How integuru-ai/integuru works
Main Components
a) Input Processing Layer:
• create_har.py (Browser Network Request Capture)
• Cookie Storage
• User Prompt Interface
b) Core Processing Components:
• Graph Builder Module
• LLM Integration Layer
• DAG Management System
• Agent System
c) Code Generation Layer:
• Integration Code Generator
• Output Handler
Component Relationships and Data Flow
[Input Layer]
• ↓
[Processing Layer]
• ↓
[Output Layer]
Specific Component Details
a) Input Components (Rectangle, Blue):
• HAR File Processor
• Cookie File Processor
• Prompt Handler
b) Core Processing (Hexagon, Green):
• Graph Builder
• DAG Manager
• Agent State Manager
• OpenAI LLM Integration
d) Output Components (Rectangle, Orange):
• Python Code Generator
• Integration Code Output
Key Interactions to Highlight
• Data flow from HAR file to Graph Builder
• Graph Builder to DAG Manager interaction
• LLM consultation flow
• Agent state transitions
• Code generation pipeline
Technical Details
Annotations for:
• File formats (.har, .json)
• LLM models used (GPT-4o, o1-preview)
• Python dependencies
• Authentication flow
Specific Implementation Notes
• The modular structure reflected in the /integuru directory
• The separation between util functions and core processing
• The model management system
• Testing infrastructure
[Browser Capture] → [HAR File] → [Cookie Storage]
• ↓
[Input Processing Layer]
• ↓
[Graph Builder] ←→ [LLM Service]
• ↓
[DAG Manager] ←→ [Agent State]
• ↓
[Code Generator]
• ↓
[Python Integration Code]
Additional Elements
• Component types
• File formats
• External service integration points