Move ProxyEndpoint into the neutral upstream module so error observations no longer depend on account types. Keep proxy matching in the account adapter and verify host/port matching, redaction, and the owner dependency graph.
Co-authored-by: Codex <noreply@openai.com>
Migrate the session binding before requesting a full client replay when safe native continuation recovery exceeds the configured wait limit. Remove stale capacity retry headers from client responses while preserving upstream diagnostics and account rotation limits.
Co-authored-by: Codex <noreply@openai.com>
Reserve a connection attempt for another eligible account when a new WebSocket request repeatedly fails before sending. Use the same rotation budget as HTTP while preserving session and replay restrictions.
Co-authored-by: Codex <noreply@openai.com>
Restore the deployment prefix on root-relative Location response headers in the Caddy example so Realtime follow-up requests stay under the configured route. Preserve external URLs and existing Cookie path rewriting.
Co-authored-by: Codex <noreply@openai.com>
* feat(frontend): support deployments under a URL prefix
Co-authored-by: Codex <noreply@openai.com>
* fix(frontend): configure URL prefixes through shared env
Preserve login redirects under a deployment prefix and use the manifest start URL as its identity. Load the shared frontend env file through Vite and align Docker and deployment documentation.
Co-authored-by: Codex <noreply@openai.com>
---------
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
* fix(openai): classify status-less stream invalid_request_error as a request error
Codex can reject a request with an in-stream `error` frame that carries
`error.type=invalid_request_error` but no HTTP status. That frame fell
through to the "unavailable" branch, so a non-streaming caller received
502 upstream_unavailable for its own invalid request.
Constraint: Capacity, flex and turn rate limit rejections keep their existing categories
Rejected: Treat every status-less stream failure as a request error | would hide real upstream outages
Confidence: high
Scope-risk: narrow
Tested: Rust 1.99.1 fmt --check; provider-openai transport::diagnostics + provider::contract (348 passed); clippy -p provider-openai --all-targets -D warnings; the new case fails without the classification change
Not-tested: End-to-end request against a live upstream on this base
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
* test(openai): cover status-less stream errors through the provider
Keep the stream failure constructor private to the crate. Exercise the original SSE error through the provider so the test covers classification, retry behavior, and the frame delivered to clients.
Co-authored-by: Codex <noreply@openai.com>
---------
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Update the pinned frontend builder image for the Node.js security release
and align the development guide with the image version.
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Rank quota scheduling strictly by configured reset windows and persist its settings independently from smart scheduling.
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Resolve explicit Claude session identities before account selection and
revalidate them after attempt middleware using the existing affinity policy.
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Apply the global request profile to account catalog discovery, warmup,
and diagnostic routing. Freeze each operation's identity through the
existing provider profile resolver.
Refresh verified offline releases and centralize their version, artifact,
and environment defaults. Cover ZIP64 MSIX metadata and keep release
update tests independent of the current bundled versions.
Fixes#703
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
* feat(settings): remove request history retention cap
Allow positive retention days beyond 31 throughout the form, domain rules,
and database constraint. Keep the existing default and saved settings.
Bound cleanup and account usage queries by stored data so large retention
windows remain safe without expanding empty history.
Co-authored-by: Codex <noreply@openai.com>
* test(admin): align retention policy coverage with positive windows
Accept long request retention windows while preserving zero rejection
for each target and independent event and audit retention settings.
Co-authored-by: Codex <noreply@openai.com>
---------
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Make event delivery respond to downstream cancellation and pool shutdown. Return or discard sockets before publishing terminal events while preserving normal connection reuse.
Co-authored-by: Codex <noreply@openai.com>
Match independently scheduled hour and minute pairs while claiming each local minute once. Keep the existing string setting and database constraint, and edit the rules as tags using UI v0.11.0.
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Use the last successful catalog for the same account, credential revision, and plan until a refresh replaces it. Apply the same exclusions in memory and in PostgreSQL while preserving explicit allowlists and accounts with no catalog.
Co-authored-by: Codex <noreply@openai.com>
Use UI v0.10.1 so built-in presets tint pages and cards while preserving explicit background and semantic color settings.
Co-authored-by: Codex <noreply@openai.com>
* fix(egress): honor deployment proxies for unbound outbound requests
Prefer the proxy selected for the request. Otherwise apply the deployment proxy and bypass rules consistently to HTTP and WebSocket connections, including background downloads. Keep explicit direct probes separate.
Co-authored-by: Codex <noreply@openai.com>
* fix(openai): preserve WebSocket dial errors
Recover the OS error or I/O kind from Hyper without retaining the endpoint. Keep refused connections distinguishable for both direct and proxied WebSocket dials.
Co-authored-by: Codex <noreply@openai.com>
---------
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
* fix(update): unlock official plugin dirs before rename on macOS
macOS rejects renaming a 0555 official plugin directory during backup
and rollback, while Linux only needs the parent writable.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(update): preserve directory permissions during macOS moves
Fall back to the existing permission-preserving copy path when macOS
rejects a directory rename. Keep chmod limited to source cleanup so
backups and restored releases retain their original directory modes.
Cover read-only staged assets, rollback, failed replacement, and
cancelled or rejected rollback with permission assertions.
Co-authored-by: Codex <noreply@openai.com>
* fix(update): handle read-only directory moves across Unix platforms
Linux also requires source directory write permission when rename moves
it between parents. Apply the same copy fallback to Unix while retaining
the cross-platform permission regression test.
Co-authored-by: Codex <noreply@openai.com>
---------
Co-authored-by: CzsGit <czsde@outlook.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Use each upstream handshake's message-size limit to select HTTP before
uploading oversized response.create frames. Remove fixed thresholds and
thread size hysteresis while preserving continuation replay and cleanup.
Co-authored-by: Zyy <zyycn@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Separate protocol identifiers from Provider identities and centralize shared
request facts and observations in gateway-protocol. Keep endpoint models
and private continuation state owned by Providers.
Reorganize the architecture documentation and clarify the shutdown guarantees
for buffered charges and observations.
Validated with workspace formatting, strict Clippy, 3013 affected-module
tests and 16 architecture checks. One reference-binary TLS test is ignored.
Co-authored-by: Codex <noreply@openai.com>