How harrishphotos/sales-forecast-ai works
Project Type & Purpose
• It’s a microservices-style backend system built with FastAPI.
• Four agents (each in its own Docker container) collaborate to process historical data, generate forecasts, and recommend strategies.
• A shared Python package (shared_models) holds common schemas and exceptions.
File Structure
• Top-level components:
– docker-compose.yml orchestrates the containers over a private Docker network.
– Four directories (each with a Dockerfile & app.py):
• analysis_agent – historical data processing
• prediction_agent – forecast generation
• recommender_agent – strategy recommendation
• orchestration_agent – pipeline controller, includes pipeline.py, repository.py, and ops_module/report_ops.py
– shared_models – common code shared by all agents
• No frontend: everything is backend API services communicating via HTTP.
README for Additional Insights
• Confirms the three‐agent pipeline (analysis, prediction, recommender) plus the orchestrator.
• Launch via docker-compose up --build.
• Each agent exposes its own FastAPI endpoint inside the Docker network.
Don’t Overthink—Be Thorough
– Capture each agent as a separate component.
– Show how they interconnect in the exact sequence defined by the orchestration pipeline.
– Include all architectural elements (Docker Compose, network, shared library, data store, final consumer) for a complete picture.