// the find
AdritPal08/TravelPlanner-CrewAi-Agents-Streamlit
Generate personalized travel itineraries based on user preferences.
A Streamlit app that orchestrates 7 CrewAI agents (destination research, accommodation, transportation, weather, itinerary, budget, orchestrator) to generate travel itineraries using LLaMA 3.1 70B via Groq. It's a demo project showing multi-agent coordination patterns, not a production travel tool. Useful if you want a working CrewAI example to learn from.
The agent decomposition is sensible — separating weather, budget, transportation, and accommodation into distinct agents with focused prompts is a reasonable architecture for this kind of task. Groq as the inference backend means the sequential agent chain runs fast enough to not be painful in a demo. The Streamlit UI keeps the setup friction low — no frontend build step, just Python. Sample output docs in Travel Doc/ let you evaluate quality before running it yourself.
Last commit was August 2024 and it pins to a specific LLaMA 3.1 70B via Groq — CrewAI has had breaking API changes since then, so the requirements.txt is likely stale and `pip install -r requirements.txt` may not produce a working environment without version pinning fixes. The repo ships a trained_agents_data.pkl file in source control, which is a bad habit (binary blobs in git, and pickle files are an arbitrary code execution vector if you run someone else's). There's no error handling visible in the architecture — if any agent in the chain fails or the Groq API rate-limits, the whole itinerary silently fails or crashes. Real travel data quality depends entirely on SERPER search results, which are not verified, so the output can hallucinate prices, hours, and availability with confidence.