# Project export: BizTrip Voice Assistant

This document was generated by HackStack to give an AI agent context about a hackathon project. Sections are labeled with their provenance; content marked as truncated was cut to keep this document small.

## Project metadata

- Hackathon: OpenAI Build Week
- Tagline: A voice-first travel companion of staff for business trips: it knows your itinerary, company policy, meetings, expenses, and local culture, and can take actions while you’re moving.
- Devpost: https://devpost.com/software/biztrip-voice-assistant
- GitHub: not linked
- Video: https://www.youtube.com/embed/qaN4FCIbzJA?enablejsapi=1&hl=en_US&rel=0&start=&version=3&wmode=transparent
- Team: contributor stats unavailable

## Devpost submission (written by the team)

### Inspiration

Business travelers often move quickly between airports, train stations, hotels, and meeting locations, where typing questions or manually searching for information is inconvenient. BizTrip Voice Assistant was inspired by the need for a hands-free assistant while users are on the move.

### What it does

It listens to user requests, understands both speech and text input, and resolves tasks using company policies, project documents, travel preferences, and organizational knowledge.

### How we built it

The system consists of: Core Agent Server (Search + Info + Memory) RAG Service with Admin Console Speech Processing Service iOS Mobile App LangSmith Tracing

### Challenges we ran into

Handling high-quality audio streaming input which aligns with websocket server contract, since audio format, sample rate, and data integrity directly affect transcription accuracy and overall task correctness. Designing an efficient retrieval pipeline to balance correctness, relevance, and low latency. Query rewrite and HNSW indexing techniques were used to improve the precision, recall and latency. Structuring the agent graph around business travel subtasks, including planning, in-trip assistance, reimbursement, and follow-up management. Evaluating agent behavior across tool selection, retrieval quality, and final answer correctness.

### Accomplishments we're proud of

The retrieval chain achieves 97.5% in recall and 67.56% in precision on average. And the current MVP reaches 88% in tool-calling accuracy and 80% in task correctness. The mobile app and backend system is fully functioning as designed.

### What we learned

We have learned that how voice interaction, enterprise knowledge retrieval, and agent-based task execution can work together in a practical tool.

### What's next

The next focus is improving the retrieval chain so BizTrip can handle longer, more complex conversations with higher accuracy. We are also working on enhancing noise deduction of audio stream processing to make the interaction with the assistant more seamless, natural, and reliable during real business travel scenarios.

## README (from the GitHub repository)

No README available.

## Detected evidence (automated analysis)

No repository was indexed for this project. Claimed technologies below could not be checked against code.
- Docker (technology) — claimed on Devpost, not found in the code
- FastAPI (technology) — claimed on Devpost, not found in the code
- OpenAI (technology) — claimed on Devpost, not found in the code
- Python (language) — claimed on Devpost, not found in the code

## Codebase structure

No repository index available.

## Key source files

No repository index available; no source files included.