airkjw/CLIProxyAPI
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(executor): adjust ApplyThinking order and add payload override test

Browse Source 
- Moved `ApplyThinking` logic earlier in `openai_compat_executor` to align with configuration application sequence.
- Added test to verify payload override precedence over Thinking suffix configuration.
This commit is contained in:
Luis Pater
2026-05-04 16:45:25 +08:00
parent a1eba112f3
commit 89d80bfff4
2 changed files with 54 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
internal/runtime/executor/openai_compat_executor.go
+10 -8
View File
@@ -96,6 +96,12 @@ func (e *OpenAICompatExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
originalPayload := originalPayloadSource
originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, opts.Stream)
translated := sdktranslator.TranslateRequest(from, to, baseModel, req.Payload, opts.Stream)
translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String(), e.Identifier())
if err != nil {
return resp, err
}
requestedModel := helps.PayloadRequestedModel(opts, req.Model)
requestPath := helps.PayloadRequestPath(opts)
translated = helps.ApplyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", translated, originalTranslated, requestedModel, requestPath)
@@ -105,11 +111,6 @@ func (e *OpenAICompatExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
}
}
translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String(), e.Identifier())
if err != nil {
return resp, err
}
url := strings.TrimSuffix(baseURL, "/") + endpoint
httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated))
if err != nil {
@@ -199,15 +200,16 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
originalPayload := originalPayloadSource
originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, true)
translated := sdktranslator.TranslateRequest(from, to, baseModel, req.Payload, true)
requestedModel := helps.PayloadRequestedModel(opts, req.Model)
requestPath := helps.PayloadRequestPath(opts)
translated = helps.ApplyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", translated, originalTranslated, requestedModel, requestPath)
translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String(), e.Identifier())
if err != nil {
return nil, err
}
requestedModel := helps.PayloadRequestedModel(opts, req.Model)
requestPath := helps.PayloadRequestPath(opts)
translated = helps.ApplyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", translated, originalTranslated, requestedModel, requestPath)
// Request usage data in the final streaming chunk so that token statistics
// are captured even when the upstream is an OpenAI-compatible provider.
translated, _ = sjson.SetBytes(translated, "stream_options.include_usage", true)