Add OpenAI Responses API (/v1/responses) endpoint

Translate Responses API requests (input→messages, instructions→system, max_output_tokens→max_tokens, text.format→response_format, flat tools→nested {function:{…}}) to upstream chat/completions, then convert responses back to Responses format (streaming SSE event lifecycle + non-streaming JSON). Verified against OpenAI migration guide and Python SDK Response model: - Echo back required fields parallel_tool_calls/tool_choice/tools - Include content:[] in reasoning items, logprobs:[] in output_text parts - Support function_call/function_call_output multi-turn input items - Map usage fields prompt_tokens→input_tokens, completion_tokens→output_tokens
2026-08-20 10:02:11 +08:00
parent c8938cb514
commit 2517b4f730
6 changed files with 1576 additions and 4 deletions
@@ -0,0 +1,304 @@
+package server
+
+import (
+	"encoding/json"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"git.misaka.ren/M1saka/zhanlu_proxy/internal/auth"
+	"git.misaka.ren/M1saka/zhanlu_proxy/internal/config"
+	"git.misaka.ren/M1saka/zhanlu_proxy/internal/store"
+)
+
+// TestResponsesNonStreaming verifies POST /v1/responses with stream:false
+// returns a properly formatted Responses API JSON object.
+func TestResponsesNonStreaming(t *testing.T) {
+	upstream, proxy, st := setupTestServer(t)
+	defer upstream.Close()
+	defer proxy.Close()
+
+	creds := auth.Credentials{
+		AccessKey: "AK", SecretKey: "SK", Token: "TOKEN",
+		APIKey: "sk-test-456", ModelBaseURL: upstream.URL, Email: "[email protected]",
+	}
+	if err := st.SaveCredentials(creds); err != nil {
+		t.Fatal(err)
+	}
+
+	body := `{"model":"GLM-4.7","input":"hi","stream":false}`
+	resp, err := http.Post(proxy.URL+"/v1/responses", "application/json", strings.NewReader(body))
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode != 200 {
+		b, _ := io.ReadAll(resp.Body)
+		t.Fatalf("status %d: %s", resp.StatusCode, string(b))
+	}
+	var result map[string]any
+	if err := json.NewDecoder(resp.Body).Decode(&result); err != nil {
+		t.Fatal(err)
+	}
+	if result["object"] != "response" {
+		t.Fatalf("object = %v", result["object"])
+	}
+	if result["status"] != "completed" {
+		t.Fatalf("status = %v", result["status"])
+	}
+	output, ok := result["output"].([]any)
+	if !ok || len(output) == 0 {
+		t.Fatalf("output = %v", result["output"])
+	}
+	msg, ok := output[0].(map[string]any)
+	if !ok {
+		t.Fatalf("output[0] type = %T", output[0])
+	}
+	if msg["type"] != "message" {
+		t.Fatalf("output[0] type = %v", msg["type"])
+	}
+	if msg["role"] != "assistant" {
+		t.Fatalf("output[0] role = %v", msg["role"])
+	}
+	content, ok := msg["content"].([]any)
+	if !ok || len(content) == 0 {
+		t.Fatalf("content = %v", msg["content"])
+	}
+	part, ok := content[0].(map[string]any)
+	if !ok {
+		t.Fatalf("content[0] type = %T", content[0])
+	}
+	if part["type"] != "output_text" {
+		t.Fatalf("content[0] type = %v", part["type"])
+	}
+	if part["text"] != "hello" {
+		t.Fatalf("content[0] text = %v", part["text"])
+	}
+	// Verify required SDK fields are present.
+	if _, ok := result["parallel_tool_calls"]; !ok {
+		t.Fatal("missing parallel_tool_calls in response")
+	}
+	if _, ok := result["tool_choice"]; !ok {
+		t.Fatal("missing tool_choice in response")
+	}
+	if _, ok := result["tools"]; !ok {
+		t.Fatal("missing tools in response")
+	}
+	// Verify output_text has logprobs field.
+	if _, ok := part["logprobs"]; !ok {
+		t.Fatal("missing logprobs in output_text content part")
+	}
+	usage, ok := result["usage"].(map[string]any)
+	if !ok {
+		t.Fatalf("usage = %v", result["usage"])
+	}
+	if totNum(usage["input_tokens"]) != 10 {
+		t.Fatalf("input_tokens = %v", usage["input_tokens"])
+	}
+	if totNum(usage["output_tokens"]) != 20 {
+		t.Fatalf("output_tokens = %v", usage["output_tokens"])
+	}
+	if totNum(usage["total_tokens"]) != 30 {
+		t.Fatalf("total_tokens = %v", usage["total_tokens"])
+	}
+}
+
+// TestResponsesStreaming verifies POST /v1/responses with stream:true
+// emits proper Responses API SSE events.
+func TestResponsesStreaming(t *testing.T) {
+	upstream, proxy, st := setupTestServer(t)
+	defer upstream.Close()
+	defer proxy.Close()
+
+	creds := auth.Credentials{
+		AccessKey: "AK", SecretKey: "SK", Token: "TOKEN",
+		APIKey: "sk-test-456", ModelBaseURL: upstream.URL, Email: "[email protected]",
+	}
+	if err := st.SaveCredentials(creds); err != nil {
+		t.Fatal(err)
+	}
+
+	body := `{"model":"GLM-4.7","input":"hi","stream":true}`
+	resp, err := http.Post(proxy.URL+"/v1/responses", "application/json", strings.NewReader(body))
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode != 200 {
+		b, _ := io.ReadAll(resp.Body)
+		t.Fatalf("status %d: %s", resp.StatusCode, string(b))
+	}
+	raw, _ := io.ReadAll(resp.Body)
+	s := string(raw)
+
+	// Check key lifecycle events are present.
+	checks := []string{
+		"event: response.created",
+		"event: response.in_progress",
+		"event: response.output_item.added",
+		"event: response.content_part.added",
+		"event: response.output_text.delta",
+		"event: response.output_text.done",
+		"event: response.content_part.done",
+		"event: response.output_item.done",
+		"event: response.completed",
+	}
+	for _, c := range checks {
+		if !strings.Contains(s, c) {
+			t.Fatalf("missing %q in SSE body:\n%s", c, s)
+		}
+	}
+
+	// Verify the text delta contains "hello".
+	if !strings.Contains(s, `"delta":"hello"`) {
+		t.Fatalf("text delta missing 'hello' in SSE body:\n%s", s)
+	}
+
+	// Verify required SDK fields are present in completed event.
+	if !strings.Contains(s, `"parallel_tool_calls"`) {
+		t.Fatalf("missing parallel_tool_calls in SSE body:\n%s", s)
+	}
+	if !strings.Contains(s, `"tool_choice"`) {
+		t.Fatalf("missing tool_choice in SSE body:\n%s", s)
+	}
+	// Verify logprobs field in output_text.
+	if !strings.Contains(s, `"logprobs"`) {
+		t.Fatalf("missing logprobs in SSE body:\n%s", s)
+	}
+
+	// Verify usage in the completed event has input_tokens.
+	if !strings.Contains(s, `"input_tokens":`) {
+		t.Fatalf("usage missing input_tokens in SSE body:\n%s", s)
+	}
+}
+
+// TestResponsesStringInput verifies the proxy correctly translates a string
+// "input" to a chat-completions "messages" array before forwarding upstream.
+func TestResponsesStringInput(t *testing.T) {
+	var capturedBody []byte
+	upstream := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path == "/chat/completions" {
+			capturedBody, _ = io.ReadAll(r.Body)
+			w.Header().Set("Content-Type", "text/event-stream")
+			_, _ = w.Write([]byte("data: {\"id\":\"x\",\"choices\":[{\"index\":0,\"delta\":{\"content\":\"ok\"},\"finish_reason\":null}]}\n\ndata: {\"id\":\"x\",\"choices\":[{\"index\":0,\"delta\":{},\"finish_reason\":\"stop\"}]}\n\ndata: [DONE]\n\n"))
+			return
+		}
+		http.NotFound(w, r)
+	}))
+	defer upstream.Close()
+
+	st, err := store.Open(filepath.Join(t.TempDir(), "test.db"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer st.Close()
+
+	cfg := config.Config{
+		MobileLoginBaseURL: upstream.URL,
+		MobileModelBaseURL: upstream.URL,
+		UpstreamPath:       "/chat/completions",
+		DBPath:             filepath.Join(t.TempDir(), "zhanlu.db"),
+		PublicKeyPEM:       defaultTestPublicKey,
+		PhonePublicKeyPEM:  defaultTestPublicKey,
+		SM2PrivateKey:      testSM2Key,
+		PluginVersion:      "1.4.2",
+	}
+	h := New(cfg, st)
+	proxy := httptest.NewServer(h)
+	defer proxy.Close()
+
+	creds := auth.Credentials{
+		APIKey: "sk-test-456", ModelBaseURL: upstream.URL, Email: "[email protected]",
+	}
+	_ = st.SaveCredentials(creds)
+
+	body := `{"model":"GLM-4.7","input":"hello world","stream":false}`
+	resp, err := http.Post(proxy.URL+"/v1/responses", "application/json", strings.NewReader(body))
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer resp.Body.Close()
+	io.Copy(io.Discard, resp.Body)
+
+	if len(capturedBody) == 0 {
+		t.Fatal("no upstream request body captured")
+	}
+	var upstreamReq map[string]any
+	if err := json.Unmarshal(capturedBody, &upstreamReq); err != nil {
+		t.Fatal(err)
+	}
+	messages, ok := upstreamReq["messages"].([]any)
+	if !ok || len(messages) != 1 {
+		t.Fatalf("messages = %v", upstreamReq["messages"])
+	}
+	msg, _ := messages[0].(map[string]any)
+	if msg["role"] != "user" || msg["content"] != "hello world" {
+		t.Fatalf("upstream message = %v", msg)
+	}
+}
+
+// TestResponsesInstructions verifies instructions are prepended as a system
+// message in the upstream request.
+func TestResponsesInstructions(t *testing.T) {
+	var capturedBody []byte
+	upstream := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path == "/chat/completions" {
+			capturedBody, _ = io.ReadAll(r.Body)
+			w.Header().Set("Content-Type", "text/event-stream")
+			_, _ = w.Write([]byte("data: {\"id\":\"x\",\"choices\":[{\"index\":0,\"delta\":{\"content\":\"ok\"},\"finish_reason\":null}]}\n\ndata: {\"id\":\"x\",\"choices\":[{\"index\":0,\"delta\":{},\"finish_reason\":\"stop\"}]}\n\ndata: [DONE]\n\n"))
+			return
+		}
+		http.NotFound(w, r)
+	}))
+	defer upstream.Close()
+
+	st, err := store.Open(filepath.Join(t.TempDir(), "test.db"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer st.Close()
+
+	cfg := config.Config{
+		MobileLoginBaseURL: upstream.URL,
+		MobileModelBaseURL: upstream.URL,
+		UpstreamPath:       "/chat/completions",
+		DBPath:             filepath.Join(t.TempDir(), "zhanlu.db"),
+		PublicKeyPEM:       defaultTestPublicKey,
+		PhonePublicKeyPEM:  defaultTestPublicKey,
+		SM2PrivateKey:      testSM2Key,
+		PluginVersion:      "1.4.2",
+	}
+	h := New(cfg, st)
+	proxy := httptest.NewServer(h)
+	defer proxy.Close()
+
+	creds := auth.Credentials{
+		APIKey: "sk-test-456", ModelBaseURL: upstream.URL, Email: "[email protected]",
+	}
+	_ = st.SaveCredentials(creds)
+
+	body := `{"model":"GLM-4.7","instructions":"be concise","input":"hello","stream":false}`
+	resp, err := http.Post(proxy.URL+"/v1/responses", "application/json", strings.NewReader(body))
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer resp.Body.Close()
+	io.Copy(io.Discard, resp.Body)
+
+	if len(capturedBody) == 0 {
+		t.Fatal("no upstream request body captured")
+	}
+	var upstreamReq map[string]any
+	_ = json.Unmarshal(capturedBody, &upstreamReq)
+	messages, _ := upstreamReq["messages"].([]any)
+	if len(messages) != 2 {
+		t.Fatalf("messages = %d items", len(messages))
+	}
+	sys, _ := messages[0].(map[string]any)
+	if sys["role"] != "system" || sys["content"] != "be concise" {
+		t.Fatalf("system message = %v", sys)
+	}
+}