How tarek-kerbedj/sql_agent works
Project type and purpose
• This is a Python-based, single-process “full-stack” web app built with Streamlit.
• Purpose: provide data insights via three key features—database analysis, document Q&A, and signal generation—driven by LLMs (via LangChain/AWS Bedrock) and a Cassandra backend.
File structure
• Top level
• insights.py – Streamlit entry point and UI layout
• requirements.txt, startup.sh – dependencies and launch script
• .github/workflows – CI/CD pipelines for Azure deployment
• .streamlit – Streamlit configuration
• components/
• database_insights.py – all Cassandra-related logic
• document_QA.py – document ingestion and Q&A
• signal_generation.py – business-logic for signal creation
• utils/
• core_funcs.py – LangChain chains and summaries
• llm_utilities.py – LLM API orchestration (prompt parsing, chain connections)
• util_funcs.py – misc. file I/O, zipping, downloads
• style.py – UI styling helpers
• other/
• cassandra/ – secure-connect bundle & tokens
• credentials/ – login CSV
• prompts/ – YAML prompt templates
• wheels/ – bundled boto3/botocore wheels for AWS Bedrock
• images/ – static assets
README for additional insights
• Confirms insights.py is the sole entry point.
• Outlines the three component modules.
• Installation and run instructions via Streamlit.