{"name":"vegaml","repo":"NandhaKishorM/vegaml","url":"https://midorreal.com/project/nandhakishorm-vegaml","repository":"https://github.com/NandhaKishorM/vegaml","description":"Typed decisions from a frozen language model and a small physics engine: calibrated probabilities, conformal sets, 73k context, images.","facts":{"license":{"value":"Apache-2.0","source":"https://github.com/NandhaKishorM/vegaml","checked":"22 Nov 2026"},"language":{"value":"Python","source":"https://github.com/NandhaKishorM/vegaml","checked":"22 Nov 2026"},"last_release":{"value":"v0.18.0, 21 Nov 2026","source":"https://github.com/NandhaKishorM/vegaml/releases","checked":"22 Nov 2026"},"releases":{"value":"37","source":"https://github.com/NandhaKishorM/vegaml/releases","checked":"21 Nov 2026"},"contributors":{"value":"10","source":"https://github.com/NandhaKishorM/vegaml/graphs/contributors","checked":"21 Nov 2026"},"archived":{"value":"no","source":"https://github.com/NandhaKishorM/vegaml","checked":"22 Nov 2026"}},"stars":317.0,"writeup":{"what_it_is":"vegaml is a Python package that produces typed decisions from a frozen language model and a 14.3M-parameter engine. Given a state and questions with fixed answer types (choice, score, yes/no), it returns answers with calibrated probabilities, conformal answer sets and an abstain flag. It generates no text.","audience":"Developers building typed decision pipelines on their own hardware","claims":[{"kind":"specific","claim":"Uses a 14.3M-parameter engine that works with any frozen reader from 0.8B to 27B parameters","excerpt":"vega-U: one 14.3M engine, any frozen reader, from 0.8B to 27B","status":"not_checked"},{"kind":"specific","claim":"Supports a context of 73,728 tokens","excerpt":"73,728-token context","status":"not_checked"},{"kind":"specific","claim":"Returns calibrated probabilities with an explicit abstain flag","excerpt":"Each carries a calibrated probability, a conformal answer set and an explicit abstain flag.","status":"not_checked"},{"kind":"specific","claim":"Can attach a user-supplied base model without retraining the engine","excerpt":"From 0.12, a base model the checkpoint ships no port for is something you attach yourself","status":"not_checked"},{"kind":"specific","claim":"Inference runs on a single T4 GPU","excerpt":"Inference on a single T4, end to end, in the browser.","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-17"},"why_now":null,"owner_supplied":null,"score":{"verdict":"sleeper","hype":18,"reality":87,"gap":-70,"momentum":1.2,"confidence":85,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T07:00:00+00:00","verdict_source":"rule"},"early_read":null,"changes":[{"date":"2026-11-21T15:37:33+00:00","text":"Latest release v0.18.0","source":"https://github.com/NandhaKishorM/vegaml/releases"},{"date":"2026-10-09T18:16:02+00:00","text":"First release","source":"https://github.com/NandhaKishorM/vegaml/releases"},{"date":"2026-10-09T17:45:54+00:00","text":"Repository created","source":"https://github.com/NandhaKishorM/vegaml"}],"cite":"vegaml is a GitHub project at https://github.com/NandhaKishorM/vegaml written mainly in Python, described there as \"Typed decisions from a frozen language model and a small physics engine: calibrated probabilities, conformal sets, 73k context, images\".","attribution":"Data from Mid or Real, https://midorreal.com/project/nandhakishorm-vegaml, checked 22 November 2026. Attribution with a link is required.","still_real":null}