feat(gemini): 接入 Veo 官方异步生成协议

将 Gemini Veo 视频请求改为 predictLongRunning 提交与 Operation 轮询,支持任务恢复、首尾帧和参考图的官方 Base64 字段映射。

完成后由网关携带 API Key 从受信任的 Gemini 文件域下载视频,再进入现有对象存储转存链路,避免暴露鉴权地址。

验证:env -u AI_GATEWAY_TEST_DATABASE_URL go test ./... -count=1;go vet ./...;真实 Veo 3.1 任务生成并下载 1633544 字节视频。
This commit is contained in:
2026-08-04 16:27:24 +08:00
parent 5c7d6ac9aa
commit 44d5cf2b9d
7 changed files with 951 additions and 1 deletions
@@ -0,0 +1,267 @@
package clients
import (
"context"
"encoding/base64"
"encoding/json"
"net/http"
"net/http/httptest"
"os"
"reflect"
"strings"
"sync/atomic"
"testing"
"time"
"github.com/easyai/easyai-ai-gateway/apps/api/internal/store"
)
func TestGeminiVeoRunUsesOfficialLongRunningProtocol(t *testing.T) {
const operationName = "models/veo-3.1-generate-preview/operations/test-operation"
videoPayload := []byte("test mp4 payload")
var server *httptest.Server
var pollCount atomic.Int32
server = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.Header.Get("x-goog-api-key") != "test-gemini-key" {
t.Errorf("missing Gemini API key header for %s", r.URL.Path)
}
switch {
case r.Method == http.MethodPost && r.URL.Path == "/v1beta/models/veo-3.1-generate-preview:predictLongRunning":
var body map[string]any
if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
t.Fatalf("decode Veo body: %v", err)
}
instances := body["instances"].([]any)
if prompt := instances[0].(map[string]any)["prompt"]; prompt != "a paper boat on a lake" {
t.Errorf("unexpected prompt: %v", prompt)
}
parameters := body["parameters"].(map[string]any)
if parameters["durationSeconds"] != float64(4) || parameters["resolution"] != "720p" || parameters["aspectRatio"] != "16:9" {
t.Errorf("unexpected parameters: %+v", parameters)
}
_ = json.NewEncoder(w).Encode(map[string]any{"name": operationName})
case r.Method == http.MethodGet && r.URL.Path == "/v1beta/"+operationName:
if pollCount.Add(1) == 1 {
_ = json.NewEncoder(w).Encode(map[string]any{"name": operationName, "done": false})
return
}
_ = json.NewEncoder(w).Encode(map[string]any{
"name": operationName,
"done": true,
"response": map[string]any{"generateVideoResponse": map[string]any{"generatedSamples": []any{
map[string]any{"video": map[string]any{"uri": server.URL + "/v1beta/files/generated-video:download", "mimeType": "video/mp4"}},
}}},
})
case r.Method == http.MethodGet && r.URL.Path == "/v1beta/files/generated-video:download":
w.Header().Set("Content-Type", "video/mp4")
_, _ = w.Write(videoPayload)
default:
http.NotFound(w, r)
}
}))
defer server.Close()
var submitted string
polled := 0
response, err := (GeminiClient{}).Run(context.Background(), Request{
Kind: "videos.generations",
ModelType: "video_generate",
Model: "veo-3.1",
Body: map[string]any{
"prompt": "a paper boat on a lake",
"duration": 4,
"resolution": "720p",
"aspect_ratio": "16:9",
},
Candidate: testGeminiVeoCandidate(server.URL),
OnRemoteTaskSubmitted: func(remoteTaskID string, _ map[string]any) error {
submitted = remoteTaskID
return nil
},
OnRemoteTaskPolled: func(remoteTaskID string, _ map[string]any) error {
if remoteTaskID != operationName {
t.Errorf("unexpected polled operation: %s", remoteTaskID)
}
polled++
return nil
},
})
if err != nil {
t.Fatalf("run Gemini Veo: %v", err)
}
if submitted != operationName || polled != 2 || pollCount.Load() != 2 {
t.Fatalf("unexpected async callbacks: submitted=%q polled=%d requests=%d", submitted, polled, pollCount.Load())
}
if response.UpstreamProtocol != ProtocolGeminiVeo || response.RequestID != operationName {
t.Fatalf("unexpected response metadata: %+v", response)
}
data := response.Result["data"].([]any)
item := data[0].(map[string]any)
if got := item["video_bytes"].([]byte); !reflect.DeepEqual(got, videoPayload) {
t.Fatalf("unexpected video payload: %q", got)
}
if _, exposed := item["uri"]; exposed {
t.Fatalf("upstream authenticated URI must not be exposed: %+v", item)
}
}
func TestGeminiVeoRunResumesOperationWithoutSubmittingAgain(t *testing.T) {
const operationName = "models/veo-3.1-generate-preview/operations/resumed"
var postCount atomic.Int32
server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.Method == http.MethodPost {
postCount.Add(1)
http.Error(w, "unexpected submit", http.StatusInternalServerError)
return
}
_ = json.NewEncoder(w).Encode(map[string]any{
"name": operationName,
"done": true,
"response": map[string]any{"generateVideoResponse": map[string]any{"generatedSamples": []any{
map[string]any{"video": map[string]any{"videoBytes": base64.StdEncoding.EncodeToString([]byte("resumed video")), "mimeType": "video/mp4"}},
}}},
})
}))
defer server.Close()
response, err := (GeminiClient{}).Run(context.Background(), Request{
Kind: "videos.generations",
ModelType: "video_generate",
Model: "veo-3.1",
Body: map[string]any{"prompt": "ignored on resume"},
Candidate: testGeminiVeoCandidate(server.URL),
RemoteTaskID: operationName,
})
if err != nil {
t.Fatalf("resume Gemini Veo: %v", err)
}
if postCount.Load() != 0 || response.Result["id"] != operationName {
t.Fatalf("resume submitted a new operation or returned wrong ID: posts=%d result=%+v", postCount.Load(), response.Result)
}
}
func TestGeminiVeoBodyMapsImageAndParameters(t *testing.T) {
first := "data:image/png;base64," + base64.StdEncoding.EncodeToString([]byte("first"))
last := "data:image/jpeg;base64," + base64.StdEncoding.EncodeToString([]byte("last"))
reference := "data:image/webp;base64," + base64.StdEncoding.EncodeToString([]byte("reference"))
body, err := geminiVeoBody(Request{Body: map[string]any{
"prompt": "animate",
"first_frame": first,
"last_frame": last,
"reference_images": []any{reference},
"duration": 8,
"resolution": "2160p",
"aspect_ratio": "9:16",
"negative_prompt": "text",
"person_generation": "allow_adult",
"enhance_prompt": false,
}})
if err != nil {
t.Fatalf("build Gemini Veo body: %v", err)
}
instance := body["instances"].([]any)[0].(map[string]any)
image := instance["image"].(map[string]any)
lastFrame := instance["lastFrame"].(map[string]any)
references := instance["referenceImages"].([]any)
if image["bytesBase64Encoded"] != base64.StdEncoding.EncodeToString([]byte("first")) || image["mimeType"] != "image/png" {
t.Fatalf("unexpected first frame: %+v", image)
}
if lastFrame["mimeType"] != "image/jpeg" || len(references) != 1 {
t.Fatalf("unexpected last/reference images: last=%+v references=%+v", lastFrame, references)
}
parameters := body["parameters"].(map[string]any)
if parameters["resolution"] != "4k" || parameters["durationSeconds"] != 8 || parameters["enhancePrompt"] != false {
t.Fatalf("unexpected Gemini Veo parameters: %+v", parameters)
}
}
func TestGeminiVeoRejectsInvalidDurationBeforeSubmission(t *testing.T) {
_, err := geminiVeoBody(Request{Body: map[string]any{"prompt": "test", "duration": 5}})
if err == nil || ErrorCode(err) != "invalid_parameter" || ErrorParam(err) != "duration" || !strings.Contains(err.Error(), "4, 6, or 8") {
t.Fatalf("unexpected invalid duration error: %v", err)
}
}
func TestGeminiVeoDownloadURLOnlyTrustsProviderAndGoogleAPIs(t *testing.T) {
for _, test := range []struct {
uri string
trusted bool
}{
{uri: "https://gemini-proxy.example.com/v1beta/files/video", trusted: true},
{uri: "https://generativelanguage.googleapis.com/v1beta/files/video", trusted: true},
{uri: "https://storage.googleapis.com/signed/video", trusted: false},
{uri: "https://attacker.example.com/video", trusted: false},
} {
_, trusted, err := geminiVeoDownloadURL("https://gemini-proxy.example.com/v1beta", test.uri)
if err != nil {
t.Fatalf("parse %s: %v", test.uri, err)
}
if trusted != test.trusted {
t.Fatalf("unexpected trust for %s: got %t want %t", test.uri, trusted, test.trusted)
}
}
}
func TestGeminiVeoLive(t *testing.T) {
apiKey := strings.TrimSpace(os.Getenv("GEMINI_VEO_LIVE_API_KEY"))
if apiKey == "" {
t.Skip("GEMINI_VEO_LIVE_API_KEY is not configured")
}
baseURL := strings.TrimSpace(os.Getenv("GEMINI_VEO_LIVE_BASE_URL"))
model := strings.TrimSpace(os.Getenv("GEMINI_VEO_LIVE_MODEL"))
if model == "" {
model = "veo-3.1-generate-preview"
}
ctx, cancel := context.WithTimeout(context.Background(), 12*time.Minute)
defer cancel()
operationName := strings.TrimSpace(os.Getenv("GEMINI_VEO_LIVE_OPERATION"))
response, err := (GeminiClient{}).Run(ctx, Request{
Kind: "videos.generations",
ModelType: "video_generate",
Model: model,
Body: map[string]any{
"prompt": "A small red paper boat gently floating on calm water, static camera, natural daylight.",
"duration": 4,
"resolution": "720p",
"aspect_ratio": "16:9",
},
Candidate: store.RuntimeModelCandidate{
Provider: "gemini",
BaseURL: baseURL,
ProviderModelName: model,
Credentials: map[string]any{"apiKey": apiKey},
PlatformConfig: map[string]any{"pollIntervalMs": 10000, "pollTimeoutMs": 660000},
},
RemoteTaskID: operationName,
OnRemoteTaskSubmitted: func(remoteTaskID string, _ map[string]any) error {
t.Logf("live Gemini Veo submitted operation %s", remoteTaskID)
return nil
},
})
if err != nil {
t.Fatalf("live Gemini Veo request failed: %v", err)
}
data, _ := response.Result["data"].([]any)
if len(data) != 1 {
t.Fatalf("live Gemini Veo returned %d videos", len(data))
}
item, _ := data[0].(map[string]any)
payload, _ := item["video_bytes"].([]byte)
if len(payload) < 1024 {
t.Fatalf("live Gemini Veo video is unexpectedly small: %d bytes", len(payload))
}
if mimeType := strings.TrimSpace(stringFromAny(item["mime_type"])); !strings.HasPrefix(mimeType, "video/") {
t.Fatalf("live Gemini Veo returned unexpected MIME type: %q", mimeType)
}
t.Logf("live Gemini Veo accepted operation %s and downloaded %d video bytes", response.RequestID, len(payload))
}
func testGeminiVeoCandidate(baseURL string) store.RuntimeModelCandidate {
return store.RuntimeModelCandidate{
Provider: "gemini",
BaseURL: baseURL,
ProviderModelName: "veo-3.1-generate-preview",
Credentials: map[string]any{"apiKey": "test-gemini-key"},
PlatformConfig: map[string]any{"pollIntervalMs": 1, "pollTimeoutMs": 1000},
}
}