minor tweaks for key and readme

.vscode/launch.json

@@ -9,7 +9,8 @@
             "type": "debugpy",
             "request": "launch",
             "program": "${file}",
-            "console": "integratedTerminal"
+            "console": "integratedTerminal",
+            "args": ["--log_file", "/Users/quinone/Documents/projects/audio-stream/prompts.log"]
         }
     ]
 }

README.md

@@ -4,16 +4,41 @@ This project is a Gradio-based application that allows users to interact with an
 
 ## Table of Contents
 
-- [Prerequisites](#prerequisites)
+- [Install](#install)
 - [Usage](#usage)
 - [Features](#features)
 - [Configuration](#configuration)
 - [Deployment](#deployment)
 - [License](#license)
 
-## Prerequisites
-
-- Python 3.7 or higher
+## Install
+Installation and environment setup is currently done locally, but with a little effort, we could 
+make a docker image.
+
+1. Install miniconda or anaconda [here](https://docs.conda.io/en/latest/miniconda.html)
+2. Create a new environment with the following command:
+    ```bash
+    conda create -n audio-stream python=3.11
+    ```
+3. Activate the environment:
+    ```bash
+    conda activate audio-stream
+    ```
+4. Install the required packages:
+    ```bash
+    pip install -r requirements.txt
+    ```
+5. Set the required environment variables or create a `.env` file (exclude the word `export` for `.env`)
+    ```bash
+    export OPENAI_API_KEY=<your_openai_api_key>
+    ```
+6. Run the Gradio application:
+    ```bash
+    python app.py
+    ```
+
+### General Requirements
+- Python 3.9 or higher  (recommend 3.11)
 - [Gradio](https://www.gradio.app/) library
 - [OpenAI](https://openai.com/) API key
 - [Whisper](https://github.com/openai/whisper) library (for speech recognition)
@@ -35,20 +60,15 @@ This project is a Gradio-based application that allows users to interact with an
 
 ## Configuration
 
-The application can be configured using command-line arguments or environment variables. The available configuration options include:
-
-- `--model`: The OpenAI model to use for language tasks.
-- `--temperature`: The temperature parameter for the OpenAI model.
-- `--max_tokens`: The maximum number of tokens to generate.
-- `--port`: The port number for the Gradio application.
+The application can be configured using command-line arguments or environment variables. Run the main command with the option `--help` to get a full list of available options.
 
 ## Deployment
 
 The application can be deployed to various platforms, such as:
 
 - **Local Machine**: Run the application on your local machine using the instructions in the [Usage](#usage) section.
-- **Cloud Platform**: Deploy the application to a cloud platform like AWS, Google Cloud, or Azure.
 - **Docker**: Package the application in a Docker container for easy deployment and scaling.
+- **Cloud Platform**: Deploy the application to a cloud platform like AWS, Google Cloud, or Azure.  At first blush [a deployment strategy like this one](https://vinaykachare.medium.com/serverless-api-with-aws-sam-fastapi-3f4d9510d6b6) seems like a good follow-up for automated deployment.
 
 
 ## License

stream_app.py

@@ -38,14 +38,10 @@ def run_gradio(config:dict):
                 whisper_model = whisper.load_model("base")
             audio = whisper.load_audio(input_audio)
             result = whisper_model.transcribe(audio)
-            result["no_speech_prob"] = 0
-            prob_scores = [x['no_speech_prob'] for x in result['segments']]
-            if len(prob_scores) > 0:    # average the probs
-                result["no_speech_prob"] = sum(prob_scores)/len(prob_scores)
 
         elif "online" in input_audio_model.lower():
             with open(input_audio, 'rb') as file_audio:
-                result = client.audio.translations.create(
+                result = client.audio.transcriptions.create(
                     model="whisper-1", file=file_audio, response_format="verbose_json",
                 )
                 if result is None:
@@ -53,7 +49,13 @@ def run_gradio(config:dict):
                 result = result.to_dict()
         prompt = result["text"]
         logger.warning(f"Transcription: {result}")
-        
+
+        if "no_speech_prob" not in result:      # look for probability of a good tanscription
+            result["no_speech_prob"] = 1.0
+            prob_scores = [x['no_speech_prob'] for x in result['segments']]
+            if len(prob_scores) > 0:    # average the probs
+                result["no_speech_prob"] = sum(prob_scores)/len(prob_scores)
+
         if result["no_speech_prob"] < (1 - config['speech_threshold']):      # threshold to avoid bad output
             return input_text + " " + prompt
         return input_text
@@ -82,6 +84,7 @@ def run_gradio(config:dict):
 
         partial_response = ""
         for stream_response in response:
+            logger.warning(f"Prompt response: {stream_response.to_dict()}")
             token = stream_response.choices[0].delta.content
             if token is None:
                 break
@@ -95,22 +98,32 @@ def run_gradio(config:dict):
         """)
         with gr.Row():
             with gr.Column():
-                input_text = gr.Textbox(
-                    label="Text Input",
-                    placeholder="Enter your prompt here",
-                    lines=5,
-                    max_lines=10,
-                )
-                input_audio_model = gr.Radio(
-                    label="Audio Model",
-                    choices=["whisper (offline)", "openai-whisper (online)"],
-                    value="openai-whisper (online)",
-                )
-                input_audio = gr.Audio(
-                    label="Speech Input",
-                    streaming=True,
-                    type="filepath",
-                )
+                with gr.Group():
+                    input_text = gr.Textbox(
+                        label="Text Input",
+                        placeholder="Enter your prompt here",
+                        lines=5,
+                        max_lines=10,
+                    )
+                    online_text_model = f"openai-{config['model']} (online)"
+                    input_audio_model = gr.Radio(
+                        label="Textual Model",
+                        choices=[online_text_model],
+                        value=online_text_model,
+                    )
+
+                with gr.Group():
+                    input_audio = gr.Audio(
+                        label="Speech Input",
+                        streaming=True,
+                        type="filepath",
+                    )
+                    input_audio_model = gr.Radio(
+                        label="Audio Model",
+                        choices=["whisper (offline)", "openai-whisper (online)"],
+                        value="openai-whisper (online)",
+                    )
+
             with gr.Column():
                 output_text = gr.Textbox(
                     label="Output",
@@ -148,8 +161,8 @@ def parse_args() -> dict:
         help="Maximum number of tokens to generate in chat completion.")
     
     opt_group = parser.add_argument_group("Speech Processing")
-    opt_group.add_argument("--speech_threshold", type=float, default=0.5,
-        help="Speech threshold for recognition to add text to a prompt. ")
+    opt_group.add_argument("--speech_threshold", type=float, default=0.10,
+        help="Speech threshold (probability) for recognition to add text to a prompt. ")
 
     opt_group = parser.add_argument_group("App Settings")
     opt_group.add_argument("--port", type=int, default=7860, 
@@ -168,6 +181,10 @@ def parse_args() -> dict:
 
 if __name__ == "__main__":
     os.environ['GRADIO_ANALYTICS_ENABLED'] = 'False'
+    api_key = os.environ.get("OPENAI_API_KEY")
+    print(api_key)
+    if not api_key:
+        raise ValueError("OPENAI_API_KEY environment variable not set as environment variable or as a setting in `.env`.  (see https://platform.openai.com/docs/quickstart/step-2-set-up-your-api-key)")
 
     config = parse_args()
     run_gradio(config)