifespan=app_lifespan_basic,
# Disable DNS rebinding protection for containerized deployments (k8s, Docker)
# MCP 1.23+ auto-enables this for localhost, breaking k8s service DNS names
transport_security=TransportSecuritySettings(
enable_dns_rebinding_protection=False
),
)
mcp = FastMCP("Nextcloud MCP", lifespan=app_lifespan_basic)
@mcp.resource("nc://capabilities")
async def nc_get_capabilities():
@@ -201,15 +201,8 @@ function vizApp() {
return `${baseUrl}/apps/calendar`;
case 'contact':
return `${baseUrl}/apps/contacts`;
case 'deck_card':
// URL pattern: /apps/deck/board/:boardId/card/:cardId
if (result.metadata && result.metadata.board_id) {
return `${baseUrl}/apps/deck/board/${result.metadata.board_id}/card/${result.id}`;
}
// Fallback if board_id not available
case 'deck':
return `${baseUrl}/apps/deck`;
case 'news_item':
return `${baseUrl}/apps/news/item/${result.id}`;
default:
return `${baseUrl}`;
}
@@ -65,12 +65,8 @@
<span>Contacts</span>
</label>
<label style="display: flex; align-items: center; cursor: pointer; font-weight: normal;">
<input type="checkbox" x-model="docTypes" value="deck_card" style="margin-right: 4px;">
<span>Deck Cards</span>
</label>
<label style="display: flex; align-items: center; cursor: pointer; font-weight: normal;">
<input type="checkbox" x-model="docTypes" value="news_item" style="margin-right: 4px;">
<span>News</span>
<input type="checkbox" x-model="docTypes" value="deck" style="margin-right: 4px;">
<span>Deck</span>
</label>
</div>
</div>
-2
View File
@@ -298,7 +298,6 @@ async def vector_visualization_search(request: Request) -> JSONResponse:
"title": r.title,
"excerpt": r.excerpt,
"score": r.score,
"metadata": r.metadata,
}
for r in search_results
],
@@ -459,7 +458,6 @@ async def vector_visualization_search(request: Request) -> JSONResponse:
), # Raw score from algorithm
"chunk_start_offset": r.chunk_start_offset,
"chunk_end_offset": r.chunk_end_offset,
"metadata": r.metadata, # Include metadata (e.g., board_id for deck_card)
}
for r in search_results
]
+3 -12
View File
@@ -228,10 +228,6 @@ class NewsClient(BaseNextcloudClient):
async def get_item(self, item_id: int) -> dict[str, Any]:
"""Get a specific item by ID.
Note: The News API doesn't have a direct single-item endpoint,
so we fetch all items and filter. For efficiency, consider
caching or using get_items with specific feed if known.
Args:
item_id: Item ID
@@ -239,15 +235,10 @@ class NewsClient(BaseNextcloudClient):
Item data
Raises:
ValueError: If item not found
HTTPStatusError: 404 if item not found
"""
# Fetch all items and find the one we need
# This is inefficient but the API doesn't provide a direct endpoint
items = await self.get_items(batch_size=-1, get_read=True)
for item in items:
if item.get("id") == item_id:
return item
raise ValueError(f"Item {item_id} not found")
response = await self._make_request("GET", f"{self.API_BASE}/items/{item_id}")
return response.json()
async def get_updated_items(
self,
+5 -13
View File
@@ -219,18 +219,6 @@ class BM25HybridSearchAlgorithm(SearchAlgorithm):
seen_chunks.add(chunk_key)
# Build metadata dict with common fields
metadata = {
"chunk_index": result.payload.get("chunk_index"),
"total_chunks": result.payload.get("total_chunks"),
"search_method": f"bm25_hybrid_{self.fusion_name}",
}
# Add deck_card-specific metadata for frontend URL construction
if doc_type == "deck_card":
if board_id := result.payload.get("board_id"):
metadata["board_id"] = board_id
# Return unverified results (verification happens at output stage)
results.append(
SearchResult(
@@ -239,7 +227,11 @@ class BM25HybridSearchAlgorithm(SearchAlgorithm):
title=result.payload.get("title", "Untitled"),
excerpt=result.payload.get("excerpt", ""),
score=result.score, # Fusion score (RRF or DBSF)
metadata=metadata,
metadata={
"chunk_index": result.payload.get("chunk_index"),
"total_chunks": result.payload.get("total_chunks"),
"search_method": f"bm25_hybrid_{self.fusion_name}",
},
chunk_start_offset=result.payload.get("chunk_start_offset"),
chunk_end_offset=result.payload.get("chunk_end_offset"),
page_number=result.payload.get("page_number"),
+2 -149
View File
@@ -209,64 +209,6 @@ async def _get_file_path_from_qdrant(
return None
async def _get_deck_metadata_from_qdrant(
user_id: str, card_id: int
) -> dict[str, int] | None:
"""Retrieve board_id and stack_id for a deck card from Qdrant payload.
Args:
user_id: User ID who owns the card
card_id: Card ID
Returns:
Dictionary with board_id and stack_id, or None if not found
"""
try:
from qdrant_client.models import FieldCondition, Filter, MatchValue
from nextcloud_mcp_server.config import get_settings
from nextcloud_mcp_server.vector.qdrant_client import get_qdrant_client
qdrant_client = await get_qdrant_client()
settings = get_settings()
# Query for any chunk of this card (we just need metadata)
scroll_result = await qdrant_client.scroll(
collection_name=settings.get_collection_name(),
scroll_filter=Filter(
must=[
FieldCondition(key="user_id", match=MatchValue(value=user_id)),
FieldCondition(key="doc_id", match=MatchValue(value=card_id)),
FieldCondition(key="doc_type", match=MatchValue(value="deck_card")),
]
),
limit=1,
with_payload=["board_id", "stack_id"],
with_vectors=False,
)
if scroll_result[0]:
point = scroll_result[0][0]
board_id = point.payload.get("board_id")
stack_id = point.payload.get("stack_id")
if board_id is not None and stack_id is not None:
logger.debug(
f"Retrieved deck metadata for card {card_id}: "
f"board_id={board_id}, stack_id={stack_id}"
)
return {"board_id": int(board_id), "stack_id": int(stack_id)}
logger.debug(
f"Could not find deck metadata in Qdrant for card {card_id} "
f"(might be legacy data without board_id/stack_id)"
)
return None
except Exception as e:
logger.debug(f"Error querying Qdrant for deck metadata: {e}")
return None
@dataclass
class ChunkContext:
"""Expanded chunk with surrounding context and position markers.
@@ -452,9 +394,7 @@ async def get_chunk_with_context(
logger.debug(f"Resolved file_id {doc_id} to file_path {file_path}")
# Fetch full document text
full_text = await _fetch_document_text(
nc_client, resolved_doc_id, doc_type, user_id
)
full_text = await _fetch_document_text(nc_client, resolved_doc_id, doc_type)
if full_text is None:
logger.warning(
f"Could not fetch document text for {doc_type} {doc_id}, "
@@ -513,7 +453,7 @@ async def get_chunk_with_context(
async def _fetch_document_text(
nc_client: NextcloudClient, doc_id: str | int, doc_type: str, user_id: str
nc_client: NextcloudClient, doc_id: str | int, doc_type: str
) -> str | None:
"""Fetch full text content of a document.
@@ -584,93 +524,6 @@ async def _fetch_document_text(
f"Error fetching file content for {doc_id}: {e}", exc_info=True
)
return None
elif doc_type == "news_item":
# Fetch news item by ID
from nextcloud_mcp_server.vector.html_processor import html_to_markdown
item = await nc_client.news.get_item(int(doc_id))
# Reconstruct full content as indexed: title + source + URL + body
# This ensures chunk offsets align with indexed content structure
body_markdown = html_to_markdown(item.get("body", ""))
item_title = item.get("title", "")
item_url = item.get("url", "")
feed_title = item.get("feedTitle", "")
content_parts = [item_title]
if feed_title:
content_parts.append(f"Source: {feed_title}")
if item_url:
content_parts.append(f"URL: {item_url}")
content_parts.append("") # Blank line
content_parts.append(body_markdown)
return "\n".join(content_parts)
elif doc_type == "deck_card":
# Fetch card from Deck API
# Try to get board_id/stack_id from Qdrant metadata (O(1) lookup)
# Otherwise fall back to iteration (legacy data)
card = None
deck_metadata = await _get_deck_metadata_from_qdrant(user_id, int(doc_id))
if deck_metadata:
# Fast path: Direct lookup with known board_id/stack_id
board_id = deck_metadata["board_id"]
stack_id = deck_metadata["stack_id"]
try:
card = await nc_client.deck.get_card(
board_id=board_id, stack_id=stack_id, card_id=int(doc_id)
)
logger.debug(
f"Retrieved deck card {doc_id} using metadata "
f"(board_id={board_id}, stack_id={stack_id})"
)
except Exception as e:
logger.warning(
f"Failed to fetch card with metadata (board_id={board_id}, "
f"stack_id={stack_id}, card_id={doc_id}): {e}, falling back to iteration"
)
# Fallback: Iterate through all boards/stacks (for legacy data or if fast path failed)
if card is None:
boards = await nc_client.deck.get_boards()
card_found = False
for board in boards:
if card_found:
break
# Skip deleted boards (soft delete: deletedAt > 0)
if board.deletedAt > 0:
logger.debug(
f"Skipping deleted board {board.id} while searching for card {doc_id}"
)
continue
stacks = await nc_client.deck.get_stacks(board.id)
for stack in stacks:
if card_found:
break
if stack.cards:
for c in stack.cards:
if c.id == int(doc_id):
card = c
card_found = True
logger.debug(
f"Found deck card {doc_id} in board {board.id}, "
f"stack {stack.id} (fallback iteration)"
)
break
if not card_found:
logger.warning(f"Deck card {doc_id} not found in any board/stack")
return None
# Reconstruct full content as indexed: title + "\n\n" + description
# This ensures chunk offsets align with indexed content structure
content_parts = [card.title]
if card.description:
content_parts.append(card.description)
return "\n\n".join(content_parts)
else:
logger.warning(f"Unsupported doc_type for context expansion: {doc_type}")
return None
+4 -12
View File
@@ -151,17 +151,6 @@ class SemanticSearchAlgorithm(SearchAlgorithm):
seen_chunks.add(chunk_key)
# Build metadata dict with common fields
metadata = {
"chunk_index": result.payload.get("chunk_index"),
"total_chunks": result.payload.get("total_chunks"),
}
# Add deck_card-specific metadata for frontend URL construction
if doc_type == "deck_card":
if board_id := result.payload.get("board_id"):
metadata["board_id"] = board_id
# Return unverified results (verification happens at output stage)
results.append(
SearchResult(
@@ -170,7 +159,10 @@ class SemanticSearchAlgorithm(SearchAlgorithm):
title=result.payload.get("title", "Untitled"),
excerpt=result.payload.get("excerpt", ""),
score=result.score,
metadata=metadata,
metadata={
"chunk_index": result.payload.get("chunk_index"),
"total_chunks": result.payload.get("total_chunks"),
},
chunk_start_offset=result.payload.get("chunk_start_offset"),
chunk_end_offset=result.payload.get("chunk_end_offset"),
page_number=result.payload.get("page_number"),
+3 -3
View File
@@ -65,13 +65,13 @@ def configure_semantic_tools(mcp: FastMCP):
database for optimal relevance. This provides the best of both semantic
understanding and keyword precision.
Requires VECTOR_SYNC_ENABLED=true. Supports indexing of notes, files,
news items, and deck cards.
Requires VECTOR_SYNC_ENABLED=true. Currently only "note" documents are
fully supported for indexing.
Args:
query: Natural language or keyword search query
limit: Maximum number of results to return (default: 10)
doc_types: Document types to search (e.g., ["note", "file", "deck_card", "news_item"]). None = search all indexed types (default)
doc_types: Document types to search (e.g., ["note", "file"]). None = search all indexed types (default)
score_threshold: Minimum fusion score (0-1, default: 0.0)
fusion: Fusion algorithm: "rrf" (Reciprocal Rank Fusion, default) or "dbsf" (Distribution-Based Score Fusion)
RRF: Good general-purpose fusion using reciprocal ranks
+5 -116
View File
@@ -6,7 +6,6 @@ Processes documents from stream: fetches content, generates embeddings, stores i
import logging
import time
import uuid
from typing import Any, cast
import anyio
from anyio.abc import TaskStatus
@@ -312,97 +311,6 @@ async def _index_document(
file_path = None
content_bytes = None
content_type = None
elif doc_task.doc_type == "deck_card":
# Fetch card from Deck API
# Use metadata from scanner if available (O(1) lookup)
# Otherwise fall back to iteration (legacy data)
card = None
board = None
stack = None
if (
doc_task.metadata
and "board_id" in doc_task.metadata
and "stack_id" in doc_task.metadata
):
# Fast path: Direct lookup with known board_id/stack_id
board_id = doc_task.metadata["board_id"]
stack_id = doc_task.metadata["stack_id"]
try:
card = await nc_client.deck.get_card(
board_id=int(board_id),
stack_id=int(stack_id),
card_id=int(doc_task.doc_id),
)
# Fetch board and stack info for metadata
boards = await nc_client.deck.get_boards()
for b in boards:
if b.id == int(board_id):
board = b
stacks = await nc_client.deck.get_stacks(b.id)
for s in stacks:
if s.id == int(stack_id):
stack = s
break
break
except Exception as e:
logger.warning(
f"Failed to fetch card with metadata (board_id={board_id}, stack_id={stack_id}, card_id={doc_task.doc_id}): {e}, falling back to iteration"
)
# Fallback: Iterate through all boards/stacks (for legacy data or if fast path failed)
if card is None:
boards = await nc_client.deck.get_boards()
card_found = False
for b in boards:
if card_found:
break
# Skip deleted boards (soft delete: deletedAt > 0)
if b.deletedAt > 0:
continue
stacks = await nc_client.deck.get_stacks(b.id)
for s in stacks:
if card_found:
break
if s.cards:
for c in s.cards:
if c.id == int(doc_task.doc_id):
card = c
board = b
stack = s
card_found = True
break
if not card_found:
raise ValueError(
f"Deck card {doc_task.doc_id} not found in any board/stack"
)
# Build content from card title and description
content_parts = [card.title]
if card.description:
content_parts.append(card.description)
content = "\n\n".join(content_parts)
title = card.title
# Store deck-specific metadata
file_metadata = {
"board_id": board.id,
"board_title": board.title,
"stack_id": stack.id,
"stack_title": stack.title,
"card_type": card.type,
"duedate": (card.duedate.isoformat() if card.duedate else None),
"archived": card.archived,
"owner": (
card.owner.uid if hasattr(card.owner, "uid") else str(card.owner)
),
}
etag = card.etag or ""
file_path = None
content_bytes = None
content_type = None
elif doc_task.doc_type == "file":
# For files, doc_id is now the numeric file ID, file_path comes from DocumentTask
if not doc_task.file_path:
@@ -491,16 +399,14 @@ async def _index_document(
# Assign page numbers to chunks if page boundaries are available (PDFs)
page_boundaries = file_metadata.get("page_boundaries")
if doc_task.doc_type == "file" and page_boundaries is not None:
# Type narrowing: page_boundaries is guaranteed to be list[dict] here
page_boundaries_list = cast(list[dict[str, Any]], page_boundaries)
with trace_operation(
"vector_sync.assign_page_numbers",
attributes={
"vector_sync.chunk_count": len(chunks),
"vector_sync.page_count": len(page_boundaries_list),
"vector_sync.page_count": len(page_boundaries),
},
):
assign_page_numbers(chunks, page_boundaries_list)
assign_page_numbers(chunks, page_boundaries)
# Diagnostic: Verify page number assignment
assigned_count = sum(1 for c in chunks if c.page_number is not None)
@@ -523,8 +429,8 @@ async def _index_document(
f"Text length: {len(content)}, "
f"Chunks: {len(chunks)}, "
f"Chunk offset range: [{chunks[0].start_offset}:{chunks[-1].end_offset}], "
f"Page boundaries: {len(page_boundaries_list)} pages, "
f"First boundary: {page_boundaries_list[0] if page_boundaries_list else 'None'}"
f"Page boundaries: {len(page_boundaries)} pages, "
f"First boundary: {page_boundaries[0] if page_boundaries else 'None'}"
)
# Extract chunk texts for embedding
@@ -598,9 +504,6 @@ async def _index_document(
logger.warning("No page boundaries available, skipping highlighting")
return
# Type narrowing: page_boundaries is guaranteed to be list[dict] here
page_boundaries_list = cast(list[dict[str, Any]], page_boundaries)
logger.info(
f"Batch generating highlighted page images for {len(chunk_data)} PDF chunks"
)
@@ -611,7 +514,7 @@ async def _index_document(
lambda: PDFHighlighter.highlight_chunks_batch(
pdf_bytes=content_bytes,
chunks=chunk_data,
page_boundaries=page_boundaries_list,
page_boundaries=page_boundaries,
full_text=content,
color="yellow",
zoom=2.0,
@@ -720,20 +623,6 @@ async def _index_document(
if doc_task.doc_type == "news_item"
else {}
),
# Deck card-specific metadata
**(
{
"board_id": file_metadata.get("board_id"),
"board_title": file_metadata.get("board_title"),
"stack_id": file_metadata.get("stack_id"),
"stack_title": file_metadata.get("stack_title"),
"card_type": file_metadata.get("card_type"),
"duedate": file_metadata.get("duedate"),
"owner": file_metadata.get("owner"),
}
if doc_task.doc_type == "deck_card"
else {}
),
# Highlighted page image (PDF only)
**(
{
+3 -221
View File
@@ -36,9 +36,6 @@ class DocumentTask:
operation: str # "index" or "delete"
modified_at: int
file_path: str | None = None # File path for files (when doc_id is file_id)
metadata: dict[str, int | str] | None = (
None # Additional metadata (e.g., board_id/stack_id for deck_card)
)
# Track documents potentially deleted (grace period before actual deletion)
@@ -82,11 +79,9 @@ async def get_last_indexed_timestamp(user_id: str) -> int | None:
if scroll_result[0]:
timestamps = [
point.payload.get("indexed_at", 0)
for point in scroll_result[0]
if point.payload is not None
point.payload.get("indexed_at", 0) for point in scroll_result[0]
]
max_timestamp = max(timestamps) if timestamps else 0
max_timestamp = max(timestamps)
logger.info(
f"Max indexed_at: {max_timestamp}, timestamps sample: {timestamps[:3]}"
)
@@ -569,23 +564,9 @@ async def scan_user_documents(
except Exception as e:
logger.warning(f"Failed to scan news items for {user_id}: {e}")
# Scan Deck cards
deck_queued = 0
try:
deck_queued = await scan_deck_cards(
user_id=user_id,
send_stream=send_stream,
nc_client=nc_client,
initial_sync=initial_sync,
scan_id=scan_id,
)
queued += deck_queued
except Exception as e:
logger.warning(f"Failed to scan deck cards for {user_id}: {e}")
if queued > 0:
logger.info(
f"Sent {queued} documents ({file_queued} files, {news_queued} news items, {deck_queued} deck cards) for incremental sync: {user_id}"
f"Sent {queued} documents ({file_queued} files, {news_queued} news items) for incremental sync: {user_id}"
)
else:
logger.debug(f"No changes detected for {user_id}")
@@ -772,202 +753,3 @@ async def scan_news_items(
_potentially_deleted[doc_key] = current_time
return queued
async def scan_deck_cards(
user_id: str,
send_stream: MemoryObjectSendStream[DocumentTask],
nc_client: NextcloudClient,
initial_sync: bool,
scan_id: int,
) -> int:
"""
Scan user's Deck cards and queue changed cards for indexing.
Indexes cards from all non-archived boards and stacks.
Args:
user_id: User to scan
send_stream: Stream to send changed documents to processors
nc_client: Authenticated Nextcloud client
initial_sync: If True, send all documents (first-time sync)
scan_id: Scan identifier for logging
Returns:
Number of cards queued for processing
"""
settings = get_settings()
queued = 0
# Get indexed deck card IDs from Qdrant (for deletion tracking)
indexed_card_ids: set[str] = set()
if not initial_sync:
qdrant_client = await get_qdrant_client()
scroll_result = await qdrant_client.scroll(
collection_name=settings.get_collection_name(),
scroll_filter=Filter(
must=[
FieldCondition(key="user_id", match=MatchValue(value=user_id)),
FieldCondition(key="doc_type", match=MatchValue(value="deck_card")),
]
),
with_payload=["doc_id"],
with_vectors=False,
limit=10000,
)
indexed_card_ids = {
point.payload["doc_id"]
for point in (scroll_result[0] or [])
if point.payload is not None
}
logger.debug(f"Found {len(indexed_card_ids)} indexed deck cards in Qdrant")
# Fetch all boards
boards = await nc_client.deck.get_boards()
logger.debug(f"[SCAN-{scan_id}] Found {len(boards)} deck boards")
card_count = 0
nextcloud_card_ids: set[str] = set()
# Iterate through boards
for board in boards:
# Skip archived boards
if board.archived:
continue
# Skip deleted boards (soft delete: deletedAt > 0)
if board.deletedAt > 0:
logger.debug(f"[SCAN-{scan_id}] Skipping deleted board {board.id}")
continue
# Get stacks for this board
stacks = await nc_client.deck.get_stacks(board.id)
# Iterate through stacks
for stack in stacks:
# Skip if stack has no cards
if not stack.cards:
continue
# Iterate through cards in stack
for card in stack.cards:
# Skip archived cards
if card.archived:
continue
card_count += 1
doc_id = str(card.id)
nextcloud_card_ids.add(doc_id)
# Use lastModified timestamp if available
modified_at = card.lastModified or 0
if initial_sync:
# Send everything on first sync - write placeholder first
await write_placeholder_point(
doc_id=doc_id,
doc_type="deck_card",
user_id=user_id,
modified_at=modified_at,
)
await send_stream.send(
DocumentTask(
user_id=user_id,
doc_id=doc_id,
doc_type="deck_card",
operation="index",
modified_at=modified_at,
metadata={"board_id": board.id, "stack_id": stack.id},
)
)
queued += 1
else:
# Incremental sync: check if card exists and compare modified_at
doc_key = (user_id, doc_id)
if doc_key in _potentially_deleted:
logger.debug(
f"Deck card {doc_id} reappeared, removing from deletion grace period"
)
del _potentially_deleted[doc_key]
# Query Qdrant for existing entry
existing_metadata = await query_document_metadata(
doc_id=doc_id, doc_type="deck_card", user_id=user_id
)
needs_indexing = False
if existing_metadata is None:
needs_indexing = True
elif existing_metadata.get("modified_at", 0) < modified_at:
needs_indexing = True
elif existing_metadata.get("is_placeholder", False):
queued_at = existing_metadata.get("queued_at", 0)
placeholder_age = time.time() - queued_at
stale_threshold = settings.vector_sync_scan_interval * 5
if placeholder_age > stale_threshold:
logger.debug(
f"Found stale placeholder for deck card {doc_id} "
f"(age={placeholder_age:.1f}s), requeuing"
)
needs_indexing = True
if needs_indexing:
await write_placeholder_point(
doc_id=doc_id,
doc_type="deck_card",
user_id=user_id,
modified_at=modified_at,
)
await send_stream.send(
DocumentTask(
user_id=user_id,
doc_id=doc_id,
doc_type="deck_card",
operation="index",
modified_at=modified_at,
metadata={"board_id": board.id, "stack_id": stack.id},
)
)
queued += 1
logger.info(
f"[SCAN-{scan_id}] Found {card_count} deck cards (non-archived) for {user_id}"
)
record_vector_sync_scan(card_count)
# Check for deleted cards (not initial sync)
if not initial_sync:
grace_period = settings.vector_sync_scan_interval * 1.5
current_time = time.time()
for doc_id in indexed_card_ids:
if doc_id not in nextcloud_card_ids:
doc_key = (user_id, doc_id)
if doc_key in _potentially_deleted:
first_missing_time = _potentially_deleted[doc_key]
time_missing = current_time - first_missing_time
if time_missing >= grace_period:
logger.info(
f"Deck card {doc_id} missing for {time_missing:.1f}s "
f"(>{grace_period:.1f}s grace period), sending deletion"
)
await send_stream.send(
DocumentTask(
user_id=user_id,
doc_id=doc_id,
doc_type="deck_card",
operation="delete",
modified_at=0,
)
)
queued += 1
del _potentially_deleted[doc_key]
else:
logger.debug(
f"Deck card {doc_id} missing for first time, starting grace period"
)
_potentially_deleted[doc_key] = current_time
return queued
+2 -3
View File
@@ -1,6 +1,6 @@
[project]
name = "nextcloud-mcp-server"
version = "0.52.0"
version = "0.50.0"
description = "Model Context Protocol (MCP) server for Nextcloud integration - enables AI assistants to interact with Nextcloud data"
authors = [
{name = "Chris Coutinho", email = "chris@coutinho.io"}
@@ -10,7 +10,7 @@ license = {text = "AGPL-3.0-only"}
requires-python = ">=3.11"
keywords = ["nextcloud", "mcp", "model-context-protocol", "llm", "ai", "claude", "webdav", "caldav", "carddav"]
dependencies = [
"mcp[cli] (>=1.23,<1.24)",
"mcp[cli] (>=1.22,<1.23)",
"httpx (>=0.28.1,<0.29.0)",
"pillow (>=10.3.0,<12.0.0)", # Compatible with fastembed
"icalendar (>=6.0.0,<7.0.0)",
@@ -101,7 +101,6 @@ extend-select = ["I"]
[tool.uv.sources]
caldav = { git = "https://github.com/cbcoutinho/caldav", branch = "feature/httpx" }
qdrant-client = { git = "https://github.com/cbcoutinho/qdrant-client", branch = "fix/fusion-score-threshold" }
[build-system]
requires = ["uv_build>=0.9.4,<0.10.0"]
+7 -26
View File
@@ -310,16 +310,14 @@ async def test_news_api_get_items_unread_only(mocker):
async def test_news_api_get_item(mocker):
"""Test that get_item fetches all items and filters for the requested ID."""
# Create multiple items, only one should be returned
items = [
create_mock_news_item(item_id=100, title="Other Item 1"),
create_mock_news_item(item_id=123, title="Single Item"),
create_mock_news_item(item_id=200, title="Other Item 2"),
]
"""Test that get_item fetches a single item by ID."""
item = create_mock_news_item(item_id=123, title="Single Item")
mock_response = create_mock_response(status_code=200, json_data=item)
mock_client = mocker.AsyncMock(spec=httpx.AsyncClient)
mock_get_items = mocker.patch.object(NewsClient, "get_items", return_value=items)
mock_make_request = mocker.patch.object(
NewsClient, "_make_request", return_value=mock_response
)
client = NewsClient(mock_client, "testuser")
result = await client.get_item(item_id=123)
@@ -327,24 +325,7 @@ async def test_news_api_get_item(mocker):
assert result["id"] == 123
assert result["title"] == "Single Item"
# Verify it fetched all items with correct params
mock_get_items.assert_called_once_with(batch_size=-1, get_read=True)
async def test_news_api_get_item_not_found(mocker):
"""Test that get_item raises ValueError when item not found."""
items = [
create_mock_news_item(item_id=100, title="Item 1"),
create_mock_news_item(item_id=200, title="Item 2"),
]
mock_client = mocker.AsyncMock(spec=httpx.AsyncClient)
mocker.patch.object(NewsClient, "get_items", return_value=items)
client = NewsClient(mock_client, "testuser")
with pytest.raises(ValueError, match="Item 999 not found"):
await client.get_item(item_id=999)
mock_make_request.assert_called_once_with("GET", "/apps/news/api/v1-3/items/123")
async def test_news_api_get_updated_items(mocker):
@@ -1,238 +0,0 @@
"""Integration tests for Deck card vector search.
These tests validate that Deck cards are properly indexed and searchable
via semantic search.
"""
import pytest
pytestmark = [pytest.mark.integration, pytest.mark.smoke]
async def test_deck_card_semantic_search(nc_mcp_client, nc_client, mocker):
"""Test that Deck cards can be indexed and searched via semantic search.
This test:
1. Creates a Deck board with a card
2. Manually triggers indexing (simulates vector sync)
3. Performs semantic search filtering by deck_card doc_type
4. Verifies the card is found in results
"""
# Skip if vector sync is not enabled
settings_response = await nc_mcp_client.call_tool("nc_get_vector_sync_status", {})
if settings_response.isError:
pytest.skip("Vector sync not enabled")
# Create a test board
board_title = "Test Board for Vector Search"
board = await nc_client.deck.create_board(title=board_title, color="ff0000")
try:
# Create a stack for the board
stack = await nc_client.deck.create_stack(
board_id=board.id, title="Test Stack", order=0
)
# Create a test card with searchable content
card_title = "Machine Learning Project Plan"
card_description = """
# ML Project Outline
## Phase 1: Data Collection
- Gather training data from multiple sources
- Clean and preprocess the dataset
## Phase 2: Model Training
- Experiment with different neural network architectures
- Use gradient descent optimization
## Phase 3: Deployment
- Deploy model to production environment
- Monitor performance metrics
"""
card = await nc_client.deck.create_card(
board_id=board.id,
stack_id=stack.id,
title=card_title,
description=card_description,
)
# Note: In a real integration test with vector sync enabled,
# we would wait for the background scanner to index the card.
# For now, we'll test the scanning function directly if needed.
# TODO: Once vector sync is running in test environment,
# add actual semantic search test here
# For now, just verify the card was created successfully
assert card.id is not None
assert card.title == card_title
assert card.description == card_description
# Test semantic search with deck_card filter
# Note: This will only work if vector sync is actually running
# and the card has been indexed
try:
search_result = await nc_mcp_client.call_tool(
"nc_semantic_search",
{
"query": "machine learning neural networks",
"doc_types": ["deck_card"],
"limit": 10,
},
)
# If vector sync is working, we should find the card
if not search_result.isError:
data = search_result.structuredContent
results = data.get("results", [])
# Check if our card is in the results
found_card = any(
r.get("doc_type") == "deck_card" and r.get("title") == card_title
for r in results
)
# Log result for debugging
if found_card:
print("✓ Successfully found Deck card in vector search")
else:
print(
"⚠ Deck card not found in search (may need time for indexing)"
)
except Exception as e:
# If search fails, it might be because indexing hasn't happened yet
print(f"⚠ Semantic search failed (indexing may not be complete): {e}")
finally:
# Cleanup: delete the board
try:
await nc_client.deck.delete_board(board.id)
except Exception as e:
print(f"Warning: Failed to cleanup test board: {e}")
async def test_deck_card_appears_in_cross_app_search(nc_mcp_client, nc_client):
"""Test that Deck cards appear in cross-app semantic search (no doc_type filter).
This verifies that when searching without specifying doc_types,
Deck cards are included in the results alongside notes, files, etc.
"""
# Skip if vector sync is not enabled
settings_response = await nc_mcp_client.call_tool("nc_get_vector_sync_status", {})
if settings_response.isError:
pytest.skip("Vector sync not enabled")
# Create a test board with a distinctive card
board_title = "Cross-App Search Test Board"
board = await nc_client.deck.create_board(title=board_title, color="00ff00")
try:
# Create a stack for the board
stack = await nc_client.deck.create_stack(
board_id=board.id, title="Test Stack", order=0
)
# Use a very distinctive term to make it easy to find
unique_term = "xylophone_banana_unicorn_test"
_card = await nc_client.deck.create_card(
board_id=board.id,
stack_id=stack.id,
title=f"Test Card with {unique_term}",
description=f"This card contains the unique search term: {unique_term}",
)
# Test cross-app search (no doc_type filter)
try:
search_result = await nc_mcp_client.call_tool(
"nc_semantic_search",
{
"query": unique_term,
"limit": 20,
},
)
if not search_result.isError:
data = search_result.structuredContent
results = data.get("results", [])
# Check if deck_card appears in cross-app results
deck_cards_found = [
r for r in results if r.get("doc_type") == "deck_card"
]
if deck_cards_found:
print(
f"✓ Found {len(deck_cards_found)} Deck card(s) in cross-app search"
)
else:
print(
"⚠ No Deck cards in cross-app search (may need time for indexing)"
)
except Exception as e:
print(f"⚠ Cross-app search failed: {e}")
finally:
# Cleanup
try:
await nc_client.deck.delete_board(board.id)
except Exception as e:
print(f"Warning: Failed to cleanup test board: {e}")
async def test_deck_card_chunk_context(nc_client):
"""Test that Deck card chunk context can be fetched for visualization.
This test validates that the vector viz UI can display Deck card previews
by fetching the chunk context via the context expansion module.
"""
from nextcloud_mcp_server.search.context import get_chunk_with_context
# Create board, stack, and card
board = await nc_client.deck.create_board(title="Test Board", color="ff0000")
try:
stack = await nc_client.deck.create_stack(
board_id=board.id, title="Test Stack", order=0
)
card_title = "Test Card for Context Expansion"
card_description = "This is a test description that should be fetched by the context expansion module when displaying chunk previews in the vector visualization UI."
card = await nc_client.deck.create_card(
board_id=board.id,
stack_id=stack.id,
title=card_title,
description=card_description,
)
# Fetch chunk context (simulates viz UI request)
# The chunk spans the title, so start=0 and end=len(card_title)
context = await get_chunk_with_context(
nc_client=nc_client,
user_id=nc_client.username,
doc_id=card.id,
doc_type="deck_card",
chunk_start=0,
chunk_end=len(card_title),
context_chars=100,
)
# Verify context was fetched successfully
assert context is not None, "Chunk context should not be None"
assert card_title in context.chunk_text, (
f"Card title '{card_title}' should be in chunk_text"
)
# Verify context includes description
assert card_description[:50] in context.after_context, (
"Card description should be in after_context"
)
print(f"✓ Successfully fetched chunk context for Deck card {card.id}")
finally:
# Cleanup
try:
await nc_client.deck.delete_board(board.id)
except Exception as e:
print(f"Warning: Failed to cleanup test board: {e}")
-174
View File
@@ -1,174 +0,0 @@
"""
Test that DNS rebinding protection is properly disabled for containerized deployments.
This test verifies that the fix for MCP 1.23.x DNS rebinding protection works correctly.
Without the fix, requests with Host headers that don't match the default allowed list
(127.0.0.1:*, localhost:*, [::1]:*) would be rejected with a 421 Misdirected Request error.
"""
import httpx
import pytest
@pytest.mark.integration
async def test_accepts_various_host_headers():
"""Test that the MCP server accepts requests with various Host headers.
This test simulates what happens in containerized deployments where the Host
header might be a k8s service DNS name, a proxied hostname, or other values
that don't match the default allowed list.
Without the DNS rebinding protection fix, these requests would fail with:
- 421 Misdirected Request (for Host header mismatch)
- 403 Forbidden (for Origin header mismatch)
"""
mcp_url = "http://localhost:8000/mcp"
# Test various Host headers that would be rejected by DNS rebinding protection
test_cases = [
{
"name": "Kubernetes service DNS",
"headers": {
"Host": "nextcloud-mcp-server.default.svc.cluster.local:8000",
"Content-Type": "application/json",
"Accept": "application/json, text/event-stream",
},
},
{
"name": "Custom domain",
"headers": {
"Host": "mcp.example.com:8000",
"Content-Type": "application/json",
"Accept": "application/json, text/event-stream",
},
},
{
"name": "Proxied hostname",
"headers": {
"Host": "proxy.internal:8000",
"Content-Type": "application/json",
"Accept": "application/json, text/event-stream",
},
},
{
"name": "Default localhost (should always work)",
"headers": {
"Host": "localhost:8000",
"Content-Type": "application/json",
"Accept": "application/json, text/event-stream",
},
},
]
# Create a simple initialize request payload
initialize_request = {
"jsonrpc": "2.0",
"method": "initialize",
"params": {
"protocolVersion": "2024-11-05",
"capabilities": {},
"clientInfo": {"name": "test-client", "version": "1.0.0"},
},
"id": 1,
}
async with httpx.AsyncClient() as client:
for test_case in test_cases:
print(f"\n🧪 Testing: {test_case['name']}")
print(f" Host header: {test_case['headers']['Host']}")
response = await client.post(
mcp_url,
json=initialize_request,
headers=test_case["headers"],
timeout=10.0,
)
# With DNS rebinding protection enabled (MCP 1.23 default), these would fail with:
# - 421 Misdirected Request (Host header not in allowed list)
# - 403 Forbidden (Origin header not in allowed list)
#
# With our fix (enable_dns_rebinding_protection=False), they should succeed
assert response.status_code in [200, 202], (
f"Request failed for {test_case['name']}: "
f"status={response.status_code}, "
f"headers={test_case['headers']}, "
f"body={response.text[:200]}"
)
print(f" ✅ Status: {response.status_code}")
# For SSE responses (status 200), verify we got SSE format
# For JSON responses (status 202), verify we got valid JSON
if response.status_code == 200:
# SSE response - should start with "event: message" or similar
response_text = response.text
assert "event:" in response_text or "data:" in response_text, (
f"Expected SSE format for {test_case['name']}, got: {response_text[:200]}"
)
print(" ✅ Received SSE stream response")
elif response.status_code == 202:
# JSON response for notifications
response_json = response.json()
assert "jsonrpc" in response_json or response_json is None, (
f"Invalid response for {test_case['name']}: {response_json}"
)
print(" ✅ Received JSON response")
@pytest.mark.integration
async def test_dns_rebinding_protection_is_disabled():
"""Verify that DNS rebinding protection is actually disabled in the configuration.
This test makes a request that would DEFINITELY fail if DNS rebinding protection
was enabled with default settings (only allowing 127.0.0.1:*, localhost:*, [::1]:*).
"""
mcp_url = "http://localhost:8000/mcp"
# Use a Host header that would NEVER be in the default allowed list
malicious_host = "evil.attacker.com:8000"
initialize_request = {
"jsonrpc": "2.0",
"method": "initialize",
"params": {
"protocolVersion": "2024-11-05",
"capabilities": {},
"clientInfo": {"name": "test-client", "version": "1.0.0"},
},
"id": 1,
}
async with httpx.AsyncClient() as client:
response = await client.post(
mcp_url,
json=initialize_request,
headers={
"Host": malicious_host,
"Content-Type": "application/json",
"Accept": "application/json, text/event-stream",
},
timeout=10.0,
)
# If DNS rebinding protection was enabled, this would return:
# - 421 Misdirected Request (Host header validation failed)
#
# Since we disabled it, this should succeed (status 200 or 202)
assert response.status_code in [200, 202], (
f"DNS rebinding protection may still be enabled! "
f"Request with Host='{malicious_host}' was rejected: "
f"status={response.status_code}, body={response.text[:500]}"
)
# Verify we got a valid response (SSE or JSON)
if response.status_code == 200:
response_text = response.text
assert "event:" in response_text or "data:" in response_text, (
f"Expected SSE format, got: {response_text[:200]}"
)
print("✅ DNS rebinding protection is properly disabled")
print(
f" Request with Host '{malicious_host}' succeeded: {response.status_code}"
)
Generated
+12 -8
View File
@@ -1671,7 +1671,7 @@ wheels = [
[[package]]
name = "mcp"
version = "1.23.2"
version = "1.22.0"
source = { registry = "https://pypi.org/simple" }
dependencies = [
{ name = "anyio" },
@@ -1689,9 +1689,9 @@ dependencies = [
{ name = "typing-inspection" },
{ name = "uvicorn", marker = "sys_platform != 'emscripten'" },
]
sdist = { url = "https://files.pythonhosted.org/packages/39/a9/0e95530946408747ae200e86553ceda0dbd851d4ae9bbe0d02a69cbd6ad5/mcp-1.23.2.tar.gz", hash = "sha256:df4e4b7273dca2aaf428f9cf7a25bbac0c9007528a65004854b246aef3d157bc", size = 599953, upload-time = "2025-12-08T15:51:02.432Z" }
sdist = { url = "https://files.pythonhosted.org/packages/a3/a2/c5ec0ab38b35ade2ae49a90fada718fbc76811dc5aa1760414c6aaa6b08a/mcp-1.22.0.tar.gz", hash = "sha256:769b9ac90ed42134375b19e777a2858ca300f95f2e800982b3e2be62dfc0ba01", size = 471788, upload-time = "2025-11-20T20:11:28.095Z" }
wheels = [
{ url = "https://files.pythonhosted.org/packages/ad/6a/1a726905cf41a69d00989e8dfd9de7bd9b4a9f3c8723dac3077b0ba1a7b9/mcp-1.23.2-py3-none-any.whl", hash = "sha256:d8e4c6af0317ad954ea0a53dfb5e229dddea2d0a54568c080e82e8fae4a8264e", size = 231897, upload-time = "2025-12-08T15:51:01.023Z" },
{ url = "https://files.pythonhosted.org/packages/a9/bb/711099f9c6bb52770f56e56401cdfb10da5b67029f701e0df29362df4c8e/mcp-1.22.0-py3-none-any.whl", hash = "sha256:bed758e24df1ed6846989c909ba4e3df339a27b4f30f1b8b627862a4bade4e98", size = 175489, upload-time = "2025-11-20T20:11:26.542Z" },
]
[package.optional-dependencies]
@@ -1962,7 +1962,7 @@ wheels = [
[[package]]
name = "nextcloud-mcp-server"
version = "0.52.0"
version = "0.50.0"
source = { editable = "." }
dependencies = [
{ name = "aiosqlite" },
@@ -2027,7 +2027,7 @@ requires-dist = [
{ name = "jinja2", specifier = ">=3.1.6" },
{ name = "langchain-text-splitters", specifier = ">=1.0.0" },
{ name = "markdownify", specifier = ">=0.14.1" },
{ name = "mcp", extras = ["cli"], specifier = ">=1.23,<1.24" },
{ name = "mcp", extras = ["cli"], specifier = ">=1.22,<1.23" },
{ name = "openai", specifier = ">=2.8.1" },
{ name = "opentelemetry-api", specifier = ">=1.28.2" },
{ name = "opentelemetry-exporter-otlp-proto-grpc", specifier = ">=1.28.2" },
@@ -2044,7 +2044,7 @@ requires-dist = [
{ name = "pymupdf4llm", specifier = ">=0.2.2" },
{ name = "python-json-logger", specifier = ">=3.2.0" },
{ name = "pythonvcard4", specifier = ">=0.2.0" },
{ name = "qdrant-client", git = "https://github.com/cbcoutinho/qdrant-client?branch=fix%2Ffusion-score-threshold" },
{ name = "qdrant-client", specifier = ">=1.7.0" },
]
[package.metadata.requires-dev]
@@ -3329,8 +3329,8 @@ wheels = [
[[package]]
name = "qdrant-client"
version = "1.16.2"
source = { git = "https://github.com/cbcoutinho/qdrant-client?branch=fix%2Ffusion-score-threshold#a62ec3098bca86af799147695a0e2b6fb759b3aa" }
version = "1.16.1"
source = { registry = "https://pypi.org/simple" }
dependencies = [
{ name = "grpcio" },
{ name = "httpx", extra = ["http2"] },
@@ -3340,6 +3340,10 @@ dependencies = [
{ name = "pydantic" },
{ name = "urllib3" },
]
sdist = { url = "https://files.pythonhosted.org/packages/d9/68/fec3816a223c0b73b0e0036460be45c61ce2770ffb9197ac371e4f615ddc/qdrant_client-1.16.1.tar.gz", hash = "sha256:676c7c10fd4d4cb2981b8fcb32fd764f5f661b04b7334d024034d07212f971fd", size = 332130, upload-time = "2025-11-25T04:31:54.212Z" }
wheels = [
{ url = "https://files.pythonhosted.org/packages/60/e2/60a20d04b0595c641516463168909c5bbcc192d3d6eacb637c1677109c6a/qdrant_client-1.16.1-py3-none-any.whl", hash = "sha256:1eefe89f66e8a468ba0de1680e28b441e69825cfb62e8fb2e457c15e24ce5e3b", size = 378481, upload-time = "2025-11-25T04:31:52.629Z" },
]
[[package]]
name = "questionary"