local-context-bridge
v1.0.7Search local documents (PDF, Word, Excel, Markdown) using semantic search.
Installation
ContextBridge Knowledge Base
ContextBridge is a local AI agent's memory bridge. It provides instant access to your documents (PDF, Office, Markdown) for AI agents like OpenClaw, Claude Desktop, and Cursor.
📚 HOW IT WORKS
ContextBridge acts as a "Semantic Bridge" between your files and OpenClaw:
- Local Indexing: Scans documents and splits them into text chunks.
- Vectorization: Converts text into semantic vectors using a local AI embedding model.
- Semantic Search: Finds the most relevant chunks from your knowledge base.
- 100% Private: All data stays on your local machine.
Key Features: - 🔌 MCP and API Ready: Native support for Model Context Protocol - 📄 Multi-format Parsing: PDF, Word, Excel, PPTX, Markdown - 🔋 Batteries Included: Embedded vector database - 👁️ Auto Sync: Automatic file monitoring and index rebuilding
🚀 INSTALLATION
1. Install ContextBridge
pip install cbridge-agent
2. Initialize Workspace
cbridge init
Creates workspace directories and generates config file (~/.cbridge/config.yaml).
3. Add Documents
cbridge watch add /path/to/your/documents
cbridge watch list # View watched folders
4. Start Service
cbridge serve # API only (http://127.0.0.1:9790)
Configuration
Edit ~/.cbridge/config.yaml:
workspace_dir: ~/.cbridge/workspace
watch_dirs:
- /path/to/documents
exclude_patterns:
- "**/node_modules/**"
- "**/.git/**"
language: en
mode: embedded
Performance Optimization (for low-end devices):
export CB_MAX_CONCURRENT_FILES=2
export CB_EMBEDDING_BATCH_SIZE=4
export CB_DISABLE_WATCHER=true
🔧 API USAGE
ContextBridge provides a REST API at http://127.0.0.1:9790.
1. Semantic Search
{
"command": "curl -s -X POST http://127.0.0.1:9790/api/v1/search -d '{"query": "project budget 2024", "top_k": 5}'"
}
Response:
{
"results": [
{
"content": "The 2024 Q1 budget is $500,000...",
"metadata": {"source": "budget.xlsx"},
"score": 0.89
}
]
}
2. Check Status
{
"command": "curl -s http://127.0.0.1:9790/api/v1/watch/status"
}
3. Add/Remove Watch Directory
// Add
{
"command": "curl -s -X POST http://127.0.0.1:9790/api/v1/watch/directories -d '{"path": "/path/to/folder"}'"
}
// Remove
{
"command": "curl -s -X DELETE http://127.0.0.1:9790/api/v1/watch/directories -d '{"path": "/path/to/folder"}'"
}
4. Manual Re-index
{
"command": "curl -s -X POST http://127.0.0.1:9790/api/v1/index/sync"
}
5. Health Check
{
"command": "curl -s http://127.0.0.1:9790/api/v1/health"
}
🎯 WHEN TO USE THIS SKILL
DIRECTIVE: If the user's request involves information NOT in general public knowledge, MUST query ContextBridge before answering.
1. Deep Fact Retrieval (Private Knowledge)
Find specific numbers, names, or facts from local files.
- Examples: "What was the 2024 marketing budget?", "Who was the lead engineer for Project X?"
- Query: "2024 marketing budget" ✅
2. Policy-Based Auditing & Compliance
Review documents against standards.
- Examples: "Does this contract follow our procurement policy?", "Review this PR based on coding standards."
- Strategy: First search "coding standards", then search document content.
3. Contextual Drafting (Style & Templates)
Write content matching previous style or templates.
- Examples: "Draft a project proposal using our template.", "Write a follow-up email in the same tone."
- Query: "project proposal template" or "client communication style"
4. Technical Support & Internal Processes
Answer "How-to" questions about internal tools or processes.
- Examples: "How do I set up the dev environment?", "What's the onboarding process?"
- Query: "development environment setup" or "onboarding process steps"
5. Codebase Understanding
Navigate and understand your own codebase.
- Examples: "Where is authentication implemented?", "How does payment processing work?"
- Query: "authentication implementation" or "payment processing flow"
💡 SEARCH BEST PRACTICES
Keyword Extraction
- DO: Extract core entities
"2024 marketing budget"✅- DON'T: Use full sentences
"What was the budget for 2024 marketing?"❌
Iterative Searching
- Start with specific keywords
- If no results, broaden query
- Try synonyms or related terms
Multiple Queries
For complex tasks, execute multiple searches:
{
"command": "curl -s -X POST http://127.0.0.1:9790/api/v1/search -d '{"query": "coding standards python", "top_k": 3}'"
}
Citation Requirement
Always cite sources:
- "According to budget.xlsx..."
- "As documented in employee_handbook.pdf..."
📖 CLI COMMANDS
# Initialization
cbridge init # Setup workspace
cbridge lang en # Switch language
# Document Management
cbridge watch add <path> # Add folder
cbridge watch remove <path> # Remove folder
cbridge watch list # List folders
cbridge index # Manual re-index
# Service Control
cbridge start # Start service
cbridge serve # API only
cbridge stop # Stop
cbridge status # Check status
cbridge logs # View logs
# Search
cbridge search <query> # Search documents
🔍 TROUBLESHOOTING
API Connection Failed
cbridge status # Check status
cbridge restart # Restart service
curl -s http://127.0.0.1:9790/api/v1/health # Health check
Cannot Find Latest Content
cbridge watch list # Check watched folders
cbridge index # Force re-index
Curl Command Fails
Ensure curl is installed:
curl --version # Check version
brew install curl # macOS
sudo apt install curl # Linux
📝 WORKFLOW EXAMPLES
Example 1: Find Project Budget
User asks: "What was the total budget for Project Alpha?"
Agent actions:
1. Extract keywords: "Project Alpha budget"
2. Execute search:
{
"command": "curl -s -X POST http://127.0.0.1:9790/api/v1/search -d '{"query": "Project Alpha budget", "top_k": 3}'"
}
- Cite source: "According to
project_alpha_proposal.pdf, the total budget was $150,000."
Example 2: Code Review
User asks: "Review this Python function for compliance with our coding standards."
Agent actions:
1. Search for standards: "Python coding standards"
2. Search for examples: "Python best practices"
3. Compare and provide feedback with citations.
📚 RESOURCES
- GitHub: whyischen/context-bridge
- API Docs:
http://127.0.0.1:9790/docs(when running) - Config:
~/.cbridge/config.yaml - Workspace:
~/.cbridge/workspace/