Atlas Components

Atlas is built using a modular component architecture that enables flexible configuration and extension. This document provides an overview of the major components and their relationships.

Component Diagram

The following diagram shows the main components of Atlas and how they interact:

Component Relationships

Agent Components

The agent components form the primary interface with users and orchestrate the entire system:

• AtlasAgent: Base agent class that provides core functionality

  • Manages conversation history
  • Interacts with language model providers
  • Retrieves knowledge from the database
  • Formats prompts and processes responses

• ControllerAgent: Extends AtlasAgent for multi-agent orchestration

  • Manages worker agents
  • Coordinates parallel tasks
  • Aggregates results from workers
  • Implements complex workflows

• WorkerAgent: Specialized agents for specific tasks

  • AnalysisWorker: Analyzes queries and identifies information needs
  • RetrievalWorker: Focuses on efficient document retrieval
  • DraftWorker: Specializes in generating draft responses

Knowledge Components

The knowledge components manage document storage and retrieval:

• KnowledgeBase: Main interface for knowledge retrieval

  • Manages ChromaDB connection
  • Provides document search functionality
  • Handles relevance scoring

• DocumentProcessor: Handles document ingestion

  • Splits documents into appropriate chunks
  • Manages metadata and source tracking
  • Creates embeddings for documents

Provider Components

The provider components abstract interactions with language model APIs:

• Provider: Base interface for all providers

  • AnthropicProvider: Implementation for Anthropic Claude models
  • OpenAIProvider: Implementation for OpenAI GPT models
  • OllamaProvider: Implementation for local Ollama models

• ProviderFactory: Creates appropriate provider instances

  • Handles provider detection and selection
  • Manages provider registration
  • Implements fallback logic

Graph Components

The graph components enable complex workflow definition:

• StateGraph: LangGraph integration for state management

  • Defines workflow steps and transitions
  • Manages state during workflow execution

• Edge: Defines transitions between graph nodes

  • Implements conditional routing
  • Provides workflow branching logic

• Node: Defines individual operations in the workflow

  • Implements state transformations
  • Encapsulates agent and system functions

Core Components

The core components provide foundational capabilities:

• Config: Configuration management

  • Environment variable integration
  • Default value handling
  • Configuration validation

• Telemetry: Performance monitoring

  • Operation tracing
  • Error tracking
  • Performance metrics

• Error: Standardized error handling

  • Custom error types
  • Error recovery mechanisms
  • Graceful degradation

Component Dependencies

The following table shows the key dependencies between components:

Component	Dependencies
AtlasAgent	Provider, KnowledgeBase, Config
ControllerAgent	AtlasAgent, WorkerAgent, StateGraph
WorkerAgent	AtlasAgent, KnowledgeBase
KnowledgeBase	ChromaDB, Config
Provider	Config, Telemetry
StateGraph	Node, Edge, Config

Next Steps

• See Agents for detailed agent documentation
• See Knowledge for knowledge system details
• See Providers for provider information
• See Graph for workflow documentation