Ollama: Tool Calling Gets a Reliability Pass
Today's merges center on making tool calling and model compatibility more robust, with fixes for silently dropped or malformed tool calls, upstream model alignment, and a handful of performance and code-quality cleanups across the codebase.
Duration: PT2M31S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-07-23T13:01:34Z
- 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 day. It's July 23rd, and here's what moved in Ollama today.
The clearest pattern across today's activity is tool calling reliability. Three separate PRs fixed cases where Ollama would silently fail instead of surfacing a usable result. PR 17250 from Daniel Hiltgen tackles the GLM parser: if a model's tool call got cut off before the closing tag, Ollama used to hand back an…
Second theme: model compatibility and upstream alignment. PR 17335 removes Ollama's local Laguna implementation in favor of upstream llama-dot-cpp, while preserving a narrow Metal-specific fix for prompt overflow, and adds translation logic so older GGUF metadata still loads correctly. PR 17336 fixes expert…
Rounding it out, several maintainability commits: hoisting regular expressions to package level in converters, avoiding an allocation in GGUF string writing, and caching a trimmed-string computation in token repeat detection — all from Matias Insaurralde, all small performance wins with no behavior change.
What's next: watch for the GGUF metadata caching PR, 17337, which aims to cut warm-request load times — it's still open and worth tracking. Also expect new library models to require version 0.32.3…
That's…
Nearby episodes from Ollama
- Scheduler Race Conditions Take Center Stage
- Command Line Consolidation and Hardware Fixes
- Thinking Models Leak Control Tokens
- Weekly Recap - Hardening the Runtime, One Edge Case at a Time
- Closing the Gaps Between API Layers
- Cleaning Up the Agent Loop and Widening Model Support
- Agent Package Gets a Major Consolidation
- Agent Hardening and a Path Traversal Fix