Ollama: Consistency Fixes for Tool Schemas and Image Input
Today's changes center on output consistency — fixing non-deterministic tool schema rendering and unpredictable image orientation — alongside new API continuation support and two README additions for the community integrations list.
Duration: PT2M25S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-09-14T13:01:34Z
- Audio duration: PT2M25S
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 September 14th, and here's your Ollama development briefing.
The throughline today is consistency. Two separate bugs shared the same root problem: the same input producing different output depending on when you ran it.
First, in PR 18433, developer L4XB fixed a tool-rendering bug where Go's map iteration order caused a tool's extra schema keys — things like "defs" alongside "required" — to shuffle position from one request to the next. It only showed up when an object carried two or more of these extra keys, which is why it…
Second, PR 18432 from QIU-Guanzong tackles a related consistency issue on the image side. JPEG files carrying EXIF rotation data — orientations two through eight — were being passed through without normalization, so the same photo could render sideways or upside down depending on the runner. The fix normalizes those…
Together, these two fixes are a good reminder: nondeterminism bugs — whether from map ordering or missing metadata handling — are easy to miss in testing and expensive in production, especially for anything user-facing or reproducibility-sensitive.
On the feature side, ParthSareen opened PR 18434 adding support for…
Nearby episodes from Ollama
- Weekly Recap - Memory Discipline and Tool-Calling Cleanup
- Precision Fixes and Cleanup
- Silent Failures Get Loud
- Taming Memory in the MLX Runner
- Tool-Call Parsing Gets a Hardening Pass
- Codex Integration Hardening and Streaming Reliability
- Tightening Up the OpenAI Compatibility Layer
- Closing the Gaps Between What You Configure and What Actually Runs