Hawkeye

AI Investigator
Dashboard

Inspiration

A trade can generate dozens of messages, but understanding what happened shouldn’t require manually piecing them together. Inspired by Lucera’s focus on reliable trading infrastructure, we created Hawkeye to make trading activity easier to investigate through clear timelines, traceable evidence, and AI-assisted explanations.

What it does

Hawkeye analyzes a supported subset of FIX messages, the messages trading systems exchange about orders and executions. It reconstructs order activity, flags duplicate executions, and highlights missing or conflicting information. Historical market-data comparisons provide context for execution prices. An AI investigator helps users understand the findings and identify what to investigate next, while deterministic code handles the underlying calculations.

How we built it

We built Hawkeye with Next.js, TypeScript, and Tailwind CSS. Server-side API routes handle message parsing, order reconstruction, and execution deduplication. We integrated Alpaca for market-data context and the Vercel AI SDK with AI Gateway for the investigator. We used Codex extensively to accelerate implementation and debugging, GitHub for version control, and Vercel for deployment.

Challenges we ran into

One challenge was distinguishing a repeated execution message from a separate fill. We also needed to handle incomplete logs without inventing missing activity. Connecting the AI assistant introduced deployment challenges involving environment variables and failed API responses. We designed the application so that analysis and reports remain available when AI assistance fails.

Accomplishments that we're proud of

We brought message analysis, market context, and AI explanations into one investigation workflow. Our central demonstration shows how duplicate executions can be identified without inflating filled quantities.

What we learned

We learned that trustworthy trading tools need clear rules for identifiers, ordering, and duplicate handling. We also learned to separate AI explanations from authoritative calculations and to distinguish deployment configuration problems from application errors.

What's next for Hawkeye

We want to expand FIX coverage, improve multi-file investigation, and strengthen testing across more complex order lifecycles. We also plan to develop richer holdings visualizations and execution comparisons while keeping every result connected to its supporting evidence.