OmniFlow Beta

OmniFlow Beta is a multi-user AI orchestration backend on Azure Functions + Azure Blob Storage.

This repository exposes a quasi-MCP pattern: one deterministic orchestration endpoint with tool registry, validation, and structured responses.

Test status

This project is currently in test/beta mode.

Beta modes

1) Native UI beta (full feature path)

Use this mode when you want the full OmniFlow experience, including Context Builder behavior and end-to-end orchestration features.

• Frontend: ai-chatbot/
• Backend: backend/
• Includes WP6 context-building flow and broader feature coverage

2) Custom GPT beta (integration path)

Use this mode when you want to connect through OpenAPI/tool calling without running the full native UI stack.

Test model (Custom GPT):

• OmniFlow Personal Assistance (test)

What you can connect

• POST /api/tool_call_handler - primary orchestration endpoint
• POST /api/read_many_blobs - batch blob reads
• POST /api/save_interaction - interaction log input for semantic processing

For full integration details, see:

• docs/shared/MCP_AND_QUASI_MCP.md

Why quasi-MCP

• Single endpoint orchestration instead of one endpoint per tool
• Registry-driven tool specs and argument normalization
• Structured JSON errors and deterministic execution paths
• Multi-user isolation via X-User-Id and namespaced storage

Current project layout

• backend/ - Azure Functions backend (active in this repo)
• ai-chatbot/ - Next.js chat frontend
• frontend/ - Streamlit lab frontend (legacy/testing)
• docs/ - source-of-truth documentation
• tests/ - unit and e2e tests

Quick start

1. Install backend dependencies: pip install -r backend/requirements.txt
2. Start Azurite: azurite
3. Run functions locally: cd backend && func start

Optional frontend runs:

• Next.js app: cd ai-chatbot && pnpm install && pnpm dev
• Streamlit lab UI: cd frontend && streamlit run app.py

For Custom GPT setup instead of Native UI, follow:

• docs/shared/MCP_AND_QUASI_MCP.md

Key docs

• Docs index: docs/README.md
• Quasi-MCP guide: docs/shared/MCP_AND_QUASI_MCP.md
• Deployment: docs/shared/DEPLOYMENT.md
• Tool call playbook: FUNCTION_CALLS_PLAYBOOK.md
• Privacy policy: docs/shared/PRIVACY_POLICY.md
• Changelog: CHANGELOG.md

Limitations

• WP6 DEEP path is not deterministically forceable end-to-end in all scenarios.
• Streamlit frontend is maintained as legacy/lab tooling.

License

MIT