Compare independent records.
Use a separate reference source to classify missing, unexpected, and late answer records.
Why rawctx
Why rawctx?
Review a disputed AI answer against the evidence recorded at the time.
rawctx groups the submitted answer record, model-run information, source references, and correction history, then reconciles that record with a separate business system. It does not by itself prove answer accuracy, actual delivery, or that a source was used in reasoning.
Independent reconciliation first
Check coverage before advanced cryptography.
A cryptographically consistent log can still be incomplete. Compare it with an independently sourced manifest to find missing, unexpected, and late records before reviewing proof details.
Capture one answer shell.
Store application, environment, question/answer text or hashes, model-run information, policy flags, actor/session hashes, idempotency key, and status.
Record submitted source references.
Attach source_refs, trace ids, and retrieval events as submitted review material, without claiming that a reference proves source use in reasoning.
Check integrity and proof state.
Review correction, void, redaction, Merkle proof, KMS-signed STH, external anchors, and explicit proof states after coverage is established.
What rawctx adds
Keep evidence available for review, export, and disputes.
The answer record stays reviewable even when text storage is hash-only, media downloads expire, or a correction arrives later.
Answer Audit is the system of record
Each submitted answer gets an append-only audit shell with evidence refs, model-run information, source refs, policy flags, and lifecycle events.
rawctx.log_answer(...)Inference commitment binds the model run
Submitted provider or weight-verification artifacts, input/output hashes, and config hashes are bound to the same answer hash without proving delivery.
rawctx.build_inference_commitment(...)Registered media evidence
Audio and video evidence can be registered in a private vault and referenced by source_ref. A reference alone does not prove retrieval or use.
rawctx.register_media_evidence_asset(...)Trace bundles and source_refs
Submitted OTel GenAI bundles, external trace ids, model run ids, answer segments, and source refs become searchable review material.
rawctx.ingest_otel_trace_bundle(...)Trust proof states
Auditors can distinguish ANCHORED records from pending or local-only evidence candidates before making claims.
rawctx trust proof answer <id>Adjacent tools
rawctx does not replace the systems around it.
rawctx records the evidence that connects these systems. Your catalog keeps metadata, your runtime executes business logic, your LLMOps stack observes model behavior, and rawctx keeps the submitted answer, model-run information, and proof trail ready for review.
- Catalogs keep live metadata and ownership.
- Runtime systems execute definitions and SQL.
- LLMOps tools trace prompts, model calls, evals, and retrieval.
- rawctx records the answer evidence receipt and model-run commitment that connects them.
Pilot shape
Bring one answer that already needs evidence.
Start with one AI workflow, one evidence source, and one independent reference system. Measure logging coverage, missing or late records, and how quickly reviewers can retrieve the submitted evidence.