feat: support DeepSeek Responses and configurable timeout

[zijiren233]

No description provided.

[Copilot]

Pull request overview

This PR extends the relay layer to better support OpenAI-compatible “Responses” streaming across providers (notably DeepSeek), introduces a channel-configurable buffering timeout for initial Responses stream events, and adds response-time monitoring. It also standardizes frontend builds to use Corepack + pnpm with lockfile enforcement.

Changes:

• Add configurable responses_first_event_timeout and wire it into the OpenAI Responses streaming handler (plus shared config schema exposure in OpenAI-compatible adaptors).
• Add DeepSeek Responses support gated to deepseek-v4-flash, including model metadata and tests.
• Refine retry-round behavior in relay channel selection and record resp_cost in the monitor plugin, with expanded tests.

Reviewed changes

Copilot reviewed 37 out of 37 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
Dockerfile	Switch frontend build to corepack pnpm with --frozen-lockfile.
.github/workflows/release.yml	Align CI release build with Corepack + pnpm lockfile usage.
core/relay/plugin/monitor/monitor.go	Record response latency (resp_cost) after upstream response handling.
core/relay/plugin/monitor/monitor_test.go	Add tests validating resp_cost is recorded for success/error paths.
core/relay/adaptor/openai/config.go	Add responses_first_event_timeout config + exported schema/options loaders.
core/relay/adaptor/openai/config_test.go	Add tests for default/configured timeout and schema presence.
core/relay/adaptor/openai/response.go	Make initial Responses stream buffering timeout configurable via parameter.
core/relay/adaptor/openai/response_test.go	Refactor tests to avoid global timeout mutation; add config-driven timeout test.
core/relay/adaptor/openai/adaptor.go	Thread configured timeout into streaming Responses path with error wrapping.
core/relay/adaptor/qianfan/config.go	Add schema fields for responses_first_event_timeout and reasoning mapping.
core/relay/adaptor/deepseek/constants.go	Add deepseek-v4-flash model definition and limits/pricing.
core/relay/adaptor/deepseek/adaptor.go	Gate Responses support to deepseek-v4-flash; expose OpenAI config schema.
core/relay/adaptor/deepseek/adaptor_test.go	Add tests for Responses URL support, mode support gating, schema exposure.
core/relay/adaptor/doubao/main.go	Delegate Responses-family handling to embedded OpenAI adaptor; expose config schema.
core/relay/adaptor/openrouter/adaptor.go	Delegate default DoResponse to embedded OpenAI adaptor; expose config schema.
core/relay/adaptor/ali/adaptor.go	Load per-channel Responses streaming options and pass into OpenAI DoResponse.
core/relay/adaptor/azure/main.go	Expose OpenAI config schema in Azure adaptor metadata.
core/relay/adaptor/azure/main_test.go	Add test asserting OpenAI config schema is exposed.
core/relay/adaptor/azure2/main.go	Expose OpenAI config schema in Azure2 adaptor metadata.
core/relay/adaptor/ai360/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/baichuan/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/cloudflare/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/geminiopenai/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/groq/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/jina/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/lingyiwanwu/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/mistral/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/novita/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/sangforaicp/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/stepfun/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/tencent/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/xai/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/relay/adaptor/xunfei/adaptor.go	Expose OpenAI config schema in adaptor metadata.
core/controller/relay-controller.go	Simplify retry state tracking and permission-failure retry budgeting.
core/controller/relay-controller_test.go	Add/expand tests for retry lifecycle decisions and state updates.
core/controller/relay-channel.go	Implement retry “round” reset behavior and designated-channel pinning.
core/controller/relay-channel_test.go	Replace min-error-rate fallback tests with round-based retry selection tests.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.