Kevin CodexandClaude Fable 5 8ba6ed42b5 feat(providerio): rate-limit-appropriate retry schedule ...

Raise retry attempts 3 -> 6 and replace the attempt*400ms linear backoff
with exponential 2s, 4s, 8s, 16s, 30s (capped) when the server supplies
no Retry-After. Rate-limit windows are measured in seconds; retrying a
429 after 400ms almost always burned the attempt while still limited.
Worst-case a headless run now survives a ~60s rate-limit window instead
of ~1.2s. Retry-After still wins when present, still capped at 30s, and
the only-replay-unaccepted-requests policy (429/503/529) is unchanged.

Co-Authored-By: Claude Fable 5 <[email protected]>

2026-07-05 11:43:48 +08:00

..

anthropic

feat(anthropic): cache conversation messages, not just system+tools

2026-07-05 11:43:48 +08:00

gemini

fix(gemini): strip unsupported JSON Schema fields from tool declarations (#374)

2026-07-02 20:20:25 +08:00

openai

fix: agent runtime resilience — stream stalls, weak-model tool args, and sub-agent reliability (#349)

2026-07-01 09:15:08 +05:30

providerio

feat(providerio): rate-limit-appropriate retry schedule

2026-07-05 11:43:48 +08:00

factory_test.go

fix(auth): propagate credentials to every provider-build surface and pin children to the live provider (#366)

2026-07-02 12:21:27 +05:30

factory.go

fix(auth): propagate credentials to every provider-build surface and pin children to the live provider (#366)

2026-07-02 12:21:27 +05:30