Agentic AI

Project Management Agentic AI

A multi-agent assistant that breaks down ambitious project requests, coordinates specialised agents, calls integrated tools, and returns traceable artefacts—via command line or Streamlit.

View on GitHub Back to Healthcare projects
Diagram: supervisor and specialist agents

Overview

Natural-language tasks flow through a supervisor that plans execution, invokes research, analytics, planning, writing, and review agents in the right order, and consolidates outputs. Short-term coordination uses Redis or in-memory state; long-term recall uses a ChromaDB-backed vector store for retrieval-augmented behaviour.

Capabilities

Structured orchestration

Task decomposition, dependency-aware scheduling, parallel work where safe, and reviewer validation before finalisation.

Tool-connected agents

SQL warehouse access, internal knowledge search, calendars, ticketing systems, and external APIs—configurable per deployment.

Memory model

Ephemeral state for the active job plus vector memory so prior successful runs inform future requests.

Multiple interfaces

One-shot CLI execution, interactive sessions, JSON task files, and a Streamlit UI for guided use.

Visual summary

Diagram: task lifecycle, CLI and Streamlit entry points
Task lifecycle and supported command-line and Streamlit entry points. The supervisor and specialist overview is shown in the hero above.

Technology stack

  • Python 3.8+
  • OpenAI APIs
  • Redis (optional)
  • ChromaDB
  • Streamlit

See the repository Documentation folder for quick start, architecture, supervisor flow, and code walkthrough.

Want a similar internal copilot for programme or product teams?

Get in touch