w

Author: ks6088ts

apps/14_streamlit_azure_ai_speech/main.py

@@ -13,57 +13,71 @@
     st.session_state["transcribed_result"] = ""
 
 with st.sidebar:
-    azure_openai_endpoint = st.text_input(
-        label="AZURE_OPENAI_ENDPOINT",
-        value=getenv("AZURE_OPENAI_ENDPOINT"),
-        key="AZURE_OPENAI_ENDPOINT",
-        type="default",
-    )
-    azure_openai_api_key = st.text_input(
-        label="AZURE_OPENAI_API_KEY",
-        value=getenv("AZURE_OPENAI_API_KEY"),
-        key="AZURE_OPENAI_API_KEY",
-        type="password",
-    )
-    azure_openai_api_version = st.text_input(
-        label="AZURE_OPENAI_API_VERSION",
-        value=getenv("AZURE_OPENAI_API_VERSION"),
-        key="AZURE_OPENAI_API_VERSION",
-        type="default",
-    )
-    azure_openai_gpt_model = st.text_input(
-        label="AZURE_OPENAI_GPT_MODEL",
-        value=getenv("AZURE_OPENAI_GPT_MODEL"),
-        key="AZURE_OPENAI_GPT_MODEL",
-        type="default",
-    )
-    azure_ai_speech_api_subscription_key = st.text_input(
-        label="AZURE_AI_SPEECH_API_SUBSCRIPTION_KEY",
-        value=getenv("AZURE_AI_SPEECH_API_SUBSCRIPTION_KEY"),
-        key="AZURE_AI_SPEECH_API_SUBSCRIPTION_KEY",
-        type="password",
-    )
-    azure_ai_speech_api_region = st.text_input(
-        label="AZURE_AI_SPEECH_API_REGION",
-        value=getenv("AZURE_AI_SPEECH_API_REGION"),
-        key="AZURE_AI_SPEECH_API_REGION",
-        type="default",
-    )
-    azure_ai_speech_api_language = st.selectbox(
-        label="AZURE_AI_SPEECH_API_LANGUAGE",
+    inference_type = st.selectbox(
+        label="INEFERENCE_TYPE",
         options=[
-            "en-US",
-            "ja-JP",
+            "local",
+            "azure",
         ],
-        key="AZURE_AI_SPEECH_API_LANGUAGE",
+        key="INEFERENCE_TYPE",
     )
+    if inference_type == "local":
+        st.warning("yet to be implemented")
+    if inference_type == "azure":
+        azure_openai_endpoint = st.text_input(
+            label="AZURE_OPENAI_ENDPOINT",
+            value=getenv("AZURE_OPENAI_ENDPOINT"),
+            key="AZURE_OPENAI_ENDPOINT",
+            type="default",
+        )
+        azure_openai_api_key = st.text_input(
+            label="AZURE_OPENAI_API_KEY",
+            value=getenv("AZURE_OPENAI_API_KEY"),
+            key="AZURE_OPENAI_API_KEY",
+            type="password",
+        )
+        azure_openai_api_version = st.text_input(
+            label="AZURE_OPENAI_API_VERSION",
+            value=getenv("AZURE_OPENAI_API_VERSION"),
+            key="AZURE_OPENAI_API_VERSION",
+            type="default",
+        )
+        azure_openai_gpt_model = st.text_input(
+            label="AZURE_OPENAI_GPT_MODEL",
+            value=getenv("AZURE_OPENAI_GPT_MODEL"),
+            key="AZURE_OPENAI_GPT_MODEL",
+            type="default",
+        )
+        azure_ai_speech_api_subscription_key = st.text_input(
+            label="AZURE_AI_SPEECH_API_SUBSCRIPTION_KEY",
+            value=getenv("AZURE_AI_SPEECH_API_SUBSCRIPTION_KEY"),
+            key="AZURE_AI_SPEECH_API_SUBSCRIPTION_KEY",
+            type="password",
+        )
+        azure_ai_speech_api_region = st.text_input(
+            label="AZURE_AI_SPEECH_API_REGION",
+            value=getenv("AZURE_AI_SPEECH_API_REGION"),
+            key="AZURE_AI_SPEECH_API_REGION",
+            type="default",
+        )
+        azure_ai_speech_api_language = st.selectbox(
+            label="AZURE_AI_SPEECH_API_LANGUAGE",
+            options=[
+                "en-US",
+                "ja-JP",
+            ],
+            key="AZURE_AI_SPEECH_API_LANGUAGE",
+        )
     "[Azure Portal](https://portal.azure.com/)"
     "[Azure OpenAI Studio](https://oai.azure.com/resource/overview)"
     "[View the source code](https://github.com/ks6088ts-labs/workshop-azure-openai/blob/main/apps/14_streamlit_azure_ai_speech/main.py)"
 
 
 def is_configured():
-    return azure_openai_api_key and azure_openai_endpoint and azure_openai_api_version and azure_openai_gpt_model
+    if inference_type == "local":
+        return False
+    if inference_type == "azure":
+        return azure_openai_api_key and azure_openai_endpoint and azure_openai_api_version and azure_openai_gpt_model
 
 
 st.title("transcribe text")
@@ -103,7 +117,7 @@ def is_configured():
         "Task",
         [
             "Create summaries from the following text",
-            "Translate the following text into English",
+            "Extract 3 main points from the following text",
             # Add more tasks here
         ],
         key="selected_task",
@@ -118,32 +132,41 @@ def is_configured():
 
 def start_recognition():
     global process
-    command = f"python apps/14_streamlit_azure_ai_speech/speech_to_text.py --output {path_to_transcribed_text} --subscription {azure_ai_speech_api_subscription_key} --region {azure_ai_speech_api_region} --language {azure_ai_speech_api_language} --verbose"  # noqa
-    process = subprocess.Popen(command, shell=True)
+    if inference_type == "local":
+        st.warning("Local inference is not yet implemented.")
+        return
+    if inference_type == "azure":
+        command = f"python apps/14_streamlit_azure_ai_speech/speech_to_text.py --output {path_to_transcribed_text} --subscription {azure_ai_speech_api_subscription_key} --region {azure_ai_speech_api_region} --language {azure_ai_speech_api_language} --verbose"  # noqa
+        process = subprocess.Popen(command, shell=True)
 
 
 def run_task(selected_task: str, input: str) -> str:
-    client = AzureOpenAI(
-        api_key=azure_openai_api_key,
-        api_version=azure_openai_api_version,
-        azure_endpoint=azure_openai_endpoint,
-    )
+    if inference_type == "local":
+        st.warning("Local inference is not yet implemented.")
+        return
+    if inference_type == "azure":
+        client = AzureOpenAI(
+            api_key=azure_openai_api_key,
+            api_version=azure_openai_api_version,
+            azure_endpoint=azure_openai_endpoint,
+        )
 
-    response = client.chat.completions.create(
-        model=azure_openai_gpt_model,
-        messages=[
-            {
-                "role": "system",
-                "content": f"""
-                    Task: {selected_task}.
-                    ---
-                    {input}
-                    ---
-                """,
-            },
-        ],
-    )
-    return response.choices[0].message.content
+        response = client.chat.completions.create(
+            model=azure_openai_gpt_model,
+            messages=[
+                {
+                    "role": "system",
+                    "content": f"""
+                        Task: {selected_task}.
+                        ---
+                        {input}
+                        ---
+                    """,
+                },
+            ],
+        )
+        return response.choices[0].message.content
+    raise ValueError(f"Inference type is not supported: {inference_type}")
 
 
 def load_transcribed_text():
@@ -153,7 +176,7 @@ def load_transcribed_text():
 
 if start_transcribe_button:
     if not st.session_state.get("process"):
-        transcription_status.info("Transcribing...")
+        transcription_status.info(f"Transcribing... (language={azure_ai_speech_api_language})")
         start_recognition()
     else:
         transcription_status.warning("Transcription is already running.")
@@ -165,10 +188,9 @@ def load_transcribed_text():
     st.rerun()
 
 if run_task_button:
-    transcribed_text = load_transcribed_text()
     with st.spinner("Running..."):
         output = run_task(
             selected_task=selected_task,
-            input=transcribed_text,
+            input=input,
         )
         st.write(output)