Update app.py
Browse files
app.py
CHANGED
|
@@ -9,9 +9,9 @@ from openai import OpenAI
|
|
| 9 |
|
| 10 |
# Use writable cache for Hugging Face
|
| 11 |
os.environ["MPLCONFIGDIR"] = "/tmp/matplotlib"
|
| 12 |
-
os.environ["HF_HOME"] = "/
|
| 13 |
-
os.environ["TRANSFORMERS_CACHE"] = "/
|
| 14 |
-
os.environ["XDG_CACHE_HOME"] = "/
|
| 15 |
|
| 16 |
chatter_owl_url = "file/images/chatter_owl.png"
|
| 17 |
|
|
@@ -20,28 +20,29 @@ model = WhisperModel("base", compute_type="int8") # Fastest CPU option
|
|
| 20 |
openai.api_key = os.getenv("OPENAI_API_KEY")
|
| 21 |
|
| 22 |
# Prompt template
|
| 23 |
-
PROMPT_TEMPLATE =
|
| 24 |
You are a communication coach evaluating a user's spoken response.
|
| 25 |
|
| 26 |
-
|
| 27 |
1. Clarity
|
| 28 |
2. Structure
|
| 29 |
3. Fluency
|
| 30 |
4. Content Relevance
|
| 31 |
5. Tone & Expression
|
| 32 |
|
| 33 |
-
For each
|
| 34 |
-
-
|
| 35 |
-
- Provide
|
| 36 |
|
| 37 |
Then:
|
| 38 |
-
-
|
| 39 |
-
- End with
|
| 40 |
|
| 41 |
-
|
|
|
|
| 42 |
|
| 43 |
Transcript:
|
| 44 |
-
{transcript}
|
| 45 |
"""
|
| 46 |
|
| 47 |
# Convert uploaded audio to 16kHz mono WAV
|
|
@@ -98,6 +99,29 @@ def tutor_feedback(audio_file):
|
|
| 98 |
|
| 99 |
return transcript, feedback_text, mp3_path
|
| 100 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 101 |
# Gradio interface
|
| 102 |
# Path to your CSS file
|
| 103 |
css_path = "light_mode_chatter_owl.css" # Make sure this file is in the same directory as app.py
|
|
@@ -105,13 +129,14 @@ css_path = "light_mode_chatter_owl.css" # Make sure this file is in the same di
|
|
| 105 |
# Path to your CSS file
|
| 106 |
css_path = "light_mode_chatter_owl.css" # Make sure this file is in the same directory as app.py
|
| 107 |
|
| 108 |
-
|
|
|
|
| 109 |
gr.Markdown(
|
| 110 |
"""
|
| 111 |
<div id="header" style="text-align: center;">
|
| 112 |
-
<img src="file/images/chatter_owl.png">
|
| 113 |
-
<h2>π¦ Meet Chatter the Owl
|
| 114 |
-
<p>Speak
|
| 115 |
</div>
|
| 116 |
"""
|
| 117 |
)
|
|
@@ -119,14 +144,26 @@ with gr.Blocks(css=css_path) as app:
|
|
| 119 |
with gr.Row():
|
| 120 |
audio_input = gr.Audio(type="filepath", label="π Speak or Upload Audio")
|
| 121 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 122 |
with gr.Row():
|
| 123 |
-
|
| 124 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 125 |
|
| 126 |
-
|
| 127 |
|
| 128 |
-
# Hook up your function (e.g., `tutor_feedback`)
|
| 129 |
-
audio_input.change(fn=tutor_feedback, inputs=audio_input, outputs=[transcript_box, feedback_box, audio_output])
|
| 130 |
|
| 131 |
if __name__ == "__main__":
|
| 132 |
print("β
App is launching...")
|
|
|
|
| 9 |
|
| 10 |
# Use writable cache for Hugging Face
|
| 11 |
os.environ["MPLCONFIGDIR"] = "/tmp/matplotlib"
|
| 12 |
+
os.environ["HF_HOME"] = "/tmp/hf"
|
| 13 |
+
os.environ["TRANSFORMERS_CACHE"] = "/tmp/hf"
|
| 14 |
+
os.environ["XDG_CACHE_HOME"] = "/tmp/hf"
|
| 15 |
|
| 16 |
chatter_owl_url = "file/images/chatter_owl.png"
|
| 17 |
|
|
|
|
| 20 |
openai.api_key = os.getenv("OPENAI_API_KEY")
|
| 21 |
|
| 22 |
# Prompt template
|
| 23 |
+
PROMPT_TEMPLATE = """
|
| 24 |
You are a communication coach evaluating a user's spoken response.
|
| 25 |
|
| 26 |
+
Evaluate the speech using these 5 dimensions:
|
| 27 |
1. Clarity
|
| 28 |
2. Structure
|
| 29 |
3. Fluency
|
| 30 |
4. Content Relevance
|
| 31 |
5. Tone & Expression
|
| 32 |
|
| 33 |
+
For each:
|
| 34 |
+
- Score from 0β10
|
| 35 |
+
- Provide brief, specific feedback
|
| 36 |
|
| 37 |
Then:
|
| 38 |
+
- Give an overall 3β4 line summary
|
| 39 |
+
- End with a motivational sentence
|
| 40 |
|
| 41 |
+
Lastly:
|
| 42 |
+
- Provide an improved version of the speech that better demonstrates effective communication, based on their original transcript.
|
| 43 |
|
| 44 |
Transcript:
|
| 45 |
+
\"\"\"{transcript}\"\"\"
|
| 46 |
"""
|
| 47 |
|
| 48 |
# Convert uploaded audio to 16kHz mono WAV
|
|
|
|
| 99 |
|
| 100 |
return transcript, feedback_text, mp3_path
|
| 101 |
|
| 102 |
+
# Generate example version of the same speech
|
| 103 |
+
def generate_example_response(transcript):
|
| 104 |
+
prompt = f"""
|
| 105 |
+
You are a communication tutor. Rewrite this speech transcript to be a more polished and confident version, while keeping the meaning and tone similar.
|
| 106 |
+
|
| 107 |
+
Transcript:
|
| 108 |
+
{transcript}
|
| 109 |
+
"""
|
| 110 |
+
response = openai.ChatCompletion.create(
|
| 111 |
+
model="gpt-4",
|
| 112 |
+
messages=[
|
| 113 |
+
{"role": "system", "content": "You are a helpful speaking coach."},
|
| 114 |
+
{"role": "user", "content": prompt}
|
| 115 |
+
],
|
| 116 |
+
temperature=0.7
|
| 117 |
+
)
|
| 118 |
+
return response["choices"][0]["message"]["content"]
|
| 119 |
+
|
| 120 |
+
# Hook up logic
|
| 121 |
+
def show_example_feedback(transcript):
|
| 122 |
+
# Extract only example portion from GPT or return a pre-generated version
|
| 123 |
+
return generate_example_response(transcript)
|
| 124 |
+
|
| 125 |
# Gradio interface
|
| 126 |
# Path to your CSS file
|
| 127 |
css_path = "light_mode_chatter_owl.css" # Make sure this file is in the same directory as app.py
|
|
|
|
| 129 |
# Path to your CSS file
|
| 130 |
css_path = "light_mode_chatter_owl.css" # Make sure this file is in the same directory as app.py
|
| 131 |
|
| 132 |
+
# App UI
|
| 133 |
+
with gr.Blocks(css="light_mode_chatter_owl.css") as app:
|
| 134 |
gr.Markdown(
|
| 135 |
"""
|
| 136 |
<div id="header" style="text-align: center;">
|
| 137 |
+
<img src="file/images/chatter_owl.png" width="120">
|
| 138 |
+
<h2>π¦ Meet <strong>Chatter the Owl</strong></h2>
|
| 139 |
+
<p>Speak into the mic, and Iβll give you structured feedback to help you grow as a communicator!</p>
|
| 140 |
</div>
|
| 141 |
"""
|
| 142 |
)
|
|
|
|
| 144 |
with gr.Row():
|
| 145 |
audio_input = gr.Audio(type="filepath", label="π Speak or Upload Audio")
|
| 146 |
|
| 147 |
+
transcript_box = gr.Textbox(label="π What You Said", interactive=False)
|
| 148 |
+
feedback_box = gr.Textbox(label="π‘ Chatterβs Feedback", interactive=False)
|
| 149 |
+
audio_output = gr.Audio(label="π Chatter Speaks", type="filepath")
|
| 150 |
+
hidden_transcript = gr.Textbox(visible=False)
|
| 151 |
+
|
| 152 |
with gr.Row():
|
| 153 |
+
try_again = gr.Button("π Try Again")
|
| 154 |
+
show_example = gr.Button("π― Show Me an Example")
|
| 155 |
+
|
| 156 |
+
example_box = gr.Textbox(label="π£ Suggested Improvement", visible=False)
|
| 157 |
+
|
| 158 |
+
# Interactions
|
| 159 |
+
audio_input.change(fn=tutor_feedback, inputs=audio_input,
|
| 160 |
+
outputs=[transcript_box, feedback_box, audio_output, hidden_transcript])
|
| 161 |
+
|
| 162 |
+
try_again.click(fn=lambda: ("", "", None, "", ""), inputs=None,
|
| 163 |
+
outputs=[transcript_box, feedback_box, audio_output, hidden_transcript, example_box])
|
| 164 |
|
| 165 |
+
show_example.click(fn=generate_example_response, inputs=hidden_transcript, outputs=example_box)
|
| 166 |
|
|
|
|
|
|
|
| 167 |
|
| 168 |
if __name__ == "__main__":
|
| 169 |
print("β
App is launching...")
|