Claim Ledger gives a project review a paper trail

Claim Ledger is a browser-based tool designed to provide a paper trail for project reviews. It allows a teammate to perform an independent second pass on a project, with the handoff including a README, setup notes, test records, and known limitations. The tool connects a specific claim to the relevant passages, keeping the review decision with that evidence.

The independent QA was fully automated, and the reviewer supplies the verdict and note. The interface labels different types of documentation to ensure clear distinctions. The workflow involves loading documents, entering or extracting claims, finding candidate receipts, inspecting their source context, recording a verdict, and exporting the handoff.

Each receipt includes its filename, line range, retrieval mode, score, and content fingerprints. The reviewer can export the review as Markdown or JSON. The tool uses a retrieval core that combines BM25 keyword search with the open mixedbread-ai/mxbai-embed-xsmall-v1 embedding model. The model ranks candidate excerpts but does not write answers or set verdicts.

The application is licensed under Apache-2.0, and the source, tests, benchmark cases, raw results, and correction history are available on GitHub. The model and runtime binaries are distributed separately. The tool is designed to be inspectable and replaceable, allowing developers to compare different AI methods and decide whether the model helps their documents.

However, the tool has only been tested on Chromium, and other browsers and mobile devices remain untested.