Fetch Bonfire - uAgent Integration

This is an agent integration between bonfires.ai and ASI:One uAgent framework.

Bonfires provides a contextual framework to process unstructured data while the uAgents bring infrastructure for distribution of knowledge and agentic frameworks.

Bonfires is a semantic back end that creates taxonomies for documents and stores them in vector storage. The service can ingest documents and use chunk search to find contextual data for the agent.

Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Run the Bonfire Agent

python agent.py

The agent will start and display its address. Note this address for the client configuration.

3. Update Client Configuration

Edit client.py and update the BONFIRE_AGENT_ADDRESS variable with the address from step 2.

4. Run the Test Client

In a new terminal:

python client.py

Architecture

System Overview

graph TB
    subgraph "uAgent Framework"
        UA[User Agent] --> |Chat Messages| BA[Bonfire Agent]
        BA --> |Responses| UA
        BA --> |Search Queries| UA
    end

    subgraph "Bonfires.ai Backend"
        BA --> |Ingest Content| IC[Ingest Content API]
        BA --> |Search| VS[Vector Store Search]
        BA --> |Taxonomy| TG[Taxonomy Generation]
        
        IC --> |Store| VDB[(Vector Database<br/>Weaviate)]
        VS --> |Query| VDB
        TG --> |Labels| VDB
        
        VDB --> |Chunks| CH[Content Chunks]
        VDB --> |Embeddings| EM[Vector Embeddings]
        VDB --> |Taxonomy| TX[Taxonomy Labels]
    end

    subgraph "Memory Flow"
        UA --> |Chat Output| BA
        BA --> |Store in Bonfire| IC
        IC --> |Persistent Memory| VDB
        VDB --> |Context Retrieval| VS
        VS --> |Relevant Context| BA
        BA --> |Enhanced Response| UA
    end

    subgraph "Query Flow"
        UA --> |Question| BA
        BA --> |Search Request| VS
        VS --> |Semantic Search| VDB
        VDB --> |Relevant Chunks| VS
        VS --> |Search Results| BA
        BA --> |Answer| UA
    end

    style BA fill:#e1f5fe
    style VDB fill:#f3e5f5
    style UA fill:#e8f5e8
    style IC fill:#fff3e0
    style VS fill:#fff3e0
    style TG fill:#fff3e0

Loading

Data Flow

1. Ingest: Content is ingested into a bonfire via /ingest_content
2. Process: Taxonomy is generated and chunks are labeled
3. Store: Content is stored in vector database (Weaviate)
4. Query: Semantic search can be performed on stored content
5. Retrieve: Relevant chunks can be retrieved for agent responses

Integration Points

• Memory Storage: Use bonfire as persistent memory store for agent conversations
• Context Retrieval: Use vector search to find relevant context for agent responses
• Taxonomy Learning: Leverage taxonomy generation for better content organization
• Job Monitoring: Track long-running operations through job system

API Documentation

See bonfires_api_notes.md for detailed API documentation and investigation notes.

TODO

1. ✅ Map bonfires API calls and data structure
2. ✅ Create an architecture diagram for the integration:
   • ✅ Create data store on bonfires
   • ✅ Create query flow: infer, search, reply
   • ✅ Create memory flow: chat outputs into bonfire, (mailbox processing)
   • Deploy uAgent

Files

• agent.py - Main uAgent that integrates with bonfires.ai
• client.py - Test client for the uAgent
• requirements.txt - Python dependencies
• bonfires_api_notes.md - API investigation notes
• architecture.md - Detailed architecture documentation
• README.md - This file