v0.26.5

What's Changed

• feat(local-llm): default Ollama to our fine-tuned gemma4-comfyui-mcp ladder by @artokun in #151
• fix(ollama): stop clamping the fine-tune's context to 16K — model-aware num_ctx by @artokun in #152
• fix(ollama): tool-loop breaker — block identical repeat calls, end the turn at 4 repeats by @artokun in #153
• feat(ollama): free local-model VRAM during generation + pause chat until it finishes by @artokun in #154

Full Changelog: v0.26.4...v0.26.5