v0.7.2

Changelog

• 7aedf73 feat(executor): emit trust_allow/deny/ask events for audit trail
• 7a5ba68 feat(genkit): make adapter InteractiveArgs configurable via ExtraArgs

v0.7.1

Changelog

• 18d9bb2 feat(policy): add GlobMatcher with ** support for path policies

v0.7.0

Changelog

• 2301a54 feat(genkit): add PromptHandler for auto-responding to PTY screen prompts
• 42887fb feat(orchestrator): extend WorkspaceSpec with per-agent mount specs
• d59e9c9 feat(policy): add PermissionStore for SQLite-persisted trust grants
• 00f760c feat(policy): add TrustEngine, mode presets, Claude Code + ratchet parsers
• 482b7b8 fix(policy): atomic Grant transaction, propagate created_at error, log Check DB errors
• 6b2176d fix(policy): resolve 5 code review issues in TrustEngine + PermissionStore
• a457b47 fix: address 16 code review issues (3 critical, security + concurrency)

v0.8.1

v0.8.1

Team Fixes

• send_message tool works — renamed type param to message_type (JSON Schema keyword collision caused LLMs to silently drop it). Made optional with default "task".
• Error messages visible — team errors now show actual messages instead of empty error:
• Thinking event handling — mesh forwarder properly maps thinking events; team converter suppresses them from output

Known Issue

Qwen3 model thinking/reasoning may still appear in team output as regular text. This is because some models output reasoning as content (not in the thinking field) for complex multi-tool prompts. A future fix will enable Ollama's think: true config for team agents.

Full Changelog

GoCodeAlone/ratchet-cli@v0.8.0...v0.8.1

v0.6.16

Changelog

• 095899d feat: InteractiveArgs for PTY adapters — auto-approve permissions

v0.6.15

Changelog

• ae91367 fix: interactive PTY as primary mode with no exec fallback

v0.6.14

Changelog

• ea97a7f feat: Copilot interactive PTY + screen-diff extraction

v0.6.13

Changelog

• d6572a0 feat: ACP client provider for orchestrating ACP-compliant agents

v0.6.12

Changelog

• 22e5dca feat: SupportsInteractivePTY() flag per adapter

v0.6.11

Changelog

• 3dd79b0 fix: Copilot response extraction — track user input before collecting responses