feat(sdk): publish AIFlow headless runtime compatibility line

.github/workflows/pr-test-build.yml

@@ -23,3 +23,9 @@ jobs:
         run: |
           go build -o test-output ./cmd/server
           rm -f test-output
+      - name: Test
+        run: go test ./... -count=1
+      - name: Focused race tests
+        run: go test -race ./sdk/cliproxy ./sdk/translator ./internal/runtime/executor -count=1
+      - name: Vet
+        run: go vet ./...

.github/workflows/release.yaml

@@ -86,10 +86,10 @@ jobs:
           } > "$updated_notes_file"
 
           if gh release view "$GITHUB_REF_NAME" >/dev/null 2>&1; then
-            gh release edit "$GITHUB_REF_NAME" --title "$GITHUB_REF_NAME" --notes-file "$updated_notes_file"
-          else
-            gh release create "$GITHUB_REF_NAME" --title "$GITHUB_REF_NAME" --notes-file "$updated_notes_file"
+            echo "release already exists for immutable tag $GITHUB_REF_NAME" >&2
+            exit 1
           fi
+          gh release create "$GITHUB_REF_NAME" --title "$GITHUB_REF_NAME" --notes-file "$updated_notes_file"
 
   build-hosted:
     name: build ${{ matrix.target }}
@@ -201,36 +201,7 @@ jobs:
             printf '%s\n' "${assets[@]}" >&2
             exit 1
           fi
-          gh release upload "$GITHUB_REF_NAME" "${assets[@]}" --clobber
-      - name: Refresh release checksums
-        continue-on-error: true
-        shell: bash
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          set -euo pipefail
-          tmp_dir="$(mktemp -d)"
-          trap 'rm -rf "$tmp_dir"' EXIT
-          gh release download "$GITHUB_REF_NAME" \
-            --pattern 'CLIProxyAPI_*.tar.gz' \
-            --pattern 'CLIProxyAPI_*.zip' \
-            --dir "$tmp_dir" \
-            --clobber
-          (
-            cd "$tmp_dir"
-            shopt -s nullglob
-            archives=(CLIProxyAPI_*.tar.gz CLIProxyAPI_*.zip)
-            if [[ ${#archives[@]} -eq 0 ]]; then
-              echo "No release archives found"
-              exit 0
-            fi
-            if command -v sha256sum >/dev/null 2>&1; then
-              sha256sum "${archives[@]}" | sort -k2 > checksums.txt
-            else
-              shasum -a 256 "${archives[@]}" | sort -k2 > checksums.txt
-            fi
-          )
-          gh release upload "$GITHUB_REF_NAME" "$tmp_dir/checksums.txt" --clobber
+          gh release upload "$GITHUB_REF_NAME" "${assets[@]}"
 
   build-linux-glibc:
     name: build linux-${{ matrix.goarch }} glibc
@@ -667,4 +638,4 @@ jobs:
             exit 0
           fi
           sha256sum "${archives[@]}" | sort -k2 > checksums.txt
-          gh release upload "$GITHUB_REF_NAME" checksums.txt --clobber
+          gh release upload "$GITHUB_REF_NAME" checksums.txt

.gitlab-ci.yml

@@ -0,0 +1,32 @@
+stages:
+  - verify
+
+variables:
+  GOPROXY: https://goproxy.cn,direct
+  GOFLAGS: -buildvcs=false
+
+default:
+  image: docker.m.daocloud.io/library/golang:1.26-bookworm
+  cache:
+    key: go-mod-${CI_COMMIT_REF_SLUG}
+    paths:
+      - .cache/go-build
+      - .cache/go-mod
+  before_script:
+    - export GOCACHE="$CI_PROJECT_DIR/.cache/go-build"
+    - export GOMODCACHE="$CI_PROJECT_DIR/.cache/go-mod"
+
+test:
+  stage: verify
+  script:
+    - go test ./... -count=1
+
+race-focused:
+  stage: verify
+  script:
+    - go test -race ./sdk/cliproxy ./sdk/translator ./internal/runtime/executor -count=1
+
+vet:
+  stage: verify
+  script:
+    - go vet ./...

internal/runtime/executor/antigravity_executor_credits_test.go

@@ -21,10 +21,17 @@ import (
 )
 
 func resetAntigravityCreditsRetryState() {
-	antigravityCreditsFailureByAuth = sync.Map{}
-	antigravityShortCooldownByAuth = sync.Map{}
-	antigravityCreditsBalanceByAuth = sync.Map{}
-	antigravityCreditsHintRefreshByID = sync.Map{}
+	clearAntigravityTestSyncMap(&antigravityCreditsFailureByAuth)
+	clearAntigravityTestSyncMap(&antigravityShortCooldownByAuth)
+	clearAntigravityTestSyncMap(&antigravityCreditsBalanceByAuth)
+	clearAntigravityTestSyncMap(&antigravityCreditsHintRefreshByID)
+}
+
+func clearAntigravityTestSyncMap(values *sync.Map) {
+	values.Range(func(key, _ any) bool {
+		values.Delete(key)
+		return true
+	})
 }
 
 type fakeAntigravityKVClient struct {

internal/runtime/executor/claude_executor.go

@@ -380,7 +380,7 @@ func (e *ClaudeExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 		ctx,
 		to,
 		responseFormat,
-		req.Model,
+		requestedModel,
 		opts.OriginalRequest,
 		bodyForTranslation,
 		data,
@@ -552,18 +552,22 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 			scanner := bufio.NewScanner(decodedBody)
 			scanner.Buffer(nil, 52_428_800) // 50MB
 			var event bytes.Buffer
+			var param any
 			flushEvent := func() bool {
 				if event.Len() == 0 {
 					return true
 				}
 				cloned := bytes.Clone(event.Bytes())
 				event.Reset()
-				select {
-				case out <- cliproxyexecutor.StreamChunk{Payload: cloned}:
-					return true
-				case <-ctx.Done():
-					return false
+				chunks := sdktranslator.TranslateStream(ctx, to, responseFormat, requestedModel, opts.OriginalRequest, bodyForTranslation, cloned, &param)
+				for i := range chunks {
+					select {
+					case out <- cliproxyexecutor.StreamChunk{Payload: chunks[i]}:
+					case <-ctx.Done():
+						return false
+					}
 				}
+				return true
 			}
 			for scanner.Scan() {
 				line := scanner.Bytes()
@@ -607,7 +611,7 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 				ctx,
 				to,
 				responseFormat,
-				req.Model,
+				requestedModel,
 				opts.OriginalRequest,
 				bodyForTranslation,
 				bytes.Clone(line),

internal/runtime/executor/claude_executor_test.go

@@ -1237,9 +1237,13 @@ func TestClaudeExecutor_ExecuteStreamStripsOpenAIEncryptedThinkingBeforeUpstream
 }
 
 func TestClaudeExecutor_ExecuteStreamDirectPassthroughEmitsCompleteSSEEvents(t *testing.T) {
+	startData := `{"type":"message_start","message":{"id":"msg_1","model":"physical-model"}}`
 	firstData := `{"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"hi"}}`
 	secondData := `{"type":"message_stop"}`
-	upstreamStream := "event: content_block_delta\n" +
+	upstreamStream := "event: message_start\n" +
+		"data: " + startData + "\n" +
+		"\n" +
+		"event: content_block_delta\n" +
 		"data: " + firstData + "\n" +
 		"\n" +
 		"event: message_stop\n" +
@@ -1262,7 +1266,10 @@ func TestClaudeExecutor_ExecuteStreamDirectPassthroughEmitsCompleteSSEEvents(t *
 	result, err := executor.ExecuteStream(context.Background(), auth, cliproxyexecutor.Request{
 		Model:   "claude-3-5-sonnet-20241022",
 		Payload: payload,
-	}, cliproxyexecutor.Options{SourceFormat: sdktranslator.FromString("claude")})
+	}, cliproxyexecutor.Options{
+		SourceFormat: sdktranslator.FromString("claude"),
+		Metadata:     map[string]any{cliproxyexecutor.RequestedModelMetadataKey: "canonical-model"},
+	})
 	if err != nil {
 		t.Fatalf("ExecuteStream() error = %v", err)
 	}
@@ -1276,6 +1283,7 @@ func TestClaudeExecutor_ExecuteStreamDirectPassthroughEmitsCompleteSSEEvents(t *
 	}
 
 	want := []string{
+		"event: message_start\n" + "data: " + strings.Replace(startData, "physical-model", "canonical-model", 1) + "\n\n",
 		"event: content_block_delta\n" + "data: " + firstData + "\n\n",
 		"event: message_stop\n" + "data: " + secondData + "\n\n",
 	}

internal/runtime/executor/codex_executor.go

@@ -986,7 +986,7 @@ func (e *CodexExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 
 		var param any
 		clientCompletedData := applyCodexIdentityExposeResponsePayload(completedData, identityState)
-		out := sdktranslator.TranslateNonStream(ctx, to, responseFormat, req.Model, originalPayload, body, clientCompletedData, &param)
+		out := sdktranslator.TranslateNonStream(ctx, to, responseFormat, requestedModel, originalPayload, body, clientCompletedData, &param)
 		resp = cliproxyexecutor.Response{Payload: out, Headers: httpResp.Header.Clone()}
 		return resp, nil
 	}
@@ -1096,7 +1096,7 @@ func (e *CodexExecutor) executeCompact(ctx context.Context, auth *cliproxyauth.A
 	reporter.EnsurePublished(ctx)
 	var param any
 	clientData := applyCodexIdentityExposeResponsePayload(upstreamData, identityState)
-	out := sdktranslator.TranslateNonStream(ctx, to, responseFormat, req.Model, originalPayload, body, clientData, &param)
+	out := sdktranslator.TranslateNonStream(ctx, to, responseFormat, requestedModel, originalPayload, body, clientData, &param)
 	resp = cliproxyexecutor.Response{Payload: out, Headers: httpResp.Header.Clone()}
 	return resp, nil
 }
@@ -1264,7 +1264,7 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 			}
 
 			translatedLine = applyCodexIdentityExposeResponsePayload(translatedLine, identityState)
-			chunks := sdktranslator.TranslateStream(ctx, to, responseFormat, req.Model, originalPayload, body, translatedLine, &param)
+			chunks := sdktranslator.TranslateStream(ctx, to, responseFormat, requestedModel, originalPayload, body, translatedLine, &param)
 			for i := range chunks {
 				select {
 				case out <- cliproxyexecutor.StreamChunk{Payload: chunks[i]}:

internal/runtime/executor/openai_compat_executor.go

@@ -194,7 +194,7 @@ func (e *OpenAICompatExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
 	reporter.EnsurePublished(ctx)
 	// Translate response back to source format when needed
 	var param any
-	out := sdktranslator.TranslateNonStream(ctx, to, responseFormat, req.Model, opts.OriginalRequest, translated, body, &param)
+	out := sdktranslator.TranslateNonStream(ctx, to, responseFormat, requestedModel, opts.OriginalRequest, translated, body, &param)
 	resp = cliproxyexecutor.Response{Payload: out, Headers: httpResp.Header.Clone()}
 	return resp, nil
 }
@@ -423,7 +423,7 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 			}
 
 			// OpenAI-compatible streams must use SSE data lines.
-			chunks := sdktranslator.TranslateStream(ctx, to, responseFormat, req.Model, opts.OriginalRequest, translated, bytes.Clone(trimmedLine), &param)
+			chunks := sdktranslator.TranslateStream(ctx, to, responseFormat, requestedModel, opts.OriginalRequest, translated, bytes.Clone(trimmedLine), &param)
 			for i := range chunks {
 				select {
 				case out <- cliproxyexecutor.StreamChunk{Payload: chunks[i]}:
@@ -443,7 +443,7 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 			// In case the upstream close the stream without a terminal [DONE] marker.
 			// Feed a synthetic done marker through the translator so pending
 			// response.completed events are still emitted exactly once.
-			chunks := sdktranslator.TranslateStream(ctx, to, responseFormat, req.Model, opts.OriginalRequest, translated, []byte("data: [DONE]"), &param)
+			chunks := sdktranslator.TranslateStream(ctx, to, responseFormat, requestedModel, opts.OriginalRequest, translated, []byte("data: [DONE]"), &param)
 			for i := range chunks {
 				select {
 				case out <- cliproxyexecutor.StreamChunk{Payload: chunks[i]}:

internal/runtime/executor/openai_compat_executor_compact_test.go

@@ -106,6 +106,31 @@ func TestOpenAICompatExecutorPayloadOverrideWinsOverThinkingSuffix(t *testing.T)
 	}
 }
 
+func TestOpenAICompatExecutorNormalizesRequestedModelInResponse(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"id":"chatcmpl_1","object":"chat.completion","model":"physical-model","choices":[]}`))
+	}))
+	defer server.Close()
+
+	executor := NewOpenAICompatExecutor("deepseek", &config.Config{})
+	auth := &cliproxyauth.Auth{Attributes: map[string]string{"base_url": server.URL, "api_key": "test"}}
+	resp, err := executor.Execute(t.Context(), auth, cliproxyexecutor.Request{
+		Model:   "physical-model",
+		Payload: []byte(`{"model":"canonical-model","messages":[]}`),
+	}, cliproxyexecutor.Options{
+		SourceFormat:   sdktranslator.FormatOpenAI,
+		ResponseFormat: sdktranslator.FormatOpenAI,
+		Metadata:       map[string]any{cliproxyexecutor.RequestedModelMetadataKey: "canonical-model"},
+	})
+	if err != nil {
+		t.Fatalf("Execute error: %v", err)
+	}
+	if model := gjson.GetBytes(resp.Payload, "model").String(); model != "canonical-model" {
+		t.Fatalf("response model=%q payload=%s", model, resp.Payload)
+	}
+}
+
 func TestOpenAICompatExecutorImagesGenerationsPassthrough(t *testing.T) {
 	var gotPath string
 	var gotBody []byte