Ollama: Backup Collisions and a Security Fix Converge
Three separate contributors independently fixed the same file-backup collision bug on August 13th, while a critical SSRF-adjacent verification bypass in image pulling was patched and merged. Structured output handling and MLX platform work also saw heavy, connected activity.
Duration: PT2M24S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-08-13T13:01:14Z
- Audio duration: PT2M24S
Transcript excerpt
This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.
Good day, and welcome to your Ollama briefing for August 13th, 2026.
The clearest signal today: three different developers - deepshekhardas, and vatsalgargg - all independently fixed the exact same bug within hours of each other. The `WriteWithBackup` function used a one-second-resolution timestamp for naming backup files, so rapid successive writes to the same file silently…
Second theme: structured output reliability during model "thinking." ekalinin shipped two related fixes - PR 17706 and 17705 - ensuring that when a model reasons before answering, the JSON format constraint doesn't get applied too early and break the response. A related fix, PR 17708, stops the generate endpoint…
Third, security and correctness in the pull pipeline. PR 15504, merged from vigneshakaviki, closes a real vulnerability: when a manifest's config and layer shared the same digest, hash verification could be silently skipped, opening a path for a rogue registry to smuggle unverified content to disk. That's now fixed…
On the platform side, Daniel Hiltgen landed a cluster of MLX work - CUDA backend enablement, a fix to avoid pulling MLX models on systems that can't run them, plus routine llama-dot-cpp and…
W…
Nearby episodes from Ollama
- Weekly Recap - Model Compatibility and the Push Into Coding Agents
- Silent Failures Get Louder
- Qwen Gets Serious About Coding Agents
- Launch Ecosystem Expands, Parser Edge Cases Get Squashed
- Fixing the Defaults That Silently Hurt You
- Model Coverage and Tool-Call Reliability
- Vision Support Lands in the MLX Runner
- Weekly Recap - Vision Runners and Robustness Cleanup