Skip to content

Fix/realtime sticky#17

Closed
vi70x3 wants to merge 13 commits into
mainfrom
fix/realtime-sticky
Closed

Fix/realtime sticky#17
vi70x3 wants to merge 13 commits into
mainfrom
fix/realtime-sticky

Conversation

@vi70x3

@vi70x3 vi70x3 commented Jun 5, 2026

Copy link
Copy Markdown
Collaborator

No description provided.

vi70x3 added 13 commits June 5, 2026 14:26
- Change activeRequests from Map to Set to allow concurrent requests from same session
- Add stale active request cleanup with 10-minute TTL
- Cache owl-alpha model ID to avoid repeated DB lookups
- Fix active request iteration to use Set-compatible syntax
- Remove package-lock.json (npm lockfile)
- Add packageManager field to package.json
- Create .npmrc with pnpm configuration
BUG-05: Abort upstream provider stream on stall detection by breaking
the for-await loop and calling gen.return() when the keepalive timer
detects MAX_STREAM_STALL_MS has elapsed without data.

BUG-06: Fix cooldown guard to use the actual routable fallback chain
(fallback_config JOIN models) instead of all enabled models, ensuring
transient cooldowns only skip models that would actually be routed to.

BUG-10: Remove double semicolon in proxy.ts.

Also adds SSE keep-alive comments during idle periods, transient model
cooldown injection before retry loops, and LongCat sticky session
cooldown support in balanced routing mode.
…, TTL refresh, collapsible pools, doc paths, cleanup
… pre-stream, cooldown gating, timer cleanup, a11y, log clarity
… pre-stream, cooldown gating, timer cleanup, a11y, log clarity, test fixes
@vi70x3 vi70x3 closed this Jun 5, 2026
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