Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -14,31 +14,31 @@ def image_to_base64(img_path):
|
|
| 14 |
img_base64 = image_to_base64("SBC6.jpg")
|
| 15 |
img_html = f'<img src="data:image/jpg;base64,{img_base64}" alt="SBC6" width="300" style="display: block; margin: auto;"/>'
|
| 16 |
|
| 17 |
-
|
| 18 |
def predict(question_choice, audio):
|
| 19 |
# Transcribe the audio using Whisper
|
| 20 |
with open(audio, "rb") as audio_file:
|
| 21 |
transcript = openai.Audio.transcribe("whisper-1", audio_file)
|
|
|
|
| 22 |
|
| 23 |
-
message = transcript["text"]
|
| 24 |
-
|
| 25 |
# Generate the system message based on the chosen question
|
| 26 |
current_question_index = data6.questions.index(question_choice)
|
|
|
|
| 27 |
|
| 28 |
# Construct the conversation with the system and user's message
|
| 29 |
conversation = [
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
|
|
|
| 36 |
|
| 37 |
response = openai.ChatCompletion.create(
|
| 38 |
model='gpt-3.5-turbo',
|
| 39 |
messages=conversation,
|
| 40 |
-
temperature=0.
|
| 41 |
-
max_tokens=
|
| 42 |
stream=True
|
| 43 |
)
|
| 44 |
|
|
@@ -48,9 +48,11 @@ def predict(question_choice, audio):
|
|
| 48 |
partial_message = partial_message + chunk['choices'][0]['delta']['content']
|
| 49 |
yield partial_message
|
| 50 |
|
|
|
|
| 51 |
def get_image_html():
|
| 52 |
return "" # Markdown syntax to embed the image
|
| 53 |
|
|
|
|
| 54 |
# Gradio Interface
|
| 55 |
iface = gr.Interface(
|
| 56 |
fn=predict,
|
|
@@ -59,7 +61,8 @@ iface = gr.Interface(
|
|
| 59 |
gr.inputs.Audio(source="microphone", type="filepath") # Audio input
|
| 60 |
],
|
| 61 |
outputs=gr.inputs.Textbox(), # Using inputs.Textbox as an output to make it editable
|
| 62 |
-
description=img_html
|
|
|
|
| 63 |
)
|
| 64 |
iface.queue().launch()
|
| 65 |
|
|
|
|
| 14 |
img_base64 = image_to_base64("SBC6.jpg")
|
| 15 |
img_html = f'<img src="data:image/jpg;base64,{img_base64}" alt="SBC6" width="300" style="display: block; margin: auto;"/>'
|
| 16 |
|
|
|
|
| 17 |
def predict(question_choice, audio):
|
| 18 |
# Transcribe the audio using Whisper
|
| 19 |
with open(audio, "rb") as audio_file:
|
| 20 |
transcript = openai.Audio.transcribe("whisper-1", audio_file)
|
| 21 |
+
message = transcript["text"] # This is the transcribed message from the audio input
|
| 22 |
|
|
|
|
|
|
|
| 23 |
# Generate the system message based on the chosen question
|
| 24 |
current_question_index = data6.questions.index(question_choice)
|
| 25 |
+
strategy, explanation = data6.strategy_text[current_question_index]
|
| 26 |
|
| 27 |
# Construct the conversation with the system and user's message
|
| 28 |
conversation = [
|
| 29 |
+
{
|
| 30 |
+
"role": "system",
|
| 31 |
+
"content": f"You are an expert English Language Teacher in a Singapore Primary school, directly guiding a Primary 6 student in Singapore. The student is answering the question: '{data6.questions[current_question_index]}'. Point out areas they did well and where they can improve. Then, provide a suggested answer using the {data6.strategy_text[current_question_index][0]} strategy. Encourage the use of sophisticated vocabulary and expressions. For the second and third questions, the picture is not relevant, so the student should not refer to it in their response. {explanation} The feedback should be in second person, addressing the student directly."
|
| 32 |
+
},
|
| 33 |
+
{"role": "user", "content": message}
|
| 34 |
+
]
|
| 35 |
+
|
| 36 |
|
| 37 |
response = openai.ChatCompletion.create(
|
| 38 |
model='gpt-3.5-turbo',
|
| 39 |
messages=conversation,
|
| 40 |
+
temperature=0.4,
|
| 41 |
+
max_tokens=400, # Limiting the response to 500 tokens
|
| 42 |
stream=True
|
| 43 |
)
|
| 44 |
|
|
|
|
| 48 |
partial_message = partial_message + chunk['choices'][0]['delta']['content']
|
| 49 |
yield partial_message
|
| 50 |
|
| 51 |
+
|
| 52 |
def get_image_html():
|
| 53 |
return "" # Markdown syntax to embed the image
|
| 54 |
|
| 55 |
+
|
| 56 |
# Gradio Interface
|
| 57 |
iface = gr.Interface(
|
| 58 |
fn=predict,
|
|
|
|
| 61 |
gr.inputs.Audio(source="microphone", type="filepath") # Audio input
|
| 62 |
],
|
| 63 |
outputs=gr.inputs.Textbox(), # Using inputs.Textbox as an output to make it editable
|
| 64 |
+
description=img_html,
|
| 65 |
+
css="custom.css" # Link to the custom CSS file
|
| 66 |
)
|
| 67 |
iface.queue().launch()
|
| 68 |
|