KB-Bridge
A Model Context Protocol (MCP) server for intelligent knowledge base search and retrieval with support for multiple backend providers.
接入你的智能体
从发布者处安装。不是 PlugMCP 端点。
远程 URL 还是本机命令? 两种安装方式的区别
pip install kbbridge
Create a .env file with your retrieval backend credentials:
# Required - Retrieval Backend Configuration
RETRIEVAL_ENDPOINT=https://api.dify.ai/v1 # Example: Dify endpoint
RETRIEVAL_API_KEY=your-retrieval-api-key
LLM_API_URL=https://your-llm-service.com/v1
LLM_MODEL=gpt-4o
LLM_API_TOKEN=your-token-here
# Optional
RERANK_URL=https://your-rerank-api.com
RERANK_MODEL=your-rerank-model
Supported Backends:
See env.example for all available configuration options.
# Start server
python -m kbbridge.server --host 0.0.0.0 --port 5210
# Or using Makefile (if available)
make start
Server runs on http://0.0.0.0:5210 with MCP endpoint at http://0.0.0.0:5210/mcp.
For local development or simple single-container deployments:
# Build the image
docker build -t kbbridge:latest .
# Run with environment variables
docker run -d \
--name kbbridge \
-p 5210:5210 \
--env-file .env \
kbbridge:latest
For production deployments, use container orchestration platforms like Kubernetes with your preferred deployment method.
KB-Bridge follows a multi-stage pipeline to ensure high-quality answers:
flowchart LR
Start([User Query]) --> Preprocess[Query Preprocessing<br/>Rewriting & Understanding]
Preprocess --> FileDiscovery[File Discovery<br/>Find Relevant Files]
FileDiscovery --> Search[Search Stages]
Search --> Direct[Direct Approach<br/>Simple Retrieval]
Search --> Advanced[Advanced Approach<br/>File-level Processing]
Direct --> Candidates
Advanced --> Candidates[Collect Candidates]
Candidates --> Synthesis[Answer Synthesis<br/>Rerank & Format]
Synthesis --> Reflection{Reflection<br/>Enabled?}
Reflection -->|Yes| Reflect[Quality Check<br/>& Refinement]
Reflection -->|No| Final
Reflect --> Final([Final Answer])
style Start fill:#e1f5ff
style Final fill:#c8e6c9
style FileDiscovery fill:#fff9c4
style Direct fill:#fff9c4
style Advanced fill:#fff9c4
style Reflect fill:#ffccbc
style Synthesis fill:#e1bee7
Query Preprocessing (Optional)
File Discovery
Search Stages (Parallel)
Answer Synthesis
Quality Reflection (Optional)
The orchestrator (DatasetProcessor) currently implements stages 1-3, 5-8. File Discovery Quality Evaluation (stage 4) is implemented but not yet integrated into the pipeline. See .doc/FILE_DISCOVERY_EVALUATION_CONFIG.md for details.
assistant: Intelligent search and answer extraction from knowledge basesfile_discover: Discover relevant files using retriever + optional rerankingfile_lister: List files in knowledge base datasetskeyword_generator: Generate search keywords using LLMretriever: Retrieve information using various search methodsfile_count: Get file count in knowledge base datasetimport asyncio
from fastmcp import Client
async def main():
async with Client("http://localhost:5210/mcp") as client:
result = await client.call_tool(
"assistant",
{
"resource_id": "resource-id",
"query": "What are the safety protocols?",
},
)
print(result.content[0].text)
asyncio.run(main())
await client.call_tool("assistant", {
"resource_id": "hr_dataset",
"query": "What is the maternity leave policy?",
"custom_instructions": "Focus on HR compliance and legal requirements."
})
await client.call_tool("assistant", {
"resource_id": "resource-id",
"query": "What are the safety protocols?",
"enable_query_rewriting": True # Enables LLM-based query expansion/relaxation
})
await client.call_tool("assistant", {
"resource_id": "resource-id",
"query": "What are the safety protocols?",
"document_name": "safety_manual.pdf" # Limit search to specific document
})
You can plug KB-Bridge into a Dify Agent Workflow instead of calling MCP tools directly:
http://localhost:5210/mcpX-RETRIEVAL-ENDPOINT, X-RETRIEVAL-API-KEY, X-LLM-API-URL, X-LLM-MODELassistantresource_id, query, and other tool parametersassistant tool → Structured answer with citations# Install development dependencies
pip install -e ".[dev]"
# Run tests
pytest tests/
# Format code
black kbbridge/ tests/
# Lint code
ruff check kbbridge/ tests/
Apache-2.0
MCP server for enhancing knowledge base search and retrieval
工具
暂无工具列表。远程服务器在首次健康检查后会获得工具列表。
常见问题
在本页复制适用于 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 配置。它使用发布者的安装方式,而不是 PlugMCP 代理。
此记录未列出必需的环境变量。
Kb Bridge 通过 stdio 在本地运行。
上次检查时间:2026-10-28。 MCP server for enhancing knowledge base search and retrieval