Core - Configuration Guide

This YAML configuration file defines the structure and functionality of the AI stack, covering decision-making, available tools, and the vector database setup for Retrieval-Augmented Generation (RAG).

Configuration Overview

The configuration file can be placed in the root directory as stack.yaml or specified via an environment variable: STACK_CONFIG="/path/stack.yaml"

stack

The stack section specifies the core components, including the decision-maker and tools.

decision_maker

This component classifies tools for incoming queries.

• name: Decision-maker name.
• role: Describes its purpose, such as classifying tools for queries.
• kind: Specifies decision-making type:
  • prompt: Prompt-based classification.
  • rag: Not supported.
• ai_client: AI client configuration.
  • provider: AI provider (e.g., openai).
  • model: AI model (e.g., gpt-4).
  • api_key: API key for accessing the service.

tools

Defines the tools available to the stack for processing queries.

Each tool has the following properties:

• name: Tool name.
• role: Describes its purpose.
• kind: Tool type:
  • prompt: For prompt-based processing.
  • rag: For RAG-based document processing.
• depth: Number of document sets to consider for RAG-based tools.
• ai_client (if kind is prompt):
  • provider: AI provider (e.g., openai).
  • model: AI model (e.g., gpt-4).
  • api_key: API key for the service.
• embeddings (if kind is rag): Embedding model for document processing (e.g., text-embedding-ada-002).

Example Tools:

1. general_queries: Handles general queries using a prompt-based approach with gpt-4.
2. document_tool: Processes documents and generates answers using a RAG approach with text-embedding-ada-002 embeddings.

vector_db

Configures the vector database for RAG.

Properties:

• provider: Vector database provider (e.g., chroma).
• kind: Storage type:
  • in-memory: Temporary storage.
  • storage: Persistent local storage.
  • remote: Hosted remote database.
• host: Database host address (e.g., 0.0.0.0).
• port: Database port number (e.g., 8000).
• secure: Boolean indicating secure connection (true or false).

Example Configuration

stack:
  decision_maker:
    name: decision_maker
    role: classify tools for queries
    kind: prompt
    ai_client:
      provider: openai
      model: gpt-4
      api_key: <your-api-key>

  tools:
    - name: general_queries
      role: Handle other queries if no suitable tool is found.
      kind: prompt
      ai_client:
        provider: openai
        model: gpt-4
        api_key: <your-api-key>

    - name: document_tool
      role: Process documents and provide answers.
      kind: rag
      embeddings: text-embedding-ada-002
      ai_client:
        provider: openai
        model: gpt-4
        api_key: <your-api-key>
      depth: 4

  vector_db:
    provider: chroma
    kind: remote
    host: 0.0.0.0
    port: 8000
    secure: false