From 89d80bfff49425a49e23052a0db481f6440bfbfe Mon Sep 17 00:00:00 2001
From: Luis Pater <webmaster@idotorg.org>
Date: Mon, 4 May 2026 16:45:25 +0800
Subject: [PATCH] fix(executor): adjust ApplyThinking order and add payload
 override test

- 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.
---
 .../executor/openai_compat_executor.go        | 18 ++++----
 .../openai_compat_executor_compact_test.go    | 44 +++++++++++++++++++
 2 files changed, 54 insertions(+), 8 deletions(-)

diff --git a/internal/runtime/executor/openai_compat_executor.go b/internal/runtime/executor/openai_compat_executor.go
index 4e44a7ae..63be2d3c 100644
--- a/internal/runtime/executor/openai_compat_executor.go
+++ b/internal/runtime/executor/openai_compat_executor.go
@@ -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)
diff --git a/internal/runtime/executor/openai_compat_executor_compact_test.go b/internal/runtime/executor/openai_compat_executor_compact_test.go
index fe281262..ac9d9b32 100644
--- a/internal/runtime/executor/openai_compat_executor_compact_test.go
+++ b/internal/runtime/executor/openai_compat_executor_compact_test.go
@@ -56,3 +56,47 @@ func TestOpenAICompatExecutorCompactPassthrough(t *testing.T) {
 		t.Fatalf("payload = %s", string(resp.Payload))
 	}
 }
+
+func TestOpenAICompatExecutorPayloadOverrideWinsOverThinkingSuffix(t *testing.T) {
+	var gotBody []byte
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		body, _ := io.ReadAll(r.Body)
+		gotBody = body
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"id":"chatcmpl_1","object":"chat.completion","choices":[{"index":0,"message":{"role":"assistant","content":"ok"},"finish_reason":"stop"}],"usage":{"prompt_tokens":1,"completion_tokens":1,"total_tokens":2}}`))
+	}))
+	defer server.Close()
+
+	executor := NewOpenAICompatExecutor("openai-compatibility", &config.Config{
+		Payload: config.PayloadConfig{
+			Override: []config.PayloadRule{
+				{
+					Models: []config.PayloadModelRule{
+						{Name: "custom-openai", Protocol: "openai"},
+					},
+					Params: map[string]any{
+						"reasoning_effort": "low",
+					},
+				},
+			},
+		},
+	})
+	auth := &cliproxyauth.Auth{Attributes: map[string]string{
+		"base_url": server.URL + "/v1",
+		"api_key":  "test",
+	}}
+	payload := []byte(`{"model":"custom-openai(high)","messages":[{"role":"user","content":"hi"}]}`)
+	_, err := executor.Execute(context.Background(), auth, cliproxyexecutor.Request{
+		Model:   "custom-openai(high)",
+		Payload: payload,
+	}, cliproxyexecutor.Options{
+		SourceFormat: sdktranslator.FromString("openai"),
+		Stream:       false,
+	})
+	if err != nil {
+		t.Fatalf("Execute error: %v", err)
+	}
+	if got := gjson.GetBytes(gotBody, "reasoning_effort").String(); got != "low" {
+		t.Fatalf("reasoning_effort = %q, want %q; body=%s", got, "low", string(gotBody))
+	}
+}