rig-core-v0.35.0

Added

• (rig-1197) handle llama.cpp tool call (#1408) (#1409) (by @inqode-lars)

Fixed

• (#1604) delay response_format on initial tool turns (#1622) (by @gold-silver-copper)
• reduce ToolServer contention during tool lookup and execution (#1607) (by @isSerge) - #1607
• (streaming) preserve tool call history, deduplicate prompt (#1590) (by @gold-silver-copper)
• (openai) capture ResponseFailed errors in stream mode (#1582) (by @gabrielrondon)

Other

• (refactor): replace legacy Anthropic constants (#1616) (by @gold-silver-copper) - #1616
• Add ModelLister for Ollama, Anthropic, Mistral, OpenAI, Gemini (#1587) (by @LHelge) - #1587
• gpt image 1.5 (#1543) (by @kevinastock) - #1543
• (rig-core) [breaking] migrate examples to integration tests (#1603) (by @gold-silver-copper)
• Do not stringify strings during tool output (#1608) (by @gold-silver-copper) - #1608
• (rig-core) upgrade rmcp integration to 1.3, gate tests (#1596) (by @gold-silver-copper)

Contributors

• @gold-silver-copper
• @LHelge
• @kevinastock
• @isSerge
• @inqode-lars
• @gabrielrondon