Ollama: Codex Integration Hardening and Streaming Reliability
Nearly half of today's activity centers on tightening the ChatGPT and Codex integration inside the Ollama app, alongside a cluster of fixes to keep OpenAI-compatible streaming and tool-calling correct under edge cases. Two data-race fixes and a rethink of MLX memory management round out a day focused on reliability over new features.
Duration: PT3M
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-09-09T13:01:15Z
- Audio duration: PT3M
Transcript excerpt
This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.
This is Ollama, your daily developer briefing for September 9th, 2026.
Today's dominant story is the Codex and ChatGPT integration inside the Ollama app, which saw more than ten related pull requests land or open in a single day. The through-line: making that connection feel trustworthy and stay correct as configuration changes.
On the user-facing side, Eva H shipped a first-use intro when connecting ChatGPT, PR 18321, plus matching status copy with Claude in PR 18325, and defaulting Codex mode on launch in PR 18323. Underneath that polish, Parth Sareen fixed a real correctness gap in PR 18247: changing the Ollama host could leave ChatGPT…
The second theme is protocol correctness on the OpenAI-compatible surface. Tool-search call IDs needed a "tsc underscore" prefix instead of "ts underscore" to pass validation when switching models, fixed in PR 18296. Compaction logic got a retry path for context overflow in PR 18324, trimming the oldest removable…
Two infrastructure fixes deserve a mention: Daniel Hiltgen's PR 18319 closed data races in progress rendering and the scheduler's log values, both triggered by concurrent access to shared state. And Jesse Gross opened PR 18327, a broader…
O…
Nearby episodes from Ollama
- Tightening Up the OpenAI Compatibility Layer
- Closing the Gaps Between What You Configure and What Actually Runs
- Weekly Recap - MLX Hardens as a First-Class Engine
- Context Handling Gets Serious
- Codex Integration Push and Schema Hardening
- Structured Output Meets Speculative Decoding
- Hardening the MLX Engine and Fixing What Feeds the Scheduler
- MLX Grows Up, Model Intent Gets Respected