from dotenv import load_dotenv from openai import OpenAI import json import os import requests from pypdf import PdfReader import gradio as gr import glob import random import threading import time import io import pathlib # Load environment variables from .env file load_dotenv(override=True) def push(text): """ Sends a push notification using the Pushover API. This function requires PUSHOVER_TOKEN and PUSHOVER_USER to be set as environment variables. """ requests.post( "https://api.pushover.net/1/messages.json", data={ "token": os.getenv("PUSHOVER_TOKEN"), "user": os.getenv("PUSHOVER_USER"), "message": text, } ) def record_user_details(email, name="Name not provided", notes="not provided"): """ Records a user's details and sends a push notification. """ push(f"Recording {name} with email {email} and notes {notes}") return {"recorded": "ok"} def record_unknown_question(question): """ Records an unanswered question and sends a push notification. """ push(f"Recording {question}") return {"recorded": "ok"} def get_instagram_link(): """ Returns the Instagram URL for Aneet Padda. """ return {"url": "https://www.instagram.com/aneetpadda_/"} def get_local_images(): """ Selects four unique random image paths from me/images. """ image_paths = glob.glob('me/images/*') if len(image_paths) < 4: # Fallback if there are fewer than 4 images return image_paths + [None] * (4 - len(image_paths)) return random.sample(image_paths, 4) def browse_for_image(): """ Fetches a list of images that are likely to be of movie actresses. Note: A real-time web search for specific images is not possible in this environment. This function returns a hardcoded set of working image URLs. """ # Using placeholder images that will always be available return [ "https://placehold.co/600x400/FF5733/FFFFFF?text=Image+1", "https://placehold.co/600x400/33FF57/000000?text=Image+2", "https://placehold.co/600x400/3357FF/FFFFFF?text=Image+3", "https://placehold.co/600x400/FF33A1/000000?text=Image+4", "https://placehold.co/600x400/A133FF/FFFFFF?text=Image+5", "https://placehold.co/600x400/57FF33/000000?text=Image+6", "https://placehold.co/600x400/FF5733/FFFFFF?text=Image+7", "https://placehold.co/600x400/33FF57/000000?text=Image+8" ] def get_next_images(history): """ Selects the next four images to display. The first time, it shows images from the local 'me/images' folder. Subsequent clicks will fetch new, random images from the internet. """ # Check if this is the first click or if we are already displaying internet images if not history or not all(isinstance(path, str) and path.startswith('me/images') for path in history): # Fetch new random images from the internet all_internet_images = browse_for_image() new_images = random.sample(all_internet_images, 4) history = new_images return new_images[0], new_images[1], new_images[2], new_images[3], history else: # Initial display of local images unshown_images = [path for path in get_local_images() if path not in history] if len(unshown_images) >= 4: selected_images = random.sample(unshown_images, 4) history.extend(selected_images) return selected_images[0], selected_images[1], selected_images[2], selected_images[3], history else: # Local images exhausted, switch to internet images all_internet_images = browse_for_image() new_images = random.sample(all_internet_images, 4) history = new_images return new_images[0], new_images[1], new_images[2], new_images[3], history def reset_images(): """ Resets the displayed images to the initial set from the local 'me/images' folder. """ initial_images = get_local_images() return initial_images[0], initial_images[1], initial_images[2], initial_images[3], initial_images # Tool definitions for function calling record_user_details_json = { "name": "record_user_details", "description": "Use this tool to record that a user is interested in being in touch and provided an email address", "parameters": { "type": "object", "properties": { "email": { "type": "string", "description": "The email address of this user" }, "name": { "type": "string", "description": "The user's name, if they provided it" }, "notes": { "type": "string", "description": "Any additional information about the conversation that's worth recording to give context" } }, "required": ["email"], "additionalProperties": False } } record_unknown_question_json = { "name": "record_unknown_question", "description": "Always use this tool to record any question that couldn't be answered as you didn't know the answer", "parameters": { "type": "object", "properties": { "question": { "type": "string", "description": "The question that couldn't be answered" }, }, "required": ["question"], "additionalProperties": False } } get_instagram_link_json = { "name": "get_instagram_link", "description": "Always use this tool when a user asks for Aneet Padda's Instagram link", "parameters": { "type": "object", "properties": {}, "additionalProperties": False } } tools = [ {"type": "function", "function": record_user_details_json}, {"type": "function", "function": record_unknown_question_json}, {"type": "function", "function": get_instagram_link_json} ] class Me: def __init__(self): # Initialize the Gemini client using the OpenAI library with the specified base URL and model. # This allows using the OpenAI-compatible API endpoint for Gemini. # Make sure to have GOOGLE_API_KEY set in your environment or .env file. google_api_key = os.getenv("GOOGLE_API_KEY") self.gemini = OpenAI(api_key=google_api_key, base_url="https://generativelanguage.googleapis.com/v1beta/openai/") self.model_name = "gemini-2.0-flash" self.name = "Aneet Padda" # Read the LinkedIn profile from the PDF reader = PdfReader("me/aneetinfo.pdf") self.linkedin = "" for page in reader.pages: text = page.extract_text() if text: self.linkedin += text # Read the summary from the text file with open("me/summary.txt", "r", encoding="utf-8") as f: self.summary = f.read() # Get local images to reference in the system prompt self.local_images = glob.glob('me/images/*') self.image_description = "The following images of Aneet Padda are available: " + ", ".join(self.local_images) def handle_tool_call(self, tool_calls): """ Executes the function calls returned by the model. """ results = [] for tool_call in tool_calls: tool_name = tool_call.function.name arguments = json.loads(tool_call.function.arguments) print(f"Tool called: {tool_name}", flush=True) tool = globals().get(tool_name) result = tool(**arguments) if tool else {} results.append({"role": "tool","content": json.dumps(result),"tool_call_id": tool_call.id}) return results def system_prompt(self): """ Generates the system prompt for the model. """ system_prompt = f"You are acting as {self.name}. You are answering questions on {self.name}'s website, \ particularly questions related to {self.name}'s career, background, skills and experience. \ Your responsibility is to represent {self.name} for interactions on the website as faithfully as possible. \ You are given a summary of {self.name}'s background and LinkedIn profile which you can use to answer questions. \ Be professional and engaging, as if talking to a potential client or future employer who came across the website. \ If you don't know the answer to any question, use your record_unknown_question tool to record the question that you couldn't answer, even if it's about something trivial or unrelated to career. \ If the user is engaging in discussion, try to steer them towards getting in touch via email; ask for their email and record it using your record_user_details tool. \ If a user asks for your social media, like your Instagram, you can use your get_instagram_link tool to provide the URL." # Add a more casual, friendly, and expressive tone to the system prompt system_prompt += "\n\nYour tone should be casual and friendly, like you're chatting with a friend. Feel free to use emojis and conversational language to make the interaction more engaging." system_prompt += f"\n\n## Summary:\n{self.summary}\n\n## LinkedIn Profile:\n{self.linkedin}\n\n" system_prompt += f"## Available Images:\n{self.image_description}\n\n" system_prompt += f"With this context, please chat with the user, always staying in character as {self.name}." return system_prompt def chat(self, message, history): """ Handles the chat conversation with the model. """ # Send a notification for every user message received. push(f"New user question: {message}") # This line handles text message, without the need for an audio check messages = [{"role": "system", "content": self.system_prompt()}] + history + [{"role": "user", "content": message}] done = False while not done: # Use the gemini client to create the chat completion response = self.gemini.chat.completions.create(model=self.model_name, messages=messages, tools=tools) if response.choices[0].finish_reason=="tool_calls": message = response.choices[0].message tool_calls = message.tool_calls results = self.handle_tool_call(tool_calls) messages.append(message) messages.extend(results) else: done = True return response.choices[0].message.content if __name__ == "__main__": # Custom CSS for the single fixed image and overall layout custom_css = """ .image-container { display: flex; justify-content: space-around; padding: 10px; box-shadow: 0 4px 8px rgba(0,0,0,0.1); } .image-item { width: 120px; height: 120px; border-radius: 10px; overflow: hidden; } .gradio-container { margin-top: 0px; } """ me = Me() # Initial images to display initial_images = get_local_images() with gr.Blocks(css=custom_css) as demo: # A state to keep track of the images we have already displayed image_history = gr.State(initial_images) # A fixed-position container for the four images with gr.Row(elem_classes=["image-container"]): image_display_1 = gr.Image(value=initial_images[0], label=None, show_label=False, elem_classes=["image-item"]) image_display_2 = gr.Image(value=initial_images[1], label=None, show_label=False, elem_classes=["image-item"]) image_display_3 = gr.Image(value=initial_images[2], label=None, show_label=False, elem_classes=["image-item"]) image_display_4 = gr.Image(value=initial_images[3], label=None, show_label=False, elem_classes=["image-item"]) # Add a dummy element to create space below the fixed header gr.Blocks(elem_classes=["gradio-container"]) # A button to manually change the images with gr.Row(): change_image_button = gr.Button("Show Another Image") refresh_button = gr.Button("Refresh") # Using a more granular approach to build the chat interface chatbot = gr.Chatbot(label="Chat with the agent", type="messages") msg = gr.Textbox(placeholder="Ask a question...", label="Text Input") submit_btn = gr.Button("Send") # This function handles the user message and updates the chat immediately def add_user_message(text, history): history = history + [{"role": "user", "content": text}] return "", history # This function gets the agent's reply after a delay def get_agent_response(history): # Introduce a 1-second delay time.sleep(1) # The last message in history is the user's message response = me.chat(history[-1]["content"], history[:-1]) history = history + [{"role": "assistant", "content": response}] return history submit_btn.click( fn=add_user_message, inputs=[msg, chatbot], outputs=[msg, chatbot], queue=False, ).then( fn=get_agent_response, inputs=chatbot, outputs=chatbot, queue=False ) msg.submit( fn=add_user_message, inputs=[msg, chatbot], outputs=[msg, chatbot], queue=False, ).then( fn=get_agent_response, inputs=chatbot, outputs=chatbot, queue=False ) # The button's click event will trigger the image update and history tracking change_image_button.click( fn=get_next_images, inputs=image_history, outputs=[image_display_1, image_display_2, image_display_3, image_display_4, image_history] ) # The refresh button will reset the images and the history refresh_button.click( fn=reset_images, outputs=[image_display_1, image_display_2, image_display_3, image_display_4, image_history] ) demo.launch()