What's Changed

• Redact provider keys in printed structs by @s-kostyaev in #276
• Switch to Open AI Response API, keep reasoning between turns for Open AI and Claude by @ahyatt in #281
• Misc README, test fixes, plus fix tool calling error behavior by @ahyatt in #282
• Add Gemini 3.5 Flash to models by @ahyatt in #284
• Fix error with logged too calls, remove obsolete defmethod by @ahyatt in #283
• Replace when-let and if-let to their starred versions by @ahyatt in #285
• Fix unit and integration tests, fixing a Claude tool call issue by @ahyatt in #286
• Add Qwen 3.7, add reasoning to Qwen 3.x models by @ahyatt in #287

Full Changelog: 0.30.3...0.31.0

What's Changed

• Fix OpenRouter chat model handling by @ahyatt in #279

Full Changelog: 0.30.2...0.30.3

What's Changed

• Fix a JSON encoding error caused by utf-8 strings for OpenAI, Ollama by @ahyatt in #264
• Fix OpenAI streaming tool calls handling by @renatofdds in #265
• Add support for Kimi K2.6, Qwen 3.6, Deepseek V4 by @ahyatt in #267
• Add Qwen 3.6 and Claude Opus 4.7 models by @ahyatt in #266
• Add reasoning controls to DeepSeek by @ahyatt in #269
• Fix Gemini reasoning by using new-style request with more models by @ahyatt in #270
• Return token counts when streaming by @ahyatt in #271
• Signal tool call errors for invalid JSON arguments by @s-kostyaev in #272
• Add Mistral Medium 3.5 model support by @ahyatt in #277
• Add XiaoMi MiMo-V2.5 model support by @ahyatt in #278

New Contributors

• @renatofdds made their first contribution in #265

Full Changelog: 0.30.1...0.30.2

What's Changed

• Fix tool call response to include reasoning by @ahyatt in #262
• Add reasoning support for streaming for Open AI compat providers by @ahyatt in #263

Full Changelog: 0.30.0...0.30.1

What's Changed

• Added Gemini 3.1 Pro model by @ahyatt in #250
• Add Qwen 3.5, LFM2 and LFM2.5 thinking by @ahyatt in #251
• Fix issue of zero-arg tools to be called by @ahyatt in #254
• Return input and output token counts in multi-output by @ahyatt in #252
• Add ChatGPT 5.4, with extra context by @ahyatt in #255
• Add Claude Sonnet 4.6 by @ahyatt in #256
• Add support for Open AI compatible reasoning_content blocks by @ahyatt in #258
• Add gemma 4 by @ahyatt in #259
• Add Gemini 3.1 Flash Lite by @ahyatt in #260
• Add OpenRouter as top-level model type by @ahyatt in #261

Full Changelog: 0.29.0...0.30.0

What's Changed

• Almost always raise new llm specific signals, add llm specific error handling by @ahyatt in #238
• Fix(llm-ollama): Set :reasoning to 'none does not take effect by @yydcnjjw in #239
• When we detect an error calling tools, use the error callback instead by @ahyatt in #240
• Normalize false values when tools are called with or return false by @ahyatt in #242
• Add Claude Opus 4.6 by @ahyatt in #243
• Substitute json false values in results without recursion by @ahyatt in #244
• Set Gemini default to 3.0 pro by @ahyatt in #245
• Fix bug calling two async tools in parallel by @ahyatt in #246
• Add Kimi k2.5, glm-5 models, modernize default context length by @ahyatt in #247
• Add Qwen 3 Coder Next by @ahyatt in #249

New Contributors

• @yydcnjjw made their first contribution in #239

Full Changelog: 0.28.5...0.29.0

What's Changed

• Fix for running tools in the original buffer with streaming by @ahyatt in #235
• Improve the tool calling docs in the README by @ahyatt in #236

Full Changelog: 0.28.4...0.28.5

What's Changed

• Add Gemini 3 Flash by @ahyatt in #223
• Fix breakage in Ollama streaming tool calls and incorrect capability by @ahyatt in #225
• Update Ollama reasoning to use the latest reaosning API by @ahyatt in #228
• Fixed bad interaction lines that were added in Ollama tool calling by @ahyatt in #227
• Run tools in the original buffer by @ahyatt in #233
• Added gpt-oss model and enabled low/medium/high reasoning for it by @ahyatt in #232

Full Changelog: 0.28.2...0.28.4

What's Changed

• Recognize post 5.0 Chat GPT models such as the new GPT-5.2 by @ahyatt in #222

Full Changelog: 0.28.0...0.28.2

What's Changed

• Fix for tool use in Ollama by @ahyatt in #217
• Add Gemini 3, add thought signatures and fix function responses by @ahyatt in #218
• Add structured output to Claude Sonnet 4.5, Opus 4.1 by @ahyatt in #219
• Add tool choice for forcing tools on or off by @ahyatt in #220
• Handle empty Claude API responses gracefully by @ahyatt in #221

Full Changelog: 0.27.3...0.28.0