v0.5.8

Changelog

• a1bf786 feat: add DisableThinking config for reasoning models

v0.5.7

Changelog

• 8a25cd9 fix: handle reasoning_content for Qwen3.5 and similar models

v0.5.6

Changelog

• e507e72 fix: support tar.gz archives for newer llama.cpp releases

v0.5.5

Changelog

v0.5.4

Changelog

• eac3565 feat: add ollama and llama_cpp factories to orchestrator provider registry
• 21eb7a1 feat: merge ratchet orchestration layer into agent plugin
• a589dfa feat: register step.model_pull in orchestrator plugin manifest

v0.5.3

Changelog

• 2d3c3fd feat: add Ollama and llama.cpp local inference providers (#2)
• 4710737 feat: add Ollama type conversion helpers
• 3c36454 feat: add OllamaProvider implementing Provider interface
• 773cfd3 feat: add Thinking fields to Response and StreamEvent
• 1c58f4d feat: add github.com/ollama/ollama/api dependency
• 447c850 feat: add llama.cpp provider, auto-download, and HuggingFace downloader
• 9eb1d57 feat: add shared local inference helpers (ParseThinking, ThinkingStreamParser, LocalAuthMode)
• 8a281c7 feat: add step.model_pull step type
• 3aa2045 feat: thread thinking traces through executor and step output
• 5310a12 feat: wire ollama and llama_cpp into registry, plugin, auth, and model listing
• f6df694 fix: address Copilot review feedback on PR #2
• ff5a6a7 fix: address spec-reviewer feedback for Tasks 7 and 8
• 8c3e3b0 fix: goroutine leaks, panic recovery, channel cleanup in daemon and gemini provider
• e9513e6 security: validate base_url to prevent SSRF via internal network access

v0.5.2

Changelog

v0.5.1

Changelog

• 8670852 feat: add sdk package with reusable daemon-client infrastructure

v0.5.0

Changelog

• 6f77025 ci: fix branch mismatch — trigger on master not main

v0.4.1

Changelog

• 3a3d80c fix: address 3 code review issues in Anthropic providers
• f7da81f fix: sort tool call indices before emitting to ensure deterministic order