Multi-Agent Text-to-SQL and ML Insights System (Production-Grade)

Overview

Enhanced for industrial use with MLOps, security, and scalability. Uses LangGraph for multi-agent workflows, integrated with BigQuery/SparkML on GCP.

Setup

1. Clone: git clone https://github.com/your-username/multi-agent-insights.git
2. Install: pip install -r requirements.txt (prod) and pip install -r dev-requirements.txt (dev)
3. Secrets: Use GCP Secret Manager or Vault; set via .env.
4. Monitoring: Set LANGCHAIN_API_KEY for LangSmith.
5. Deploy: docker build -t multi-agent-insights . then use k8s/ for GCP GKE.
6. CI/CD: Configured via GitHub Actions; triggers on push/PR/merge.

Security Best Practices

• Prompt Injection: Secondary LLM guard .
• Access: GCP IAM RBAC .
• Compliance: OWASP LLM mitigations .

MLOps

• CI/CD: GitHub Actions for test/build/deploy .
• Monitoring: LangSmith for LLMs , Prometheus for metrics .
• Retraining: Scheduled via Airflow/Vertex AI .

Architecture Diagram

system.png

Contributing

Follow guidelines in CONTRIBUTING.md (added: code reviews, security scans).

License

MIT