Update app.py

app.py

@@ -22,16 +22,19 @@ from elevenlabs.client import ElevenLabs
 from huggingface_hub import hf_hub_download
 from gradio.themes.utils import colors, fonts, sizes
 from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
-from prompt_examples import TEXT_CHAT_EXAMPLES, IMG_GEN_PROMPT_EXAMPLES, AUDIO_EXAMPLES, TEXT_CHAT_EXAMPLES_LABELS, IMG_GEN_PROMPT_EXAMPLES_LABELS, AUDIO_EXAMPLES_LABELS
+from prompt_examples import TEXT_CHAT_EXAMPLES, IMG_GEN_PROMPT_EXAMPLES, AUDIO_EXAMPLES, TEXT_CHAT_EXAMPLES_LABELS, IMG_GEN_PROMPT_EXAMPLES_LABELS, AUDIO_EXAMPLES_LABELS, AYA_VISION_PROMPT_EXAMPLES
 from preambles import CHAT_PREAMBLE, AUDIO_RESPONSE_PREAMBLE, IMG_DESCRIPTION_PREAMBLE
 from constants import LID_LANGUAGES, NEETS_AI_LANGID_MAP, AYA_MODEL_NAME, BATCH_SIZE, USE_ELVENLABS, USE_REPLICATE
+from aya_vision_utils import get_aya_vision_response, get_aya_vision_prompt_example
+# from dotenv import load_dotenv
 
+# load_dotenv()
 
 HF_API_TOKEN =  os.getenv("HF_API_KEY")
 ELEVEN_LABS_KEY = os.getenv("ELEVEN_LABS_KEY")
 NEETS_AI_API_KEY = os.getenv("NEETS_AI_API_KEY")
 GROQ_API_KEY = os.getenv("GROQ_API_KEY")
-IMG_COHERE_API_KEY = os.getenv("IMG_COHERE_API_KEY")
+IMG_GEN_COHERE_API_KEY = os.getenv("IMG_GEN_COHERE_API_KEY")
 AUDIO_COHERE_API_KEY = os.getenv("AUDIO_COHERE_API_KEY")
 CHAT_COHERE_API_KEY = os.getenv("CHAT_COHERE_API_KEY")
 
@@ -39,7 +42,7 @@ DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 
 # Initialize cohere clients
 img_prompt_client = cohere.Client(
-    api_key=IMG_COHERE_API_KEY,
+    api_key=IMG_GEN_COHERE_API_KEY,
     client_name="c4ai-aya-expanse-img"
 )
 chat_client = cohere.Client(
@@ -326,6 +329,8 @@ def groq_whisper_tts(filename):
     return transcriptions.text
 
 
+
+
 # setup gradio app theme    
 theme = gr.themes.Base(
     primary_hue=gr.themes.colors.teal, 
@@ -334,8 +339,8 @@ theme = gr.themes.Base(
     text_size=gr.themes.sizes.text_lg,
 ).set(
     # Primary Button Color
-    button_primary_background_fill="#114A56",
-    button_primary_background_fill_hover="#114A56",
+    button_primary_background_fill="#2F70E3", #"#114A56",
+    button_primary_background_fill_hover="#2F70E3", #"#114A56",
     # Block Labels
     block_title_text_weight="600",
     block_label_text_weight="600",
@@ -348,13 +353,13 @@ demo = gr.Blocks(theme=theme, analytics_enabled=False)
 with demo:
     with gr.Row(variant="panel"):
         with gr.Column(scale=1):
-            gr.Image("AyaExpanse.png", elem_id="logo-img", show_label=False, show_share_button=False, show_download_button=False, show_fullscreen_button=False)
+            gr.Image("1.png", elem_id="logo-img", show_label=False, show_share_button=False, show_download_button=False, show_fullscreen_button=False)
         with gr.Column(scale=30):
-            gr.Markdown("""C4AI Aya Expanse is a state-of-art model with highly advanced capabilities to connect the world across languages. 
+            gr.Markdown("""C4AI Aya model family covers state-of-art models like Aya Vision and Aya Expanse with highly advanced capabilities to connect the world across languages. 
             <br/> 
-            You can use this space to chat, speak and visualize with Aya Expanse in 23 languages.
+            You can use this space to chat, speak, visualize and see with Aya models in 23 languages.
             <br/> 
-            **Model**: [aya-expanse-32B](https://huggingface.co/CohereForAI/aya-expanse-32b)
+            **Model**: [aya-vision-32b](https://huggingface.co/CohereForAI/aya-vision-32b), [aya-expanse-32B](https://huggingface.co/CohereForAI/aya-expanse-32b)
             <br/>
             **Developed by**: [Cohere for AI](https://cohere.com/research) and [Cohere](https://cohere.com/)
             <br/>
@@ -400,7 +405,6 @@ with demo:
                     examples_per_page=25,
                     run_on_click=True
                 )
-
     # End to End Testing Pipeline for speak with Aya  
     with gr.TabItem("Speak with Aya") as speak_with_aya:
         
@@ -410,6 +414,7 @@ with demo:
                 e2_audio_submit_button = gr.Button(value="Get Aya's Response", variant="primary")
                 
                 clear_button_microphone = gr.ClearButton()
+                
                 gr.Examples(
                     examples=AUDIO_EXAMPLES,
                     inputs=e2e_audio_file,
@@ -425,19 +430,16 @@ with demo:
                 e2e_aya_audio_response = gr.Audio(type="filepath", label="Aya's Audio Response")
 
         with gr.Accordion("See Details", open=False):
-            gr.Markdown("To enable voice interaction with Aya Expanse, this space uses [Whisper large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) and [Groq](https://groq.com/) for STT and [neets.ai](http://neets.ai/) for TTS.")
-                    
-
+            gr.Markdown("To enable voice interaction with Aya Expanse, this space uses [Whisper large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) and [Groq](https://groq.com/) for STT and [neets.ai](http://neets.ai/) for TTS.")  
+    
     # Generate Images
     with gr.TabItem("Visualize with Aya") as visualize_with_aya:
         with gr.Row():
             with gr.Column():
-                input_img_prompt = gr.Textbox(placeholder="Ask anything in our 23 languages ...", label="Describe an image", lines=3)
-                # generated_img_desc = gr.Textbox(label="Image Description generated by Aya", interactive=False, lines=3, visible=False) 
+                input_img_prompt = gr.Textbox(placeholder="Ask anything in our 23 languages ...", label="Ask anything about an image", lines=3)
                 submit_button_img = gr.Button(value="Submit", variant="primary")
                 clear_button_img = gr.ClearButton()
                 
-            
             with gr.Column():
                 generated_img = gr.Image(label="Generated Image", interactive=False)
                 input_prompt_lang = gr.Textbox(visible=False)
@@ -465,7 +467,56 @@ with demo:
         with gr.Row():
             with gr.Accordion("See Details", open=False):
                 gr.Markdown("This space uses Aya Expanse for translating multilingual prompts and generating detailed image descriptions and [Flux Schnell](https://huggingface.co/black-forest-labs/FLUX.1-schnell) for Image Generation.")
+
+        
+    # Generate Images
+    with gr.TabItem("Aya Vision") as see_with_aya:
+        with gr.Row():
+            with gr.Column():
+                aya_vision_prompt = gr.Textbox(placeholder="Ask anything in our 23 languages ...", label="Input Prompt", lines=3)
+                aya_vision_input_img = gr.Image(label="Input Image", interactive=True, type="filepath")
+                submit_aya_vision = gr.Button(value="Submit", variant="primary")
+                clear_button_aya_vision = gr.ClearButton()
+            with gr.Column():
+                aya_vision_response = gr.Textbox(lines=3,label="Aya Vision's Response", show_copy_button=True, container=True, interactive=False)
+                lang_textbox = gr.Textbox(visible=False)
+
+        with gr.Row():
+            gr.Examples(
+                    examples=[[lang] for lang in AYA_VISION_PROMPT_EXAMPLES.keys()],
+                    inputs=lang_textbox,
+                    outputs=[aya_vision_prompt, aya_vision_input_img],
+                    fn=get_aya_vision_prompt_example,
+                    label="Load example prompt for:",
+                    examples_per_page=25,
+                    run_on_click=True
+                )
+        
+        # increase spacing between examples and Accordion components
+        with gr.Row():
+            pass
+        with gr.Row():
+            pass
+        with gr.Row():
+            pass
+
+        with gr.Row():
+            with gr.Accordion("See Details", open=False):
+                gr.Markdown("This space uses [Aya Vision](https://huggingface.co/CohereForAI/aya-vision-32b) for understanding images.")
+
     
+    # Aya Vision
+    clear_button_aya_vision.click(lambda: None, None, aya_vision_input_img)
+    clear_button_aya_vision.click(lambda: None, None, aya_vision_prompt)
+    clear_button_aya_vision.click(lambda: None, None, aya_vision_response)
+
+    submit_aya_vision.click(
+        get_aya_vision_response,
+        inputs=[aya_vision_prompt, aya_vision_input_img],
+        outputs=[aya_vision_response]
+    )
+
+        
     # Image Generation
     clear_button_img.click(lambda: None, None, input_img_prompt)
     clear_button_img.click(lambda: None, None, generated_img_desc)
@@ -509,5 +560,5 @@ with demo:
     )
     
     demo.load(lambda: secrets.token_hex(16), None, token)
-
+    
 demo.queue(api_open=False, max_size=20, default_concurrency_limit=4).launch(show_api=False, allowed_paths=['/home/user/app'])