pengjunjie/ragflow
1
0
Fork 1
mirror of https://github.com/infiniflow/ragflow.git synced 2025-12-08 20:42:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

Docs: Initial editorial pass to MCP server (#7359)

Browse Source 
### What problem does this PR solve?


### Type of change

- [x] Documentation Update
This commit is contained in:
writinwaters
2025-05-07 19:40:45 +08:00
committed by GitHub
parent 9849230a04
commit 87317bcfc4
20 changed files with 238 additions and 254 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
docs/develop/acquire_ragflow_api_key.md
View File

@ -5,7 +5,7 @@ slug: /acquire_ragflow_api_key
# Acquire RAGFlow API key
A key is required for the RAGFlow server to authenticate your requests via HTTP or a Python API. This documents provides instructions on obtaining a RAGFlow API key.
An API key is required for the RAGFlow server to authenticate your HTTP/Python or MCP requests. This documents provides instructions on obtaining a RAGFlow API key.
1. Click your avatar in the top right corner of the RAGFlow UI to access the configuration page.
2. Click **API** to switch to the **API** page.

202
docs/develop/launch_mcp_server.md Normal file
View File

@ -0,0 +1,202 @@
---
sidebar_position: 4
slug: /launch_mcp_server
---
# Launch RAGFlow MCP server
Launch an MCP server from source or via Docker.
---
A RAGFlow Model Context Protocol (MCP) server is designed as an independent component to complement the RAGFlow server. Note that an MCP server must operate alongside a properly functioning RAGFlow server.
An MCP server can start up in either self-host mode (default) or host mode:
- **Self-host mode**:
When launching an MCP server in self-host mode, you must provide an API key to authenticate the MCP server with the RAGFlow server. In this mode, the MCP server can access *only* the datasets (knowledge bases) of a specified tenant on the RAGFlow server.
- **Host mode**:
In host mode, each MCP client can access their own knowledge bases on the RAGFlow server. However, each client request must include a valid API key to authenticate the client with the RAGFlow server.
Once a connection is established, an MCP server communicates with its client in MCP HTTP+SSE (Server-Sent Events) mode, unidirectionally pushing responses from the RAGFlow server to its client in real time.
## Prerequisites
1. Ensure RAGFlow is upgraded to v0.18.0 or later.
2. Have your RAGFlow API key ready. See [Acquire a RAGFlow API key](./acquire_ragflow_api_key.md).
:::tip INFO
If you wish to try out our MCP server without upgrading RAGFlow, community contributor **yiminghub2024** 👏 shares their recommended steps [here](#launch-an-mcp-server-without-upgrading-ragflow).
::
## Launch an MCP server
You can start an MCP server either from source code or via Docker.
### Launch from source code
1. Ensure that a RAGFlow server v0.18.0+ is properly running.
2. Launch the MCP server:
```bash
# Launch the MCP server to work in self-host mode, run either of the following
uv run mcp/server/server.py --host=127.0.0.1 --port=9382 --base_url=http://127.0.0.1:9380 --api_key=ragflow-xxxxx
# uv run mcp/server/server.py --host=127.0.0.1 --port=9382 --base_url=http://127.0.0.1:9380 mode=self-host --api_key=ragflow-xxxxx
# To launch the MCP server to work in host mode, run the following instead:
# uv run mcp/server/server.py --host=127.0.0.1 --port=9382 --base_url=http://127.0.0.1:9380 mode=host
```
Where:
- `host`: The MCP server's host address.
- `port`: The MCP server's listening port.
- `base_url`**: The address of the running RAGFlow server.
- `mode`: The launch mode.
- `self-host`: (default) self-host mode.
- `host`: host mode.
- `api_key`: Required in self-host mode to authenticate the MCP server with the RAGFlow server.
### Launch from Docker
#### 1. Enable MCP server
The MCP server is designed as an optional, component that complements the RAGFlow server and is disable by default. To enable MCP server:
1. Navigate to `docker/docker-compose.yml`.
2. Uncomment the `services.ragflow.command` section as shown below:
```yaml {6-13}
services:
ragflow:
...
image: ${RAGFLOW_IMAGE}
# Example configuration to set up an MCP server:
command:
- --enable-mcpserver
- --mcp-host=0.0.0.0
- --mcp-port=9382
- --mcp-base-url=http://127.0.0.1:9380
- --mcp-script-path=/ragflow/mcp/server/server.py
- --mcp-mode=self-host
- --mcp-host-api-key=ragflow-xxxxxxx
```
Where:
- `mcp-host`: The MCP server's host address.
- `mcp-port`: The MCP server's listening port.
- `mcp-base_url`: The address of the running RAGFlow server.
- `mcp-script-path`: The file path to the MCP server’s main script.
- `mcp-mode`: The launch mode.
- `self-host`: (default) self-host mode.
- `host`: host mode.
- `mcp-host-api_key`: Required in self-host mode to authenticate the MCP server with the RAGFlow server.
#### 2. Launch a RAGFlow server alongside an MCP server
Run `docker compose -f docker-compose.yml` to launch the RAGFlow server together with the MCP server.
*The following ASCII art confirms a successful launch:*
```bash
ragflow-server | Starting MCP Server on 0.0.0.0:9382 with base URL http://127.0.0.1:9380...
ragflow-server | Starting 1 task executor(s) on host 'dd0b5e07e76f'...
ragflow-server | 2025-04-18 15:41:18,816 INFO 27 ragflow_server log path: /ragflow/logs/ragflow_server.log, log levels: {'peewee': 'WARNING', 'pdfminer': 'WARNING', 'root': 'INFO'}
ragflow-server |
ragflow-server | __ __ ____ ____ ____ _____ ______ _______ ____
ragflow-server | | \/ |/ ___| _ \ / ___|| ____| _ \ \ / / ____| _ \
ragflow-server | | |\/| | | | |_) | \___ \| _| | |_) \ \ / /| _| | |_) |
ragflow-server | | | | | |___| __/ ___) | |___| _ < \ V / | |___| _ <
ragflow-server | |_| |_|\____|_| |____/|_____|_| \_\ \_/ |_____|_| \_\
ragflow-server |
ragflow-server | MCP launch mode: self-host
ragflow-server | MCP host: 0.0.0.0
ragflow-server | MCP port: 9382
ragflow-server | MCP base_url: http://127.0.0.1:9380
ragflow-server | INFO: Started server process [26]
ragflow-server | INFO: Waiting for application startup.
ragflow-server | INFO: Application startup complete.
ragflow-server | INFO: Uvicorn running on http://0.0.0.0:9382 (Press CTRL+C to quit)
ragflow-server | 2025-04-18 15:41:20,469 INFO 27 found 0 gpus
ragflow-server | 2025-04-18 15:41:23,263 INFO 27 init database on cluster mode successfully
ragflow-server | 2025-04-18 15:41:25,318 INFO 27 load_model /ragflow/rag/res/deepdoc/det.onnx uses CPU
ragflow-server | 2025-04-18 15:41:25,367 INFO 27 load_model /ragflow/rag/res/deepdoc/rec.onnx uses CPU
ragflow-server | ____ ___ ______ ______ __
ragflow-server | / __ \ / | / ____// ____// /____ _ __
ragflow-server | / /_/ // /| | / / __ / /_ / // __ \| | /| / /
ragflow-server | / _, _// ___ |/ /_/ // __/ / // /_/ /| |/ |/ /
ragflow-server | /_/ |_|/_/ |_|\____//_/ /_/ \____/ |__/|__/
ragflow-server |
ragflow-server |
ragflow-server | 2025-04-18 15:41:29,088 INFO 27 RAGFlow version: v0.18.0-285-gb2c299fa full
ragflow-server | 2025-04-18 15:41:29,088 INFO 27 project base: /ragflow
ragflow-server | 2025-04-18 15:41:29,088 INFO 27 Current configs, from /ragflow/conf/service_conf.yaml:
ragflow-server | ragflow: {'host': '0.0.0.0', 'http_port': 9380}
...
ragflow-server | * Running on all addresses (0.0.0.0)
ragflow-server | * Running on http://127.0.0.1:9380
ragflow-server | * Running on http://172.19.0.6:9380
ragflow-server | ______ __ ______ __
ragflow-server | /_ __/___ ______/ /__ / ____/ _____ _______ __/ /_____ _____
ragflow-server | / / / __ `/ ___/ //_/ / __/ | |/_/ _ \/ ___/ / / / __/ __ \/ ___/
ragflow-server | / / / /_/ (__ ) ,< / /____> </ __/ /__/ /_/ / /_/ /_/ / /
ragflow-server | /_/ \__,_/____/_/|_| /_____/_/|_|\___/\___/\__,_/\__/\____/_/
ragflow-server |
ragflow-server | 2025-04-18 15:41:34,501 INFO 32 TaskExecutor: RAGFlow version: v0.18.0-285-gb2c299fa full
ragflow-server | 2025-04-18 15:41:34,501 INFO 32 Use Elasticsearch http://es01:9200 as the doc engine.
...
```
#### Launch an MCP server without upgrading RAGFlow
:::tip KUDOS-TO
This section is contributed by our community contributor **yiminghub2024**. 👏
:::
1. Prepare all MCP-specific files and directories.
i. Copy the [mcp/](https://github.com/infiniflow/ragflow/tree/main/mcp) directory to your local working directory.
ii. Copy [docker/docker-compose.yml](https://github.com/infiniflow/ragflow/blob/main/docker/docker-compose.yml) locally.
iii. Copy [docker/entrypoint.sh](https://github.com/infiniflow/ragflow/blob/main/docker/entrypoint.sh) locally.
iv. Install required dependencies using `uv`:
- Run `uv add mcp` or
- Copy pyproject.toml and run `uv sync --python 3.10 --all-extras`.
2. Edit **docker-compose.yml** to enable MCP (disable by default).
3. Launch the MCP server:
```bash
docker compose -f docker-compose.yml up -d`
```
### Check MCP server status
Run the following to check the logs the RAGFlow server and the MCP server:
```bash
docker logs ragflow-server
```
## MCP client example
We provide a *prototype* MCP client example for testing [here](https://github.com/infiniflow/ragflow/blob/main/mcp/client/client.py).
:::danger IMPORTANT
If your MCP server is running in host mode, include your acquired API key in your client's `headers` as shown below:
```python
async with sse_client("http://localhost:9382/sse", headers={"api_key": "YOUR_KEY_HERE"}) as streams:
# Rest of your code...
```
:::
## API
The MCP server currently offers an API to assist users in searching for relevant information powered by RAGFlow DeepDoc technology:
- **retrieve**: Fetches relevant chunks from specified `dataset_ids` and optional `document_ids` using the RAGFlow retrieve interface, based on a given question. Details of all available datasets, namely, `id` and `description`, are provided within the tool description for each individual dataset.
## Security considerations
As MCP technology is still at early stage and no official best practices for authentication or authorization have been established, RAGFlow currently uses [API key](./acquire_ragflow_api_key.md) to validate identity for the operations described earlier. However, in public environments, this makeshift solution could expose your MCP server to potential network attacks. Therefore, when running a local SSE server, it is recommended to bind only to localhost (`127.0.0.1`) rather than to all interfaces (`0.0.0.0`).
For further guidance, see the [official MCP documentation](https://modelcontextprotocol.io/docs/concepts/transports#security-considerations).

230
docs/develop/mcp.md
View File

@ -1,230 +0,0 @@
---
sidebar_position: 4
slug: /mcp_server
---
# RAGFlow MCP server overview
The RAGFlow Model Context Protocol (MCP) server operates as an independent component that complements the RAGFlow server. However, it requires a RAGFlow server to work functionally well, meaning, the MCP client and server communicate with each other in MCP HTTP+SSE mode (once the connection is established, server pushes messages to client only), and responses are expected from RAGFlow server.
The MCP server currently offers a specific tool to assist users in searching for relevant information powered by RAGFlow DeepDoc technology:
- **retrieve**: Fetches relevant chunks from specified `dataset_ids` and optional `document_ids` using the RAGFlow retrieve interface, based on a given question. Details of all available datasets, namely, `id` and `description`, are provided within the tool description for each individual dataset.
## Launching the MCP Server
Similar to launching the RAGFlow server, the MCP server can be started either from source code or via Docker.
### Launch Modes
The MCP server supports two launch modes:
1. **Self-Host Mode**:
- In this mode, the MCP server is launched to access a specific tenant's datasets.
- This is the default mode.
- The `--api_key` argument is **required** to authenticate the server with the RAGFlow server.
- Example:
```bash
uv run mcp/server/server.py --host=127.0.0.1 --port=9382 --base_url=http://127.0.0.1:9380 --mode=self-host --api_key=ragflow-xxxxx
```
1. **Host Mode**:
- In this mode, the MCP server allows each user to access their own datasets.
- To ensure secure access, a valid API key must be included in the request headers to identify the user.
- The `--api_key` argument is **not required** during server launch but must be provided in the headers on each client request for user authentication.
- Example:
```bash
uv run mcp/server/server.py --host=127.0.0.1 --port=9382 --base_url=http://127.0.0.1:9380 --mode=host
```
### Launching from Source Code
All you need to do is stand on the right place and strike out command, assuming you are on the project working directory.
```bash
uv run mcp/server/server.py --host=127.0.0.1 --port=9382 --base_url=http://127.0.0.1:9380 --api_key=ragflow-xxxxx
```
For testing purposes, there is an [MCP client example](#example_mcp_client) provided, free to take!
#### Required Arguments
- **`host`**: Specifies the server's host address.
- **`port`**: Defines the server's listening port.
- **`base_url`**: The address of the RAGFlow server that is already running and ready to handle tasks.
- **`mode`**: Launch mode, only accept `self-host` or `host`.
- **`api_key`**: Required when `mode` is `self-host` to authenticate the MCP server with the RAGFlow server.
Here are three augments required, the first two,`host` and `port`, are self-explained. The`base_url` is the address of the ready-to-serve RAGFlow server to actually perform the task.
### Launching from Docker
Building a standalone MCP server image is straightforward and easy, so we just proposed a way to launch it with RAGFlow server here.
#### Alongside RAGFlow {#alongside_ragflow}
As MCP server is an extra and optional component of RAGFlow server, we consume that not everybody going to use it. Thus, it is disable by default.
To enable it, simply find `docker/docker-compose.yml` to uncomment `services.ragflow.command` section.
```yaml
services:
ragflow:
...
image: ${RAGFLOW_IMAGE}
# example to setup MCP server
command:
- --enable-mcpserver
- --mcp-host=0.0.0.0
- --mcp-port=9382
- --mcp-base-url=http://127.0.0.1:9380
- --mcp-script-path=/ragflow/mcp/server/server.py
- --mcp-mode=self-host # `self-host` or `host`
- --mcp-host-api-key=ragflow-xxxxxxx # only need to privide when mode is `self-host` and use bare string without quotation marks here.
```
To troubleshoot, launch the service in the foreground using `docker compose -f docker-compose.yml`.
### For those upgrading from versions before v0.18.0
1. Get all MCP related files ready.
1. copy `mcp/` directory to local.
1. copy `docker/docker-compose.yml` to local.
1. copy `docker/entrypoint.sh` to local.
1. resolve necessary dependencies via `uv`.
- simply run `uv add mcp` if it works for you. Or:
- copy `pyproject.toml` and run `uv sync --python 3.10 --all-extras`.
1. Change `docker-compose.yml` to enable MCP as it is disable by default, [see last section](#alongside_ragflow).
1. Launch the service with `docker compose -f docker-compose.yml up -d`
### Check the MCP server status
Checking logs of RAGFlow server with `docker logs ragflow-server`. If you see the MCP server ASCII art there, it means all is OK!
```bash
ragflow-server | Starting MCP Server on 0.0.0.0:9382 with base URL http://127.0.0.1:9380...
ragflow-server | Starting 1 task executor(s) on host 'dd0b5e07e76f'...
ragflow-server | 2025-04-18 15:41:18,816 INFO 27 ragflow_server log path: /ragflow/logs/ragflow_server.log, log levels: {'peewee': 'WARNING', 'pdfminer': 'WARNING', 'root': 'INFO'}
ragflow-server |
ragflow-server | __ __ ____ ____ ____ _____ ______ _______ ____
ragflow-server | | \/ |/ ___| _ \ / ___|| ____| _ \ \ / / ____| _ \
ragflow-server | | |\/| | | | |_) | \___ \| _| | |_) \ \ / /| _| | |_) |
ragflow-server | | | | | |___| __/ ___) | |___| _ < \ V / | |___| _ <
ragflow-server | |_| |_|\____|_| |____/|_____|_| \_\ \_/ |_____|_| \_\
ragflow-server |
ragflow-server | MCP launch mode: self-host
ragflow-server | MCP host: 0.0.0.0
ragflow-server | MCP port: 9382
ragflow-server | MCP base_url: http://127.0.0.1:9380
ragflow-server | INFO: Started server process [26]
ragflow-server | INFO: Waiting for application startup.
ragflow-server | INFO: Application startup complete.
ragflow-server | INFO: Uvicorn running on http://0.0.0.0:9382 (Press CTRL+C to quit)
ragflow-server | 2025-04-18 15:41:20,469 INFO 27 found 0 gpus
ragflow-server | 2025-04-18 15:41:23,263 INFO 27 init database on cluster mode successfully
ragflow-server | 2025-04-18 15:41:25,318 INFO 27 load_model /ragflow/rag/res/deepdoc/det.onnx uses CPU
ragflow-server | 2025-04-18 15:41:25,367 INFO 27 load_model /ragflow/rag/res/deepdoc/rec.onnx uses CPU
ragflow-server | ____ ___ ______ ______ __
ragflow-server | / __ \ / | / ____// ____// /____ _ __
ragflow-server | / /_/ // /| | / / __ / /_ / // __ \| | /| / /
ragflow-server | / _, _// ___ |/ /_/ // __/ / // /_/ /| |/ |/ /
ragflow-server | /_/ |_|/_/ |_|\____//_/ /_/ \____/ |__/|__/
ragflow-server |
ragflow-server |
ragflow-server | 2025-04-18 15:41:29,088 INFO 27 RAGFlow version: v0.18.0-285-gb2c299fa full
ragflow-server | 2025-04-18 15:41:29,088 INFO 27 project base: /ragflow
ragflow-server | 2025-04-18 15:41:29,088 INFO 27 Current configs, from /ragflow/conf/service_conf.yaml:
ragflow-server | ragflow: {'host': '0.0.0.0', 'http_port': 9380}
...
ragflow-server | * Running on all addresses (0.0.0.0)
ragflow-server | * Running on http://127.0.0.1:9380
ragflow-server | * Running on http://172.19.0.6:9380
ragflow-server | ______ __ ______ __
ragflow-server | /_ __/___ ______/ /__ / ____/ _____ _______ __/ /_____ _____
ragflow-server | / / / __ `/ ___/ //_/ / __/ | |/_/ _ \/ ___/ / / / __/ __ \/ ___/
ragflow-server | / / / /_/ (__ ) ,< / /____> </ __/ /__/ /_/ / /_/ /_/ / /
ragflow-server | /_/ \__,_/____/_/|_| /_____/_/|_|\___/\___/\__,_/\__/\____/_/
ragflow-server |
ragflow-server | 2025-04-18 15:41:34,501 INFO 32 TaskExecutor: RAGFlow version: v0.18.0-285-gb2c299fa full
ragflow-server | 2025-04-18 15:41:34,501 INFO 32 Use Elasticsearch http://es01:9200 as the doc engine.
...
```
You are ready to brew🍺!
#### Getting API Keys for Host Mode
When running the MCP server in `host` mode (by setting `--mcp-mode=host` in the configuration), each client needs to provide their own API key in requests. This API key is **different** from the `--mcp-host-api-key` specified in the server configuration.
To get a valid API key for use in your client scripts (like `test_mcp.py`):
1. Access the RAGFlow UI in your browser (typically `http://localhost:9380`)
2. Log in to your account
3. Click on your avatar/profile in the top-right corner
4. Select **API** from the dropdown menu
5. On the API page, generate a new API key or copy an existing one
6. Use this key in your client script as follows:
```python
# Client script example (test_mcp.py)
async with sse_client("http://localhost:9382/sse", headers={"api_key": "YOUR_KEY_HERE"}) as streams:
# Rest of your code...
```
## Testing and Usage
Typically, there are various ways to utilize an MCP server. You can integrate it with LLMs or use it as a standalone tool. You find the way.
### Example MCP Client {#example_mcp_client}
```python
#
# Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
from mcp.client.session import ClientSession
from mcp.client.sse import sse_client
async def main():
try:
# To access RAGFlow server in `host` mode, you need to attach `api_key` for each request to indicate identification.
# async with sse_client("http://localhost:9382/sse", headers={"api_key": "ragflow-IyMGI1ZDhjMTA2ZTExZjBiYTMyMGQ4Zm"}) as streams:
async with sse_client("http://localhost:9382/sse") as streams:
async with ClientSession(
streams[0],
streams[1],
) as session:
await session.initialize()
tools = await session.list_tools()
print(f"{tools.tools=}")
response = await session.call_tool(name="ragflow_retrieval", arguments={"dataset_ids": ["ce3bb17cf27a11efa69751e139332ced"], "document_ids": [], "question": "How to install neovim?"})
print(f"Tool response: {response.model_dump()}")
except Exception as e:
print(e)
if __name__ == "__main__":
from anyio import run
run(main)
```
## Security and Concerns
Since MCP technology is still in booming age and there are still no official Authentication and Authorization best practices to follow, RAGFlow uses `api_key` to validate the identification, and it is required to perform any operations mentioned in the preview section. Obviously, this is not a premium solution to do so, thus this RAGFlow MCP server is not expected to exposed to public use as it could be highly venerable to be attacked. For local SSE server, bind only to localhost (127.0.0.1) instead of all interfaces (0.0.0.0). For additional guidance, you can refer to [MCP official website](https://modelcontextprotocol.io/docs/concepts/transports#security-considerations).