Ollama: Tool Schema Fidelity, Desktop Launch, and Download Safety
Merged fix PR #19686 stops silent loss of tool schema keywords like default and minimum. The broader pattern is contract fidelity across tool calling, OpenAI compatibility, desktop model switching, and blob downloads.
Duration: PT2M31S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-11-14T14:01:00Z
- Audio duration: PT2M31S
Transcript excerpt
This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.
Good morning. It's Saturday, November 14th, 2026 — this is your Ollama developer briefing.
The big pattern today is contract fidelity: making sure what clients send is what models actually get, and what models compute is what you actually see.
First, tool calling. Merged work in PR 19686 fixes silent data loss in tool schemas. When a client defined parameters with keywords like default or minimum, the decoder only kept a handful of known fields and dropped the rest. On write-back, downstream validation saw an incomplete schema. That now preserves…
Related, PR 19692 tackles the Open A I compatibility path. The completions endpoint accepted log probabilities and forwarded them internally, but never returned them to the caller. The proposed fix wires that response through, closing a long-standing gap from issue 12899. If you rely on token confidences for…
Second, desktop orchestration. PR 19691 adds a command line entry point to configure and launch Claude Desktop on Mac, sharing model mappings with app settings. PR 19693 fixes the other end: Save and Restart for Codex now stops its managed background daemon before saving the catalog, so a newly selected model…
Finally, two sharp…
Nearby episodes from Ollama
- Tool Reliability and Mac Inference Tuning
- Faster Concurrent Inference and Unified Desktop Launch
- MLX Goes Parallel
- VRAM Fixes and Launch Polish
- Stricter Model Imports and CLI Output Fixes
- Weekly Recap - MLX Model Expansion and Launch Polish
- Ripple on MLX and Validation Hardening
- MLX Memory Cuts and Tool-Call Fixes