Shannon: One Path for Custom Endpoints
A single merged change simplifies how Shannon talks to OpenAI and other providers, dropping an older gateway format in favor of the Responses API and repositioning custom base URLs as a general-purpose override rather than a special case.
Duration: PT2M24S
Episode overview
This episode is a short developer briefing from Shannon.
It explains recent repository work in plain language.
- Show: Shannon
- Published: 2026-09-04T06:04:54Z
- 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 morning. It's September 4th, 2026, and this is Shannon.
Today's focus is a single, well-contained change with outsized implications for how the system connects to language model providers.
Pull request 445, from ezl-keygraph, does two things that matter together. First, it drops support for the older OpenAI chat-completions gateway format, standardizing on the newer Responses API only. Second, it reframes the custom base URL setting — previously understood as an OpenAI-specific option — as a universal…
Why this matters: maintaining two request formats for the same provider is a common source of subtle bugs, especially around error handling and response parsing. By committing to Responses only, the codebase sheds that duplication. And by generalizing the base URL override, the same configuration now applies…
The commit history backs this up directly. Commit 4b8131f shows the same PR broken into its logical pieces: drop the old chat-completions path, then two documentation passes — one reframing the base URL as universal, and one adding an example showing the optional base URL for "any other provider" setups. That's a…
One thing to remember: if you're running against the old…