TraceVox Research documentation

TraceVox separates exploring existing evidence (works immediately on this website, nothing to install) from generating new evidence (requires a TraceVox Research Runner on your own compute or a self-hosted instance — the website never trains models on its own servers).

Machine-readable documentation

• /llms.txt — concise overview and canonical URLs
• /llms-guide.txt — practical guide (explore, self-host, publish)
• /llms-full.txt — full reference: architecture, research.v1 step schema, training event schema, private/public APIs, runner protocol and typed-job security model, visibility and publication, reproducibility semantics (Recorded Replay vs Computational Rerun vs Reproduction), provenance manifests, extension points

Public data

Published experiments are plain JSON bundles (tracevox.public.bundle.v1): /public-research/index.json. Private research is private by default; publication is explicit and granular.

TraceVox Research home · Public Research Library · Documentation · llms.txt