Skip to main content

Embedded RAG Knowledge Base

Reasoning now includes a built-in RAG (Retrieval-Augmented Generation) server that allows you to create and manage a custom knowledge base directly within the plugin, with no external installation required.

This significantly improves the quality and speed of code generation by providing project-specific context to the AI model.

Overview​

The RAG server is automatically initialized when Reasoning is activated and offers:

  • Local Storage: The knowledge base is saved in the plugin's data directory.
  • Semantic Search: Uses vector embeddings to find relevant knowledge.
  • Multiple Embedding Providers: OpenAI, Ollama, or local.
  • Automatic Persistence: Data is automatically saved in JSON format.

Why Use RAG?​

Traditional code generation often lacks project-specific context, resulting in generic and inconsistent code. RAG solves this by retrieving relevant information from your project and injecting it into the prompt, which leads to:

  • 🎯 Precision: Code that aligns with your project's standards.
  • ⚡ Speed: A 30-50% reduction in development time.
  • 🔄 Consistency: Code generation that follows established conventions.
  • 📚 Quality: A 40-60% improvement in the quality of the generated code.

Available Tools​

You can interact with the knowledge base using the following tools through the chat:

1. Add Knowledge​

{
  "tool": "add_rag_knowledge",
  "arguments": {
    "content": "React is a JavaScript library for building user interfaces",
    "title": "React - Introduction",
    "source": "react-documentation",
    "tags": ["react", "javascript", "frontend"]
  }
}

2. Search Knowledge​

{
  "tool": "search_rag_knowledge",
  "arguments": {
    "query": "React hooks useState",
    "maxResults": 5,
    "minScore": 0.8,
    "tags": ["react"]
  }
}

3. List Knowledge​

{
  "tool": "list_rag_knowledge",
  "arguments": {
    "limit": 20,
    "tags": ["javascript"]
  }
}

4. Update Knowledge​

{
  "tool": "update_rag_knowledge",
  "arguments": {
    "id": "abc123def456",
    "content": "Updated content",
    "title": "Updated Title",
    "tags": ["updated", "react"]
  }
}

5. Delete Knowledge​

{
  "tool": "delete_rag_knowledge",
  "arguments": {
    "id": "abc123def456"
  }
}

6. Get Stats​

{
  "tool": "get_rag_stats",
  "arguments": {}
}

Chat Usage Examples​

Add Documentation​

Add to the knowledge base: "TypeScript is a superset of JavaScript that adds static typing."

Search for Information​

Search the knowledge base for information about TypeScript

Configuration​

  1. Open VS Code settings.
  2. Search for "Reasoning RAG".
  3. Configure embedding providers and other options.
{
  "reasoning.rag.embeddingProvider": "local",
  "reasoning.rag.embeddingModel": "text-embedding-3-small",
  "reasoning.rag.maxResults": 10,
  "reasoning.rag.minScore": 0.7,
  "reasoning.rag.openaiApiKey": "your-openai-key",
  "reasoning.rag.ollamaUrl": "http://localhost:11434"
}

Use Cases​

  • Project Documentation: Store architectural decisions, code patterns, and specific documentation.
  • Code Snippets: Save code examples, solutions to common problems, and templates.
  • Domain Knowledge: Document business rules and information about APIs.
  • Troubleshooting: Keep a history of bugs, solutions, and workarounds.

For a detailed testing guide, refer to the RAG Knowledge Base Testing Guide.