How shaheennabi/production-ready-tripplanner-multi-ai-agents-project works
Project type and purpose
• This is a modular, full-stack AI pipeline application (“TripPlanner Multi-AI Agents”) that uses Taskflow AI to orchestrate multiple AI agents.
• Purpose: take a single user query (destination + preferences) and, via a Streamlit UI, produce a complete trip report by calling specialized agents (web research, travel info, reporter).
File structure
• Top-level:
• “demo.py” or “deployment/app.py” – entrypoint for the UI/pipeline
• “src/agentic/agents/…” – definitions of three agents
• “src/agentic/tools/…” – five tool modules
• “.github/workflows/deploy.yml”, “Dockerfile”, “scripts.sh” – CI/CD and containerization
• Pattern: layered modular design – UI layer, orchestrator, agent components, tool plugins, external API integrations, deployment pipeline.
Readme insights
• Describes a “main pipeline” that wires together agents via Taskflow AI.
• Lists three agents: Web Research Agent, Travel Agent, Reporter Agent.
• Lists five tools: SerperSearch, WikiArticles, WikiImages, SearchFlights, GetWeatherData.
• Shows a deployment pipeline: GitHub Actions → Docker → AWS ECR → EC2.
Components
a. Main components:
• User Interface (Streamlit app in demo.py / deployment/app.py)
• Pipeline Orchestrator (Taskflow AI core)
• Agents:
• Web Research Agent
• Travel Agent
• Reporter Agent
• Tools / Plugins:
• Serper Search
• Wikipedia Article Search
• Wikipedia Image Search
• Flight Search (Amadeus)
• Weather Fetch (Weather.com)
• External Services / APIs: OpenAI, Serper API, Wikipedia API, Amadeus API, Weather.com API
• CI/CD & Deployment: GitHub Actions, Docker, AWS ECR, EC2 self-hosted runner
b. Relationships & Interaction:
• UI submits user input to Pipeline Orchestrator
• Orchestrator instantiates Agents in sequence (Web → Travel → Reporter)
• Each Agent calls its Tools, which in turn call external APIs
• Reporter Agent aggregates outputs and returns final report to UI
• On code push, GitHub Actions builds Docker image, pushes to AWS ECR, and deploys container to EC2
c. Architectural patterns & principles:
• Modular “plugin” pattern: each Agent is a self-contained module with its own tools
• Layered design: UI / Orchestrator / Agents / Tools / External Services
• Separation of concerns: Agents focus on a single domain, Tools abstract external API interactions
• CI/CD pipeline ensures reproducible, containerized deployment
d. Key technologies & frameworks:
• Python 3.10+, Taskflow AI
• Streamlit (UI)
• Docker, GitHub Actions
• AWS ECR & EC2
• Third-party APIs: OpenAI, SerperSearch, WikipediaTools, Amadeus, Weather.com