Rackwise: A Eurorack Agent That Separates Calculation From Evidence
This is a submission for the Sanity Challenge, Path One: Ship an Agent That Queries Real Content . What I Built I started this challenge knowing almost nothing about Sanity. Eurorack was new to me too. I had seen modular synthesizers before, but I didn't understand what HP meant, why module depth mattered, or how the different power rails were calculated. After reading the challenge prompt, I…
This project, called Rackwise, tackles a challenge in the Sanity Challenge by building an agent that queries real content. The author discovered that modular synthesizer systems can have compatibility issues when specifications change between product variants. To address this, Rackwise was created to assemble a rack row, choose a power reserve, and answer questions about space and power requirements.
The agent uses a TypeScript calculation engine for numerical checks and relies on Sanity for source-backed knowledge. Instead of adding every useful page and letting the model search through everything, the corpus was reduced to a small set of focused manufacturer sources and specification files. This allowed Rackwise to work with a more intentional corpus, giving cleaner evidence.
One unexpected benefit of using Sanity was that it highlighted inconsistencies in the Knowledge Base. For example, if there were conflicting claims about a module's recording duration, Sanity would surface both claims with their sources and ask the user to resolve the issue. This feature helped Rackwise handle module variants separately and mark unverified ones.
Rackwise treats different modules like Expert Sleepers' Disting EX and Disting NT as separate products due to their varying dimensions and power requirements. Each specification is converted into individual claims containing hardware details, properties checked, values, units, Knowledge Base paths, and original sources. This separation allows Rackwise to accurately compare claims with the catalog and generate a constrained explanation of the answer.
The project is built without a dedicated agent framework backend, focusing on key boundaries such as request validation, local calculation, reading the Sanity Knowledge Base through MCP, selecting entries for selected hardware, retrieving relevant content, extracting source-backed claims, rejecting unsupported sources and wrong variants, comparing claims with the catalog, generating a constrained explanation, and returning the result, evidence, gaps, and trace.
Rackwise currently includes 33 tests covering various aspects of the planner, MCP responses, claim extraction, source validation, module routing, coverage, and variant separation. It also features per-device and per-IP request limits to prevent overloading the model endpoint. The project's code repository is available at https://github.com/Kavinhbn/Eurostack. A live demo can be accessed at https://eurostack.vercel.app/.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
