Updated app.py

app.py

@@ -1,3 +1,277 @@
+# """
+# References:
+# - https://medium.com/@turna.fardousi/building-a-multimodal-chatbot-with-gemini-api-8015bfbee538
+# """
+
+# import os
+# import time
+# from typing import List, Tuple, Optional
+# import google.generativeai as genai
+# import gradio as gr
+# from PIL import Image
+# from dotenv import load_dotenv
+
+# load_dotenv()
+
+# GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
+
+# # ============================== Stable - START ==============================
+# TITLE = """<h1 align="center">🎮Chat with Gemini 1.5🔥 -- Beta Preview</h1>"""
+# NOTICE = """
+# Notices 📜:
+# - This app is still in development
+# - Some features may not work as expected
+# """
+# ABOUT = """
+# Updates (2024-8-12): Created the App
+
+# Info:
+# - Model: Gemini 1.5 Flash
+# """
+# ERRORS = """
+# Known errors ⚠️:
+# """
+# FUTURE_IMPLEMENTATIONS = """
+# To be implemented 🚀:
+# - Select other Gemini / Gemma models
+# - Upload files
+# - More tools other than web search
+# """
+# IMAGE_WIDTH = 512
+
+# def preprocess_stop_sequences(stop_sequences: str) -> Optional[List[str]]:
+#     return [seq.strip() for seq in stop_sequences.split(",")] if stop_sequences else None
+
+# def preprocess_image(image: Image.Image) -> Image.Image:
+#     image_height = int(image.height * IMAGE_WIDTH / image.width)
+#     return image.resize((IMAGE_WIDTH, image_height))
+
+# def user(text_prompt: str, chatbot: List[Tuple[str, str]]):
+#     return "", chatbot + [[text_prompt, None]]
+
+# def bot(
+#     google_key: str,
+#     image_prompt: Optional[Image.Image],
+#     temperature: float,
+#     max_output_tokens: int,
+#     stop_sequences: str,
+#     top_k: int,
+#     top_p: float,
+#     chatbot: List[Tuple[str, str]]
+# ):
+#     google_key = google_key or GEMINI_API_KEY
+#     if not google_key:
+#         raise ValueError("GOOGLE_API_KEY is not set. Please set it up.")
+
+#     text_prompt = chatbot[-1][0]
+#     genai.configure(api_key=google_key)
+#     generation_config = genai.types.GenerationConfig(
+#         temperature=temperature,
+#         max_output_tokens=max_output_tokens,
+#         stop_sequences=preprocess_stop_sequences(stop_sequences),
+#         top_k=top_k,
+#         top_p=top_p,
+#     )
+
+#     model_name = "gemini-1.5-flash" # if image_prompt is None else "gemini-pro-vision"
+#     model = genai.GenerativeModel(model_name)
+#     inputs = [text_prompt] if image_prompt is None else [text_prompt, preprocess_image(image_prompt)]
+
+#     response = model.generate_content(inputs, stream=True, generation_config=generation_config)
+#     response.resolve()
+
+#     chatbot[-1][1] = ""
+#     for chunk in response:
+#         for i in range(0, len(chunk.text), 10):
+#             chatbot[-1][1] += chunk.text[i:i + 10]
+#             time.sleep(0.01)
+#             yield chatbot
+
+# google_key_component = gr.Textbox(
+#     label = "GOOGLE API KEY",
+#     type = "password",
+#     placeholder = "...",
+#     visible = GEMINI_API_KEY is None
+# )
+
+# image_prompt_component = gr.Image(
+#     type = "pil",
+#     label = "Image"
+# )
+# chatbot_component = gr.Chatbot(
+#     # label = 'Gemini',
+#     bubble_full_width = False
+# )
+# text_prompt_component = gr.Textbox(
+#     placeholder = "Chat with Gemini",
+#     label = "Ask me anything and press Enter"
+# )
+# run_button_component = gr.Button(
+#     "Run"
+# )
+# temperature_component = gr.Slider(
+#     minimum = 0,
+#     maximum = 1.0,
+#     value = 0.5,
+#     step = 0.05,
+#     label = "Temperature"
+# )
+# max_output_tokens_component = gr.Slider(
+#     minimum = 1,
+#     maximum = 8192,
+#     value = 4096,
+#     step = 1,
+#     label = "Max Output Tokens"
+# )
+# stop_sequences_component = gr.Textbox(
+#     label = "Add stop sequence",
+#     placeholder = "STOP, END"
+# )
+# top_k_component = gr.Slider(
+#     minimum = 1,
+#     maximum = 40,
+#     value = 32,
+#     step = 1,
+#     label = "Top-K"
+# )
+# top_p_component = gr.Slider(
+#     minimum = 0,
+#     maximum = 1,
+#     value = 1,
+#     step = 0.01,
+#     label = "Top-P"
+# )
+
+# user_inputs = [
+#     text_prompt_component,
+#     chatbot_component
+# ]
+# bot_inputs = [
+#     google_key_component,
+#     image_prompt_component,
+#     temperature_component,
+#     max_output_tokens_component,
+#     stop_sequences_component,
+#     top_k_component,
+#     top_p_component,
+#     chatbot_component
+# ]
+# # ============================== Stable - END ==============================
+
+# # ============================== Nightly - START ==============================
+# """
+# References:
+# - https://medium.com/latinxinai/simple-chatbot-gradio-google-gemini-api-4ce02fbaf09f
+# """
+# GEMINI_API_KEY_NIGHTLY = os.getenv("GEMINI_API_KEY_NIGHTLY")
+# model_nightly_name = "gemini-1.5-flash"
+
+# def transform_history(history):
+#     new_history = []
+#     for chat in history:
+#         new_history.append({"parts": [{"text": chat[0]}], "role": "user"})
+#         new_history.append({"parts": [{"text": chat[1]}], "role": "model"})
+#     return new_history
+
+# def response(message, history):
+#     genai.configure(api_key=GEMINI_API_KEY_NIGHTLY)
+#     model_nightly = genai.GenerativeModel(model_nightly_name)
+
+#     global chat
+#     chat = model_nightly.start_chat(history=[])
+#     chat.history = transform_history(history)
+#     response = chat.send_message(message)
+#     response.resolve()
+
+#     for i in range(len(response.text)):
+#         time.sleep(0.05)
+#         yield response.text[:i+1]
+
+# # ============================== Nightly - END ==============================
+
+# with gr.Blocks(theme = gr.themes.Soft()) as demo:
+# # ============================== Stable - START ==============================
+#     with gr.Tab("Chat with Gemini 1.5 Flash"):
+#         gr.HTML(TITLE)
+#         with gr.Row():
+#             gr.Markdown(NOTICE)
+#             gr.Markdown(ABOUT)
+#             gr.Markdown(ERRORS)
+#             gr.Markdown(FUTURE_IMPLEMENTATIONS)
+#         with gr.Column():
+#             google_key_component.render()
+#             with gr.Row():
+#                 image_prompt_component.render()
+#                 chatbot_component.render()
+#             text_prompt_component.render()
+#             run_button_component.render()
+#             with gr.Accordion("Parameters", open=False):
+#                 temperature_component.render()
+#                 max_output_tokens_component.render()
+#                 stop_sequences_component.render()
+#                 with gr.Accordion("Advanced", open=False):
+#                     top_k_component.render()
+#                     top_p_component.render()
+
+#         run_button_component.click(
+#             fn = user,
+#             inputs = user_inputs,
+#             outputs = [
+#                 text_prompt_component,
+#                 chatbot_component
+#             ],
+#             queue = False
+#         ).then(
+#             fn = bot,
+#             inputs = bot_inputs,
+#             outputs = [
+#                 chatbot_component
+#             ]
+#         )
+#         text_prompt_component.submit(
+#             fn = user,
+#             inputs = user_inputs,
+#             outputs = [
+#                 text_prompt_component,
+#                 chatbot_component
+#             ],
+#             queue = False
+#         ).then(
+#             fn = bot,
+#             inputs = bot_inputs,
+#             outputs = [
+#                 chatbot_component
+#             ]
+#         )
+# # ============================== Stable - END ==============================
+
+#     with gr.Tab("Chat with Gemma 2"):
+#         gr.HTML(
+#             """
+#             <h1 align="center">Still in development</h1>
+#             """
+#         )
+
+# # ============================== Nightly - START ==============================
+#     with gr.Tab("Nightly -- Chat with Gemini 1.5"):
+#         gr.HTML(
+#             """
+#             <h1 align="center">This section will test out the next version of the stable version.</h1>
+#             """
+#         )
+#         gr.ChatInterface(
+#             response,
+#             chatbot = gr.Chatbot(height=600),
+#             title = 'Chat with Gemini 1.5',
+#             retry_btn = "Retry",
+#             undo_btn = "Undo",
+#             clear_btn = "Clear",
+#             fill_height = True
+#         )
+# # ============================== Nightly - END ==============================
+
+# demo.queue().launch(debug = True, show_error = True)
+
 """
 References:
 - https://medium.com/@turna.fardousi/building-a-multimodal-chatbot-with-gemini-api-8015bfbee538
@@ -5,6 +279,7 @@ References:
 
 import os
 import time
+import PIL.Image
 from typing import List, Tuple, Optional
 import google.generativeai as genai
 import gradio as gr
@@ -39,16 +314,27 @@ To be implemented 🚀:
 """
 IMAGE_WIDTH = 512
 
-def preprocess_stop_sequences(stop_sequences: str) -> Optional[List[str]]:
+
+def preprocess_stop_sequences(
+    stop_sequences: str
+) -> Optional[List[str]]:
     return [seq.strip() for seq in stop_sequences.split(",")] if stop_sequences else None
 
-def preprocess_image(image: Image.Image) -> Image.Image:
+
+def preprocess_image(
+    image: Image.Image
+) -> Image.Image:
     image_height = int(image.height * IMAGE_WIDTH / image.width)
     return image.resize((IMAGE_WIDTH, image_height))
 
-def user(text_prompt: str, chatbot: List[Tuple[str, str]]):
+
+def user(
+    text_prompt: str,
+    chatbot: List[Tuple[str, str]]
+):
     return "", chatbot + [[text_prompt, None]]
 
+
 def bot(
     google_key: str,
     image_prompt: Optional[Image.Image],
@@ -73,11 +359,17 @@ def bot(
         top_p=top_p,
     )
 
-    model_name = "gemini-1.5-flash" # if image_prompt is None else "gemini-pro-vision"
+    model_name = "gemini-1.5-flash"
     model = genai.GenerativeModel(model_name)
-    inputs = [text_prompt] if image_prompt is None else [text_prompt, preprocess_image(image_prompt)]
-    
-    response = model.generate_content(inputs, stream=True, generation_config=generation_config)
+    inputs = [text_prompt] if image_prompt is None else [
+        text_prompt, preprocess_image(image_prompt)
+    ]
+
+    response = model.generate_content(
+        inputs,
+        stream=True,
+        generation_config=generation_config
+    )
     response.resolve()
 
     chatbot[-1][1] = ""
@@ -87,74 +379,59 @@ def bot(
             time.sleep(0.01)
             yield chatbot
 
+
 google_key_component = gr.Textbox(
-    label = "GOOGLE API KEY",
-    type = "password",
-    placeholder = "...",
-    visible = GEMINI_API_KEY is None
+    label="GOOGLE API KEY",
+    type="password",
+    placeholder="...",
+    visible=GEMINI_API_KEY is None
 )
 
-image_prompt_component = gr.Image(
-    type = "pil",
-    label = "Image"
-)
-chatbot_component = gr.Chatbot(
-    # label = 'Gemini',
-    bubble_full_width = False
-)
+image_prompt_component = gr.Image(type="pil", label="Image")
+chatbot_component = gr.Chatbot(bubble_full_width=False)
 text_prompt_component = gr.Textbox(
-    placeholder = "Chat with Gemini",
-    label = "Ask me anything and press Enter"
-)
-run_button_component = gr.Button(
-    "Run"
+    placeholder="Chat with Gemini",
+    label="Ask me anything and press Enter"
 )
+run_button_component = gr.Button("Run")
 temperature_component = gr.Slider(
-    minimum = 0,
-    maximum = 1.0,
-    value = 0.5,
-    step = 0.05,
-    label = "Temperature"
+    minimum=0,
+    maximum=1.0,
+    value=0.5,
+    step=0.05,
+    label="Temperature"
 )
 max_output_tokens_component = gr.Slider(
-    minimum = 1,
-    maximum = 8192,
-    value = 4096,
-    step = 1,
-    label = "Max Output Tokens"
+    minimum=1,
+    maximum=8192,
+    value=4096,
+    step=1,
+    label="Max Output Tokens"
 )
 stop_sequences_component = gr.Textbox(
-    label = "Add stop sequence",
-    placeholder = "STOP, END"
+    label="Add stop sequence",
+    placeholder="STOP, END"
 )
 top_k_component = gr.Slider(
-    minimum = 1,
-    maximum = 40,
-    value = 32,
-    step = 1,
-    label = "Top-K"
+    minimum=1,
+    maximum=40,
+    value=32,
+    step=1,
+    label="Top-K"
 )
 top_p_component = gr.Slider(
-    minimum = 0,
-    maximum = 1,
-    value = 1,
-    step = 0.01,
-    label = "Top-P"
+    minimum=0,
+    maximum=1,
+    value=1,
+    step=0.01,
+    label="Top-P"
 )
 
-user_inputs = [
-    text_prompt_component,
-    chatbot_component
-]
+user_inputs = [text_prompt_component, chatbot_component]
 bot_inputs = [
-    google_key_component,
-    image_prompt_component,
-    temperature_component,
-    max_output_tokens_component,
-    stop_sequences_component,
-    top_k_component,
-    top_p_component,
-    chatbot_component
+    google_key_component, image_prompt_component, temperature_component,
+    max_output_tokens_component, stop_sequences_component, top_k_component,
+    top_p_component, chatbot_component
 ]
 # ============================== Stable - END ==============================
 
@@ -166,31 +443,74 @@ References:
 GEMINI_API_KEY_NIGHTLY = os.getenv("GEMINI_API_KEY_NIGHTLY")
 model_nightly_name = "gemini-1.5-flash"
 
+
 def transform_history(history):
     new_history = []
-    for chat in history:
-        new_history.append({"parts": [{"text": chat[0]}], "role": "user"})
-        new_history.append({"parts": [{"text": chat[1]}], "role": "model"})
+    for user_msg, model_msg in history:
+        new_history.append({"role": "user", "parts": [{"text": user_msg}]})
+        new_history.append({"role": "model", "parts": [{"text": model_msg}]})
     return new_history
 
+
+def upload_to_gemini(path, mime_type=None):
+    file = genai.upload_file(path, mime_type=mime_type)
+    print(f"Uploaded file '{file.display_name}' as: {file.uri}")
+    return file.uri
+
+
 def response(message, history):
+    message_text = message["text"]
+    message_files = message["file"]
+    if message_files and isinstance(message_files, list):
+        image_uris = [upload_to_gemini(file_path) for file_path in message_files]
+        message_content = [{"text": message_text}] + image_uris
+    else:
+        message_content = {"text": message_text}
     genai.configure(api_key=GEMINI_API_KEY_NIGHTLY)
-    model_nightly = genai.GenerativeModel(model_nightly_name)
+    model_nightly = genai.GenerativeModel(
+        model_nightly_name,
+        safety_settings=[
+            {
+                "category": "HARM_CATEGORY_HARASSMENT",
+                "threshold": "BLOCK_NONE"
+            },
+            {
+                "category": "HARM_CATEGORY_HATE_SPEECH",
+                "threshold": "BLOCK_NONE"
+            },
+            {
+                "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+                "threshold": "BLOCK_NONE"
+            },
+            {
+                "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
+                "threshold": "BLOCK_NONE"
+            },
+        ],
+        generation_config={
+            "temperature": 1,
+            "top_p": 0.95,
+            "top_k": 64,
+            "max_output_tokens": 8192,
+            "response_mime_type": "text/plain",
+        }
+    )
 
     global chat
     chat = model_nightly.start_chat(history=[])
     chat.history = transform_history(history)
-    response = chat.send_message(message)
+    response = chat.send_message(message_content)
     response.resolve()
 
     for i in range(len(response.text)):
         time.sleep(0.05)
-        yield response.text[:i+1]
+        yield response.text[:i + 1]
+
 
 # ============================== Nightly - END ==============================
 
-with gr.Blocks(theme = gr.themes.Soft()) as demo:
-# ============================== Stable - START ==============================
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    # ============================== Stable - START ==============================
     with gr.Tab("Chat with Gemini 1.5 Flash"):
         gr.HTML(TITLE)
         with gr.Row():
@@ -214,60 +534,43 @@ with gr.Blocks(theme = gr.themes.Soft()) as demo:
                     top_p_component.render()
 
         run_button_component.click(
-            fn = user,
-            inputs = user_inputs,
-            outputs = [
-                text_prompt_component,
-                chatbot_component
-            ],
-            queue = False
+            fn=user,
+            inputs=user_inputs,
+            outputs=[text_prompt_component, chatbot_component],
+            queue=False
         ).then(
-            fn = bot,
-            inputs = bot_inputs,
-            outputs = [
-                chatbot_component
-            ]
+            fn=bot,
+            inputs=bot_inputs,
+            outputs=[chatbot_component]
         )
         text_prompt_component.submit(
-            fn = user,
-            inputs = user_inputs,
-            outputs = [
-                text_prompt_component,
-                chatbot_component
-            ],
-            queue = False
+            fn=user,
+            inputs=user_inputs,
+            outputs=[text_prompt_component, chatbot_component],
+            queue=False
         ).then(
-            fn = bot,
-            inputs = bot_inputs,
-            outputs = [
-                chatbot_component
-            ]
+            fn=bot,
+            inputs=bot_inputs,
+            outputs=[chatbot_component]
         )
-# ============================== Stable - END ==============================
-    
+    # ============================== Stable - END ==============================
+
     with gr.Tab("Chat with Gemma 2"):
-        gr.HTML(
-            """
-            <h1 align="center">Still in development</h1>
-            """
-        )
+        gr.HTML("""<h1 align="center">Still in development</h1>""")
 
-# ============================== Nightly - START ==============================
+    # ============================== Nightly - START ==============================
     with gr.Tab("Nightly -- Chat with Gemini 1.5"):
-        gr.HTML(
-            """
-            <h1 align="center">This section will test out the next version of the stable version.</h1>
-            """
-        )
+        gr.HTML("""<h1 align="center">This section will test out the next version of the stable version.</h1>""")
         gr.ChatInterface(
             response,
-            chatbot = gr.Chatbot(height=600),
-            title = 'Chat with Gemini 1.5',
-            retry_btn = "Retry",
-            undo_btn = "Undo",
-            clear_btn = "Clear",
-            fill_height = True
+            chatbot=gr.Chatbot(height=600),
+            title='Chat with Gemini 1.5',
+            retry_btn="Retry",
+            undo_btn="Undo",
+            clear_btn="Clear",
+            fill_height=True,
+            multimodal=True
         )
 # ============================== Nightly - END ==============================
 
-demo.queue().launch(debug = True, show_error = True)
+demo.queue().launch(debug=True, show_error=True, share=True)