Release v1.2.0

What changed

• Added end-to-end streaming support across the runtime stack:
  • provider stream contract (generate_stream(prompt))
  • CLI stream mode via --stream
  • runner stream orchestration with deterministic final payload assembly
• Implemented protocol-level streaming for:
  • OpenAI-compatible provider
  • Anthropic provider
  • Google provider
  • Mock provider (deterministic test stream)
• Preserved safe fallback behavior:
  • providers without streaming support (for example http) automatically use non-stream
    execution
• Expanded validation coverage for streaming paths:
  • provider unit tests (nominal, retries, malformed events, error mapping)
  • provider contract tests
  • runner and CLI stream behavior tests
• Updated technical documentation to reflect stream behavior and provider support:
  • README.md
  • docs/cli-reference.md
  • docs/architecture.md
  • docs/configuration.md

Why it matters

• Improves execution feedback and UX by showing response chunks progressively during long-
  running calls.
• Keeps integration safety unchanged by preserving the same final JSON/Markdown artifacts.
• Strengthens confidence in production usage through explicit streaming failure-path
  coverage and contract validation.
• Extends multi-provider parity by supporting stream mode consistently across OpenAI-
  compatible, Anthropic, and Google protocols.

Notes

• v1.2.0 is a stable feature release on top of the v1.x baseline.
• No output schema change was introduced.
• No CLI breaking change was introduced.

Stability Statement

• The CLI interface remains stable.
• The provider contract remains stable and now includes optional streaming behavior with
  explicit fallback semantics.
• The output JSON contract remains unchanged and fully compatible with previous v1.x
  expectations.
• Future breaking changes will require a major version increment.

Scope Clarification

• The project remains a deterministic, stateless execution layer.
• Streaming in v1.2.0 is a runtime UX enhancement, not a shift in architecture.
• Out of scope (unchanged):
  • conversational memory
  • orchestration logic
  • agent behavior
  • tool-calling workflows
  • multi-step stateful execution

GitHub Pull Requests Included

• feat(v1.2): add streaming contract and openai-compatible stream path by @damienSavoldelli in #10
• feat(v1.2): add anthropic streaming support by @damienSavoldelli in #11
• feat(v1.2): add google streaming support and contract coverage by @damienSavoldelli in #12
• release(v1.2.0): prepare streaming release and documentation updates by @damienSavoldelli in #13

Full Changelog: v1.1.0...v1.2.0