# TeamAIWork

- **Event:** [The Recursive Agents Hackathon](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon)
- **When:** Sat, Oct 17 at 9:00 AM – 10:00 PM (PDT)
- **Where:** San Francisco, CA
- **Team:** [Burak Keskin](https://cerebralvalley.ai/u/Burak123)
- **GitHub:** https://github.com/Bkeskin10/recursive-travel-agents
- **Demo video:** https://youtu.be/Ym7-RE7ipoA
- **Gallery:** https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/16

# Recursive Travel AI — Self-Improving Multi-Agent Travel Planner

**Project overview**

Recursive Travel AI is a multi-agent travel planning application that generates personalized itineraries and iteratively improves them through automated research, verification, and revision.

Unlike a traditional one-shot AI travel planner, the system uses a recursive feedback loop to identify potential problems in travel schedules, transportation logistics, geographic routing, and budget assumptions.

**How it works**

1. **Planner Agent:** Generates an initial itinerary based on destination, duration, budget, and traveler preferences.
2. **Researcher Agent:** Reviews the itinerary for practical travel considerations and potential inaccuracies.
3. **Verifier Agent:** Evaluates the itinerary and identifies inconsistencies or unresolved issues.
4. **Improver Agent:** Revises the itinerary using feedback from the researcher and verifier.
5. **Recursive Orchestrator:** Repeats the evaluation and improvement process for up to three improvement cycles, with a final verification step.
6. **Budget Validator:** Checks structured budget arithmetic when sufficient expense information is available.

**Key features**

- Multi-agent collaboration using Crusoe Managed Inference
- Recursive itinerary evaluation and self-improvement
- Transparent execution history and verification reports
- Budget allocation checks
- Interactive Streamlit interface
- Downloadable Markdown itineraries
- Publicly deployed demonstration

**What makes it different**

The system doesn't simply generate an answer and stop. It attempts to evaluate its own work, revise detected problems, and make remaining limitations visible to the user.

When an itinerary cannot be fully verified, the application reports the unresolved issues instead of presenting an unsupported claim of accuracy.

**Current limitations**

The prototype uses estimated travel costs and model-based logistical reviews rather than live transportation, booking, or mapping APIs. Some generated itineraries may retain unresolved issues after the maximum improvement cycles.

**Technology**

Python, Streamlit, Crusoe Managed Inference, and a custom multi-agent orchestration workflow.

**Live demo:** https://recursive-travel-ai.streamlit.app

**GitHub:** https://github.com/Bkeskin10/recursive-travel-agents

## More from The Recursive Agents Hackathon

- [Unfoolable](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/13)
- [Repro](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/14)
- [CORE GENESIS](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/15)
- [HOLDOUT](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/17)
- [AutoFab](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/18)
- [Crusoe-Cooler-System](https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/19)

---

Markdown version of https://cerebralvalley.ai/e/crusoe-recursive-agents-hackathon/hackathon/gallery/16. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt