Skip to content

feat(sse): add Dario provider limits adapter - #12843

Closed
andrewmalov wants to merge 6 commits into
cheaper-inference:release/v3.8.50from
andrewmalov:feat/dario-provider-limits
Closed

andrewmalov wants to merge 6 commits into
cheaper-inference:release/v3.8.50from
andrewmalov:feat/dario-provider-limits

Conversation

@andrewmalov

Copy link
Copy Markdown

Summary

  • add a statically allowlisted Dario usage capability for Anthropic-compatible API-key connections
  • normalize fresh Dario 5-hour and 7-day utilization into Provider Limits windows
  • make Provider Limits and quota-aware routing resolve fetchers per connection
  • preserve connection-scoped caches and last-known-good reset-aware snapshots
  • add SSRF, credential-boundary, freshness, pool-selection, visibility, and preflight regression tests

Verification

  • focused Dario/quota tests: 49 passed
  • npm run typecheck:core: passed
  • npm run lint: passed
  • npm run test:vitest: 44 files / 405 tests passed
  • npm run check:cycles: passed
  • npm run build: passed
  • npm run typecheck:noimplicit:core: inherited failures in unchanged open-sse/translator/response/openai-responses/pureHelpers.ts
  • npm run test:unit: base v3.8.50 has 8 unrelated failures; focused Dario suite passed

The live demo deployment and Dario ingress hardening are maintained separately from this upstream code change.

diegosouzapw and others added 3 commits August 26, 2026 14:25
Release v3.8.50 — see CHANGELOG.md for the full entry.
cheaper-inference#11683)

The fragments were written during the pre-flight but the aggregation step was
left uncommitted, so the release PR merged with the ten files still loose in
changelog.d/ and CHANGELOG.md missing their entries. Tagging from that state
would have published v3.8.50 without documenting the discarded upstream call
burning quota (cheaper-inference#11552), the configured search connection being silently ignored
(cheaper-inference#11524), the /v1/models SWR refresh blocking its own stale response (cheaper-inference#11551),
the combo builder's manual model entry (dark since cheaper-inference#8285) and the local CLI's
health view (dark since cheaper-inference#11040) — the entries an operator actually reads.

Runs the aggregator on top of main and deletes the fragments in the same commit,
per changelog.d/README.md.

Co-authored-by: Xiangzhe <bakryun0718@proton.me>
@diegosouzapw

Copy link
Copy Markdown
Collaborator

Thanks for the thorough work here — the SSRF/credential-boundary test coverage is great. I
compared this against the current release tip and Dario was actually shipped as a full embedded
service in v3.8.50 (#8523) via a different path: executor, combo/headroom integration,
dashboard, and a migration numbered 138 already exist there too (which collides with this PR's
own migration 138 — different content, same number). Given that, I don't think we can merge
this as-is, but if the specific "normalize 5-hour/7-day Dario utilization into Provider Limits
windows" piece isn't covered by what's already live, I'd welcome a much smaller follow-up PR
just for that, and you'd keep authorship credit on it.

Triage note: this is the review recommendation — the close itself happens only after the maintainer's per-PR sign-off (and, where a superseding PR is named, after it has landed). Nothing is being closed by this comment.

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.

2 participants