add text_to_speech & change field name

.github/workflows/ci.yml

@@ -17,12 +17,16 @@ jobs:
     - name: checkout repository
       uses: actions/checkout@v4
 
-    - name: Build
+    - name: Replace secrets for auth.yaml
       run: |
+        sed -i "s/OPENAI_API_KEY/${{ secrets.OPENAI_API_KEY }}/g" ./conf/auth.yaml
+        sed -i "s/QWEN_API_KEY/${{ secrets.QWEN_API_KEY }}/g" ./conf/auth.yaml
         sed -i "s/OSS_ACCESS_ENDPOINT/${{ secrets.OSS_ACCESS_ENDPOINT }}/g" ./conf/auth.yaml
         sed -i "s/OSS_ACCESS_KEY_ID/${{ secrets.OSS_ACCESS_KEY_ID }}/g" ./conf/auth.yaml
         sed -i "s/OSS_ACCESS_KEY_SECRET/${{ secrets.OSS_ACCESS_KEY_SECRET }}/g" ./conf/auth.yaml
-        sed -i "s/OPENAI_API_KEY/${{ secrets.OPENAI_API_KEY }}/g" ./conf/auth.yaml
+
+    - name: Build
+      run: |
         bazel build --action_env=CC=clang-14 //...
         mkdir -p output/bin output/conf
         mv ./bazel-bin/src/ielts_ai output/bin/

WORKSPACE

@@ -17,16 +17,16 @@ http_archive(
 http_archive(
     name = "com_github_d7ead_liboai",
     build_file = "//bazel:liboai.BUILD",
-    sha256 = "71d1c7fe55fdec465e3f1ec376104c1bac92e014f5984a9024749a73e6a312ff",
-    strip_prefix = "liboai-1.0.7/liboai",
-    urls = ["https://github.com/pkusunjy/liboai/archive/v1.0.7.tar.gz"],
+    sha256 = "83108eb34c71b6bf0d18d8b7071a76a6c1609fea60f32afedaf7f69c95f0b827",
+    strip_prefix = "liboai-1.0.11/liboai",
+    urls = ["https://github.com/pkusunjy/liboai/archive/v1.0.11.tar.gz"],
 )
 
 http_archive(
     name = "com_github_pkusunjy_openai_server_proto",
-    sha256 = "5c6cf3341f68cc9ac54534363fe2e0c923bfa22c1583a476934cd98d2e2f2540",
-    strip_prefix = "openai-server-proto-1.1.6",
-    urls = ["https://github.com/pkusunjy/openai-server-proto/archive/v1.1.6.tar.gz"],
+    sha256 = "d4d1e58c1f72e4ac93068e9f4bbc381d9a5ea1ddc37407635d175138b06ee92c",
+    strip_prefix = "openai-server-proto-1.1.13",
+    urls = ["https://github.com/pkusunjy/openai-server-proto/archive/v1.1.13.tar.gz"],
 )
 
 http_archive(

conf/auth.yaml

@@ -1,4 +1,10 @@
+# ai api keys
 openai_api_key: OPENAI_API_KEY
+qwen_api_key: QWEN_API_KEY
+# ai url roots
+openai_url_root: https://mikiai-tokyo-onctmshaci.ap-northeast-1.fcapp.run/v1
+qwen_url_root: https://dashscope.aliyuncs.com/compatible-mode/v1
+# aliyun object storage service
 oss_endpoint: OSS_ACCESS_ENDPOINT
 oss_access_key_id: OSS_ACCESS_KEY_ID
 oss_access_key_secret: OSS_ACCESS_KEY_SECRET

src/ielts_ai.cpp

@@ -45,14 +45,6 @@ int32_t main(int32_t argc, char* argv[]) {
     return 0;
   }
 
-  std::string openai_api_key = token_instance.get_token_by_name("openai_api_key");
-  LOG(INFO) << "openai_api_key: " << openai_api_key;
-
-  if (!liboai::Authorization::Authorizer().SetKey(openai_api_key)) {
-    LOG(WARNING) << "OPENAI_API_KEY not found, server quit";
-    return 0;
-  }
-
   grpc::EnableDefaultHealthCheckService(true);
   grpc::reflection::InitProtoReflectionServerBuilderPlugin();
 

src/service/BUILD

@@ -25,6 +25,7 @@ cc_library(
         "//src/plugin:log_sink",
         "//src/plugin:oss",
         "//src/plugin:prompt",
+        "//src/plugin:token",
         "@boost.scope_exit",
         "@com_github_d7ead_liboai//:oai",
         "@com_github_grpc_grpc//:grpc++",

src/service/ielts_ai.cpp

@@ -3,18 +3,26 @@
 namespace chat_completion {
 
 int32_t IeltsAI::initialize() {
+  auto& token_instance = plugin::TokenFactory::instance();
+  // audio auth
+  _audio_auth.SetKey(token_instance.get_token_by_name("openai_api_key"));
+  _chat_auth.SetKey(token_instance.get_token_by_name("qwen_api_key"));
   // audio
   _audio = std::make_unique<liboai::Audio>();
   if (_audio == nullptr) {
     LOG(WARNING) << "liboai audio ctor failed";
     return -1;
   }
+  _audio->SetAuth(_audio_auth);
+  _audio->UpdateOpenAIRoot(token_instance.get_token_by_name("openai_url_root"));
   // chat
   _chat_completion = std::make_unique<liboai::ChatCompletion>();
   if (_chat_completion == nullptr) {
     LOG(WARNING) << "liboai completion ctor failed";
     return -1;
   }
+  _chat_completion->SetAuth(_chat_auth);
+  _chat_completion->UpdateOpenAIRoot(token_instance.get_token_by_name("qwen_url_root"));
   // aliyun oss
   _oss = std::make_unique<plugin::OssClient>();
   if (_oss == nullptr || _oss->initialize() != 0) {
@@ -65,12 +73,62 @@ grpc::Status IeltsAI::transcribe_judge(grpc::ServerContext* ctx, const ChatMessa
   resp->set_content(transcribe_res);
   // 4. TODO: delete aliyun oos
   absl::Time step4 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid()
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid()
             << " get from oss cost: " << absl::ToDoubleMilliseconds(step2 - step1)
             << " transcribe cost: " << absl::ToDoubleMilliseconds(step3 - step2)
             << " unlink cost: " << absl::ToDoubleMilliseconds(step4 - step3)
             << ", total cost: " << absl::ToDoubleMilliseconds(step4 - step1);
   return grpc::Status::OK;
 }
 
+grpc::Status IeltsAI::text_to_speech(grpc::ServerContext* ctx, const ChatMessage* req, ChatMessage* resp) {
+  if (_audio == nullptr) {
+    LOG(WARNING) << "text_to_speech not ready";
+    return grpc::Status(grpc::StatusCode::FAILED_PRECONDITION, "text_to_speech not ready");
+  }
+  // 1. call api
+  absl::Time step1 = absl::Now();
+  LOG(INFO) << "received stage:" << req->stage() << " text: " << req->content();
+  std::string voice = req->stage() == 1 ? "alloy" : "nova";
+  std::string input = req->content();
+
+  auto res = _audio->speech("tts-1", voice, input);
+
+  std::string local_filename = absl::StrFormat("text_to_speech_%s_%llu.mp3", req->userid(), req->logid());
+  {
+    std::ofstream ocout(local_filename, std::ios::binary);
+    ocout << res.content;
+    ocout.close();
+  }
+
+  BOOST_SCOPE_EXIT(&req, &local_filename) {
+    // delete audio file on disk
+    if (unlink(local_filename.c_str()) < 0) {
+      char buf[256];
+      strerror_r(errno, buf, 256);
+      LOG(WARNING) << "logid " << req->logid() << "unlink failed file: " << local_filename << ", errno: " << errno
+                   << ", errmsg: " << buf;
+    }
+  }
+  BOOST_SCOPE_EXIT_END
+  // 2. upload to oss
+  absl::Time step2 = absl::Now();
+  if (_oss->put_object(local_filename, local_filename) != 0) {
+    LOG(WARNING) << "OssClient put_object failed";
+    return grpc::Status(grpc::StatusCode::FAILED_PRECONDITION, "oss error");
+  }
+  // 3. response
+  absl::Time step3 = absl::Now();
+  LOG(INFO) << "logid " << req->logid() << " text_to_speech input: " << input << " result: " << local_filename;
+
+  resp->set_content(local_filename);
+  // 4. TODO: delete aliyun oos
+  absl::Time step4 = absl::Now();
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid()
+            << " text_to_speech cost: " << absl::ToDoubleMilliseconds(step2 - step1)
+            << " upload_oss cost: " << absl::ToDoubleMilliseconds(step3 - step2)
+            << ", total cost: " << absl::ToDoubleMilliseconds(step3 - step1);
+  return grpc::Status::OK;
+}
+
 } // namespace chat_completion

src/service/ielts_ai.h

@@ -16,8 +16,10 @@
 #include "chat_completion/chat_completion.grpc.pb.h"
 #include "components/audio.h"
 #include "components/chat.h"
+#include "core/authorization.h"
 #include "src/plugin/oss.h"
 #include "src/plugin/prompt.h"
+#include "src/plugin/token.h"
 
 namespace chat_completion {
 
@@ -26,8 +28,9 @@ class IeltsAI final : public ChatService::Service {
   IeltsAI() = default;
   virtual ~IeltsAI() = default;
   int32_t initialize();
-  // 音频转文字
+  // 音频文字互转
   grpc::Status transcribe_judge(grpc::ServerContext*, const ChatMessage*, ChatMessage*) override;
+  grpc::Status text_to_speech(grpc::ServerContext*, const ChatMessage*, ChatMessage*) override;
   // 雅思口语P1
   grpc::Status ielts_speaking_p1_generate(grpc::ServerContext*, const ChatMessage*,
                                           grpc::ServerWriter<ChatMessage>*) override;
@@ -100,6 +103,8 @@ class IeltsAI final : public ChatService::Service {
   void stream_handler() {}
 
  private:
+  liboai::Authorization _audio_auth{};
+  liboai::Authorization _chat_auth{};
   std::unique_ptr<liboai::Audio> _audio{nullptr};
   std::unique_ptr<liboai::ChatCompletion> _chat_completion{nullptr};
   std::string _model{"qwen-plus"};

src/service/ielts_speaking_p1.cpp

@@ -41,8 +41,8 @@ grpc::Status IeltsAI::ielts_speaking_p1_generate(grpc::ServerContext* ctx, const
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 
@@ -85,8 +85,8 @@ grpc::Status IeltsAI::ielts_speaking_p1_enrich(grpc::ServerContext* ctx, const C
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 

src/service/ielts_speaking_p2.cpp

@@ -41,8 +41,8 @@ grpc::Status IeltsAI::ielts_speaking_p2_generate(grpc::ServerContext* ctx, const
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 
@@ -85,8 +85,8 @@ grpc::Status IeltsAI::ielts_speaking_p2_enrich(grpc::ServerContext* ctx, const C
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 
@@ -129,8 +129,8 @@ grpc::Status IeltsAI::ielts_speaking_p2_score(grpc::ServerContext* ctx, const Ch
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 

src/service/ielts_speaking_p3.cpp

@@ -41,8 +41,8 @@ grpc::Status IeltsAI::ielts_speaking_p3_generate(grpc::ServerContext* ctx, const
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 
@@ -85,8 +85,8 @@ grpc::Status IeltsAI::ielts_speaking_p3_enrich(grpc::ServerContext* ctx, const C
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 

src/service/ielts_words.cpp

@@ -40,8 +40,8 @@ grpc::Status IeltsAI::ielts_speaking_words_synonyms(grpc::ServerContext* ctx, co
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 
@@ -83,8 +83,8 @@ grpc::Status IeltsAI::ielts_speaking_words_usage(grpc::ServerContext* ctx, const
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 
@@ -127,8 +127,8 @@ grpc::Status IeltsAI::ielts_writing_words_theme(grpc::ServerContext* ctx, const
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 
@@ -170,8 +170,8 @@ grpc::Status IeltsAI::ielts_writing_words_synonyms(grpc::ServerContext* ctx, con
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 
@@ -213,8 +213,8 @@ grpc::Status IeltsAI::ielts_writing_words_combination(grpc::ServerContext* ctx,
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content() << " total cost time "
-            << absl::ToDoubleMilliseconds(step2 - step1);
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
+            << " total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
 

src/service/ielts_writing_t1.cpp

@@ -40,7 +40,7 @@ grpc::Status IeltsAI::ielts_writing_t1_enrich(grpc::ServerContext* ctx, const Ch
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content()
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
             << ", enrich total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
@@ -83,7 +83,7 @@ grpc::Status IeltsAI::ielts_writing_t1_score(grpc::ServerContext* ctx, const Cha
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content()
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
             << ", score total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }

src/service/ielts_writing_t2.cpp

@@ -40,7 +40,7 @@ grpc::Status IeltsAI::ielts_writing_t2_generate(grpc::ServerContext* ctx, const
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << "generate cost time "
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << "generate cost time "
             << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
@@ -84,7 +84,7 @@ grpc::Status IeltsAI::ielts_writing_t2_enrich(grpc::ServerContext* ctx, const Ch
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content()
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
             << ", enrich total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }
@@ -127,7 +127,7 @@ grpc::Status IeltsAI::ielts_writing_t2_score(grpc::ServerContext* ctx, const Cha
   openai_resp.wait();
 
   absl::Time step2 = absl::Now();
-  LOG(INFO) << "logid " << req->logid() << " uid " << req->uid() << " content " << req->content()
+  LOG(INFO) << "logid " << req->logid() << " uid " << req->userid() << " content " << req->content()
             << ", score total cost time " << absl::ToDoubleMilliseconds(step2 - step1);
   return grpc::Status::OK;
 }