Release v1.6.0
What changed
- Added execution provenance metadata to normalized JSON output:
- metadata.model
- metadata.execution_context
- metadata.execution_context.runtime
- Added reproducibility context fields in execution_context:
- provider_protocol
- api_endpoint
- model_requested
- model_resolved
- runner_version
- prompt_hash (sha256:)
- runtime snapshot (stream, system_prompt_provided, runtime controls, timeout, retries)
- Extended provider contract with resolved-model metadata hook (get_last_model_resolved) and
integrated it across protocol providers. - Extended output schema and runtime validators for new additive provenance fields.
- Expanded unit/schema coverage for provenance serialization, validation, and runner guard/error
paths. - Updated technical documentation to reflect v1.6 provenance behavior:
- README.md
- docs/cli-reference.md
- docs/architecture.md
- docs/configuration.md
- docs/output-contract.md
Why it matters
- Improves traceability and auditability of each execution without changing the stateless
architecture. - Makes runs more reproducible by recording effective execution context (protocol, endpoint,
requested/resolved model, runner version, runtime settings, prompt hash). - Preserves the stable output contract while extending metadata in a backward-compatible additive
way.
Stability Statement
- v1.6.0 is a stable feature release on top of the v1.x baseline.
- No CLI interface breaking changes.
- No provider contract breaking changes.
- No output contract breaking changes (additive metadata only).
Scope Clarification
- The project remains a deterministic, stateless execution layer.
- No orchestration, no memory, no agents, no multi-turn workflow behavior.
- No conversation state was introduced in this release.
GitHub Pull Requests Included
- feat(v1.6): add execution provenance metadata contract by @damienSavoldelli in #26
- release(v1.6.0): prepare provenance metadata release by @damienSavoldelli in #27
Full Changelog: v1.5.0...v1.6.0