mcp_mirror_jeanibarz_knowledge_base_mcp_server
by MCP-MirrorKnowledge Base MCP Server
Overview
The Knowledge Base MCP Server is a tool designed to list and retrieve content from various knowledge bases. It leverages the Hugging Face Inference API for generating embeddings and uses FAISS for efficient similarity search. This server is particularly useful for managing and querying structured knowledge bases in a scalable manner.
Setup Instructions
Prerequisites
Steps
-
Clone the repository:
shell git clone <repository_url> cd knowledge-base-mcp-server -
Install dependencies:
shell npm install -
Configure environment variables:
HUGGINGFACE_API_KEY: API key for Hugging Face Inference API.KNOWLEDGE_BASES_ROOT_DIR: Directory for knowledge base subdirectories (defaults to$HOME/knowledge_bases).FAISS_INDEX_PATH: Path to the FAISS index (defaults to$HOME/knowledge_bases/.faiss).-
HUGGINGFACE_MODEL_NAME: Hugging Face model for embeddings (defaults tosentence-transformers/all-MiniLM-L6-v2). -
Build the server:
shell npm run build -
Add the server to the MCP settings:
-
Edit
cline_mcp_settings.jsonand add the following configuration:
json "knowledge-base-mcp": { "command": "node", "args": ["/path/to/knowledge-base-mcp-server/build/index.js"], "disabled": false, "autoApprove": [], "env": { "KNOWLEDGE_BASES_ROOT_DIR": "/path/to/knowledge_bases", "HUGGINGFACE_API_KEY": "YOUR_HUGGINGFACE_API_KEY" }, "description": "Retrieves similar chunks from the knowledge base based on a query." } -
Create knowledge base directories:
- Create subdirectories within
KNOWLEDGE_BASES_ROOT_DIRfor each knowledge base (e.g.,company,it_support,onboarding). - Place text files (e.g.,
.txt,.md) containing the knowledge base content within these subdirectories.
Usage
The server exposes two tools:
list_knowledge_bases: Lists the available knowledge bases.retrieve_knowledge: Retrieves similar chunks from the knowledge base based on a query.
Example Output of retrieve_knowledge:
## Semantic Search Results
**Result 1:**
[Content of the most similar chunk]
**Source:**
```json
{
"source": "[Path to the file containing the chunk]"
}
Result 2:
[Content of the second most similar chunk]
Source:
{
"source": "[Path to the file containing the chunk]"
}
Disclaimer: The provided results might not all be relevant. Please cross-check the relevance of the information.
```
About
This repository is a mirror of https://github.com/jeanibarz/knowledge-base-mcp-server.
Invite Gmail Contacts
