What it does

contextweigh reads a recorded run as a versioned JSON contract and reports exactly which content blocks repeat, how many bytes and whitespace tokens the duplicates cost, and whether the run passes, fails, or is unknown (unknown is kept distinct from pass and fail). It runs entirely on your machine: no upload, no account, no external calls.

Try it locally

git clone https://github.com/LibertyDigitalSystems/contextweigh.com
cd contextweigh.com
node src/cli.mjs path/to/run.json            # machine-readable JSON report
node src/cli.mjs path/to/run.json --markdown # human-readable evidence report

Exit code is part of the contract: 0 pass, 1 fail, 2 unknown or unsupported input.

Supported inputs

Native versioned block contract (conereplay-v1).
One-way import of a recorded OpenAI chat-completions run (openai-chat-v1), with documented message and size caps.

Read the documentation for the full JSON contract, the exit-code contract, and adapter caps, or the implementation notes for the exact acceptance fixtures.

Status

The local CLI, the dual (human and machine) evidence report, and the acceptance fixture matrix ship today. Hosted history, teams, and billing are not yet available: by design they stay off until authentication, isolation, retention, and deletion tests pass. There are no usage metrics to show yet, and no sign-up to collect.