Architecture

Krya.ai is composed of three main layers:

1. Frontend (Streamlit UI)

   • Accepts natural-language prompts
   • Displays generated code, logs, and live status

2. Core Service (Automation Engine)

   • LLM Interface:
     • Sends prompts + system instructions to Google Gemini 2.0 Flash multimodal
     • Receives Python code snippets
   • Execution Layer:
     • Writes code to output.py
     • Runs via subprocess (with error capture)
     • PyAutoGUI scripts for GUI automation
   • Feedback Loop:
     • On error: captures traceback
     • Re-submits prompt+error to the LLM for refined code
     • Retries until success or max attempts

3. Persistence & Logging

   • All events and errors logged in vaxinsight.log (RotatingFileHandler)
   • Session history kept in memory for quick status reporting

flowchart TD
    A[User Prompt] --> B[Streamlit UI]
    B --> C[LLM Call (Gemini)]
    C --> D[Generate output.py]
    D --> E[Execute Code]
    E -->|Success| F[Log & Display]
    E -->|Error| G[Capture Traceback]
    G --> C