the-dreimar/Dispatcher-AI

HR Agentic Ticketing System V2

Welcome to the cutting-edge Agentic RAG Ticketing system!

Architecture

1. 1.Dispatcher Engine (Gemini 3.1 Pro via `google-genai`)
2. 2.Automatically understands intent and Workday modules from employee issue tickets.
3. 3.Calculates dynamic assignments using workload constraints and resolver skills.
4. 4.Self-reflective RAG pre-fetches similar resolutions and rulebook guidelines, providing analysts a head start.

1. 1.Self-Improving RAG Loop
2. 2.Each time an analyst closes a ticket, the system automatically uses Gemini to summarize the solution steps.
3. 3.The summary gets mapped into a vector(768) coordinate and placed directly back into the pgvector database via pgvector.
4. 4.Future issues fetching similar intent will surface this newly learned fix.

1. 1.Database Architecture (Supabase)
2. 2.Powered by PostgreSQL with pgvector enabled out of the box.
3. 3.Synthetic Data generated to model an HR ecosystem (500 Employees, Active Matrix, and mock historical closures).

Hugging Face Spaces Deployment

The structure complies exactly with the native Streamlit App space guidelines.

1. 1.Add Secrets: In your Hugging Face Space settings, populate the following secrets:
2. 2.SUPABASE_URL
3. 3.SUPABASE_KEY
4. 4.GEMINI_API_KEY
5. 5.Space requirements will naturally fetch requirements.txt and launch app.py via st.navigation.