Release v1.7.0

What changed

• Added a structured runtime error taxonomy with stable normalized codes:
  • auth_error
  • rate_limit
  • timeout
  • invalid_request
  • network_error
  • provider_error
• Added per-run observability artifacts via --log-run-dir:
  • request.json (sanitized request diagnostics)
  • response.json on success
  • error.json on runtime failures
• Added centralized runtime error normalization in core and integrated it into CLI failure
  diagnostics.
• Extended runtime configuration resolution to support log_run_dir through CLI/TOML.
• Updated technical documentation to reflect observability behavior and structured errors:
  • README.md
  • docs/cli-reference.md
  • docs/architecture.md
  • docs/configuration.md

Why it matters

• Improves CI/ops reliability with machine-readable error classes.
• Makes failures easier to triage using deterministic per-run artifacts.
• Preserves the existing stateless execution boundary while adding operational observability.

Notes

• v1.7.0 is a stable observability release.
• No CLI breaking changes.
• No output schema/contract breaking changes.

Stability Statement

• The CLI interface, provider contract, and output schema remain stable.
• Existing execution model invariants are preserved.

Scope Clarification

• The project remains a deterministic, stateless execution layer.
• No orchestration, memory, agent behavior, or multi-turn state was introduced.

GitHub Pull Requests Included

• feat(v1.7): add core runtime error taxonomy mapping by @damienSavoldelli in #31
• feat(v1.7): add --log-run-dir with structured request/response/error artifacts by @damienSavoldelli in #32
• docs(v1.7): document observability logs and structured runtime errors by @damienSavoldelli in #33
• release: prepare v1.7.0 by @damienSavoldelli in #34

Full Changelog: v1.6.2...v1.7.0