Skip to content

blog: auto-router keeps the prompt cache warm across a session - #811

Open
tin-berri wants to merge 4 commits into
mainfrom
autorouter_aug_improvements
Open

tin-berri wants to merge 4 commits into
mainfrom
autorouter_aug_improvements

Conversation

@tin-berri

@tin-berri tin-berri commented Aug 7, 2026

Copy link
Copy Markdown
Contributor

New blog post leading on one improvement, with the rest as a short round-up. No PR links in the post body, per request.

Headline: auto-router session affinity now pins to a deployment rather than a model group, so a conversation stays on the deployment whose prompt cache is warm instead of bouncing across a fanned group roughly every other turn. Session pins are also scoped per API key now.

Also covered, briefly: the current-UTC-day fix on the cost dashboard, the expired-miss denominator change, and wildcard matching for 1-click presets.

Note: the session affinity change is still open as BerriAI/litellm#36146. This post should not publish until it lands.

Separate from the classifier benchmark post in #805, which stays open.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@vercel

vercel Bot commented Aug 7, 2026

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated (UTC)
litellm Ready Ready Preview Aug 7, 2026 3:54am

Request Review

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@tin-berri tin-berri changed the title blog: 4 Auto-Router improvements (caching, dashboards, presets) blog: auto-router keeps the prompt cache warm across a session Aug 7, 2026
tin-berri and others added 2 commits August 6, 2026 20:49
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant