Merge branch 'main' of hf.co:spaces/gospacedev/friday

README.md

@@ -1,13 +1,13 @@
 ---
 title: Friday
-emoji: 👩‍🏭
+emoji: 🚀
 colorFrom: pink
 colorTo: purple
 sdk: gradio
-sdk_version: 4.28.3
+sdk_version: 4.39.0
 app_file: app.py
 pinned: false
 license: mit
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -6,28 +6,30 @@ from gtts import gTTS
 from transformers import pipeline
 from huggingface_hub import InferenceClient
 
-
+# Model names
 ASR_MODEL_NAME = "openai/whisper-small"
 LLM_MODEL_NAME = "mistralai/Mistral-7B-Instruct-v0.2"
 
+# Initial system prompt
+system_prompt = """"<s>[INST] You are Friday, a helpful and conversational AI assistant, and you respond with one to two sentences. [/INST] Hello there! I'm Friday, how can I help you?</s>"""
 
-system_prompt = """"<s>[INST] You are Friday, a helpful and conversational AI assistant and You respond with one to two sentences. [/INST] Hello there! I'm friday, umm, how can I help you?</s>"""
-
-instruct_history = system_prompt + """"""
-
-formatted_history = """"""
+# Global variables for history
+instruct_history = system_prompt
+formatted_history = ""
 
+# Create inference client for text generation
 client = InferenceClient(LLM_MODEL_NAME)
 
+# Set device for ASR pipeline
 device = 0 if torch.cuda.is_available() else "cpu"
 
+# ASR pipeline
 pipe = pipeline(
     task="automatic-speech-recognition",
     model=ASR_MODEL_NAME,
     device=device,
 )
 
-
 def generate(instruct_history, temperature=0.1, max_new_tokens=128, top_p=0.95, repetition_penalty=1.0):
     temperature = float(temperature)
     if temperature < 1e-2:
@@ -48,50 +50,50 @@ def generate(instruct_history, temperature=0.1, max_new_tokens=128, top_p=0.95,
 
     return output
 
-
 @spaces.GPU(duration=60)
-def transcribe(audio, instruct_history=instruct_history, formatted_history=formatted_history):
+def transcribe(audio, past_history):
+    global instruct_history, formatted_history
+
     sr, y = audio
     y = y.astype(np.float32)
     y /= np.max(np.abs(y))
 
     transcribed_user_audio = pipe({"sampling_rate": sr, "raw": y})["text"]
 
-    formatted_history += f"""Human: {transcribed_user_audio}\n\n"""
+    formatted_history += past_history
 
-    instruct_history += f""" <s>[INST] {transcribed_user_audio} [/INST] """
+    formatted_history += f"😃 Human: {transcribed_user_audio}\n\n"
+    instruct_history += f"<s>[INST] {transcribed_user_audio} [/INST] "
 
+    # Generate LLM response
     llm_response = generate(instruct_history)
 
-    instruct_history += f""" {llm_response}</s> """
-
-    formatted_history += f"""Friday: {llm_response}\n\n"""
+    instruct_history += f" {llm_response}</s>"
+    formatted_history += f"🤖 Friday: {llm_response}\n\n"
 
+    # Convert AI response to audio
     audio_response = gTTS(llm_response)
     audio_response.save("response.mp3")
 
-    print(instruct_history)
+    print("Formatted History: ", formatted_history)
 
+    # Return the full conversation history
     return "response.mp3", formatted_history
 
-
 with gr.Blocks() as demo:
-    gr.HTML("<center><h1>Friday: AI Virtual Assistant<h1><center>")
+    gr.HTML("<center><h1>Friday: AI Virtual Assistant 🤖</h1><center>")
 
     with gr.Row():
         audio_input = gr.Audio(label="Human", sources="microphone")
-        output_audio = gr.Audio(label="Friday", type="filepath",
-                                interactive=False,
-                                autoplay=True,
-                                elem_classes="audio")
+        output_audio = gr.Audio(label="Friday", type="filepath", interactive=False, autoplay=True, elem_classes="audio")
 
     transcribe_btn = gr.Button("Transcribe")
 
-    transcription_box = gr.Textbox(label="Transcription")
+    # Textbox to display the full conversation history
+    transcription_box = gr.Textbox(label="Transcription", lines=10, placeholder="Conversation History...")
 
-    transcribe_btn.click(fn=transcribe, inputs=[audio_input],
-                         outputs=[output_audio, transcription_box])
+    transcribe_btn.click(fn=transcribe, inputs=[audio_input, transcription_box], outputs=[output_audio, transcription_box])
 
 if __name__ == "__main__":
     demo.queue()
-    demo.launch()
+    demo.launch()