Guardian Goose

• 

  Landing

• 

  Suspicion Detected

• 

  Success

Inspiration

The blockchain is often a "black box" where users sign transactions they don't fully understand and lose assets to predators they never see. I wanted to build a tool that pierces through that complexity, transforming raw data into a clear forensic story, helping people understand what they and others are doing.

What it does

Guardian Goose is a forensic post-mortem investigator that decodes the "black box" of exploited Solana transactions. Once a user inputs a suspicious transaction signature, Detective Sherlock Honk (our intrepid investigator) hits the streets of the blockchain to sniff out the data using the Helius API. He then consults with Gemini to perform a forensic analysis and reports back with a "Case File". These reports identify predatory signatures and explain the exploit in plain English, ensuring you finally know exactly how your nest was raided.

How I built it

The architecture consists of an Astro frontend deployed on Vercel and a FastAPI backend hosted on Railway. I integrated Helius for transaction parsing on the Solana chain and Gemini Pro to handle the high-level forensic reasoning. To enhance the investigative experience, I used ElevenLabs to provide the detective's voice, allowing the "Guardian Goose" to walk users through the post-mortem results.

Challenges I ran into

The deployment phase was a battle against infrastructure, specifically resolving Node.js version mismatches on Railway and debugging Vite resolution errors during the Vercel build. I also ran into a wall with the Gemini API rate limits during the final hour of testing, which forced me to migrate to a new account and API key to keep the forensics engine live. Structuring the complex, nested transaction data from the Helius API so that an LLM could consistently identify predatory signatures required constant iteration on the seed.

Accomplishments that I'm proud of

I successfully deployed a functional, full-stack forensic tool that identifies the predator signatures in a live production environment. Maintaining the project's viability through multiple infrastructure failures on Vercel and Railway in the final hour was a significant test of engineering persistence. Seeing the integration of Helius and Gemini Pro actually turn an opaque transaction into a human-readable detective report was a major technical milestone.

What I learned

I learned that spending the first few hours on architecture instead of jumping straight into code pays off massively when things break later. When I had to rip out the Gemini SDK and rewrite the integration mid-hackathon, the modular structure held up because it was designed to be swapped, not patched. I also learned to write tests from the start, not after; having a unit test suite running the whole time meant I could verify core logic in seconds after every breaking change. That confidence to move fast without second-guessing was probably the biggest thing I'm taking away from this project.

What's next for Guardian Goose

The immediate priority for Guardian Goose is to move beyond mere reporting and into agentic workflows that suggest and execute corrective actions for the user. Instead of simply explaining a raid after the fact, the next phase involves an autonomous forensic assistant that can guide a victim through the immediate "post-mortem" recovery steps.

Built With

• astro
• elevenlabs
• fastapi
• gemini
• helius
• react