Reese | ReStructure AI
Reese is a company-analysis software sample. It turns a public evidence snapshot and a defined rubric into structured decisions through a customer-configured inference backend.
This repository contains code and a rubric, not trained model weights, adapters, a checkpoint, or a live hosted app. No claim of a new foundation model is made.
Run privately
Install requirements.txt. Set INFERENCE_ENDPOINT, INFERENCE_MODEL and RESTRUCTURE_API_KEY for your authorized backend. It must support the request and response schema used in run_sample.py. Then run python run_sample.py.
The adapter has no preset provider endpoint or model. It is not runnable until you configure a compatible backend. Keep credentials server-side and outside this repository. Use only a dedicated ReStructure credential; never substitute another business's key.
Evidence and limits
The example subject is Cloudflare's public product documentation. Its rubric covers ten company-analysis categories, using an ordered ten-level scale. Thin evidence does not support an overall company score. Missing data is not a negative verdict.
Output is saved locally and is not included in this repository. No performance measurements, raw model outputs, rankings or comparative benchmark claims are published.
This is not investment advice, suitability screening, a trade recommendation or permission to execute an order. Backend output requires independent review. Review your backend's license and terms before public deployment, training or redistribution.