How andrejsstepanovs/bobik works
Main Components
a) Core Application Layer:
• Entry Point (run.py)
• Application Manager (src/manager.py)
• State Management (src/state.py)
• Configuration Handler (src/config.py)
b) I/O Layer:
• Input Handlers (src/io_input.py)
• Text Input
• Voice Input (Deepgram)
• Clipboard Input
• File Input
• Output Handlers (src/io_output.py)
• Text Output
• Voice Output (ffplay)
• File Output (History)
c) LLM Integration Layer:
• LLM Provider Interface (src/llm_provider.py)
• LLM Agent Handler (src/llm_agent.py)
• Supported Providers:
• OpenAI
• Groq
• Mistral
• Ollama
• RunPod
• LM Studio
d) Tools Layer:
• Tool Loader (src/tool_loader.py)
• Built-in Tools:
• Weather
• Calendar
• News
• DateTime
• State Management
e) Support Systems:
• History Management (src/history.py)
• Enrichers (src/enrichers.py)
• Parsers (src/parsers.py)
[User Interface Layer]
• ↕
[I/O Handler Layer]
• ↕
[Core Application Layer]
• ↕
[LLM Integration Layer] ←→ [External LLM Providers]
• ↕
[Tools Layer] ←→ [External Services]
Flow Representation
• User input flow through the system
• Configuration loading and distribution
• State changes and management
• Tool execution paths
• LLM provider interactions
Detailed Component Interactions
• How the pre-parser interacts with input processing
• How the LLM agent coordinates with tools
• How state changes propagate through the system
• How history management integrates with the core flow
Notes Section
Annotations for:
• Platform-specific considerations (Windows/Linux/Mac)
• Configuration requirements
• External dependencies
• Key feature toggles
Remember to emphasize the dynamic nature of the system, particularly how it handles runtime state changes and model switching, as these are key features of the application.