Update app.py

app.py

@@ -6,7 +6,6 @@ import requests
 from bs4 import BeautifulSoup
 from urllib.parse import urljoin, urlparse
 from groq import Groq
-import asyncio
 import re
 import json
 
@@ -15,23 +14,12 @@ CEREBRAS_API_KEY = os.getenv("CEREBRAS_API_KEY")
 if not CEREBRAS_API_KEY:
     raise ValueError("CEREBRAS_API_KEY environment variable is not set.")
 
-client_cerebras = Cerebras(api_key=CEREBRAS_API_KEY)
-client_groq = Groq()
-
-# --- Rate Limiting ---
-CEREBRAS_REQUESTS_PER_MINUTE = 30
-CEREBRAS_TOKENS_PER_MINUTE = 6000 # using lowest token limit for versatile model
-GROQ_REQUESTS_PER_MINUTE = 30
-GROQ_TOKENS_PER_MINUTE = 15000  # using token limit for tool-use-preview model
-
+GROQ_API_KEY = os.getenv("GROQ_API_KEY")
+if not GROQ_API_KEY:
+    raise ValueError("GROQ_API_KEY environment variable is not set.")
 
-cerebras_request_queue = asyncio.Queue()
-groq_request_queue = asyncio.Queue()
-
-last_cerebras_request_time = 0
-last_groq_request_time = 0
-cerebras_token_count = 0
-groq_token_count = 0
+client_cerebras = Cerebras(api_key=CEREBRAS_API_KEY)
+client_groq = Groq(api_key=GROQ_API_KEY)
 
 # --- Model Rate Limit Info ---
 CHAT_COMPLETION_MODELS_INFO = """
@@ -98,81 +86,28 @@ def extract_text_from_html(html):
     text = soup.get_text(separator=' ', strip=True)
     return text
 
-# --- Asynchronous Rate Limit Logic ---
-
-async def check_cerebras_rate_limit(num_tokens):
-    global last_cerebras_request_time
-    global cerebras_token_count
-    current_time = time.time()
-    elapsed_time = current_time - last_cerebras_request_time
-
-    if elapsed_time < 60 and cerebras_request_queue.qsize() >= CEREBRAS_REQUESTS_PER_MINUTE:
-      await asyncio.sleep(60-elapsed_time)
-
-
-    if elapsed_time < 60 and (cerebras_token_count + num_tokens) > CEREBRAS_TOKENS_PER_MINUTE :
-          time_to_wait = 60 - elapsed_time
-          await asyncio.sleep(time_to_wait)
-
-    cerebras_request_queue.put_nowait(current_time)
-    last_cerebras_request_time = time.time()
-    cerebras_token_count = num_tokens if (elapsed_time > 60) else (cerebras_token_count + num_tokens)
-
-async def check_groq_rate_limit(num_tokens):
-    global last_groq_request_time
-    global groq_token_count
-    current_time = time.time()
-    elapsed_time = current_time - last_groq_request_time
-    if elapsed_time < 60 and groq_request_queue.qsize() >= GROQ_REQUESTS_PER_MINUTE:
-      await asyncio.sleep(60 - elapsed_time)
-
-    if elapsed_time < 60 and (groq_token_count + num_tokens) > GROQ_TOKENS_PER_MINUTE :
-      time_to_wait = 60 - elapsed_time
-      await asyncio.sleep(time_to_wait)
-
-    groq_request_queue.put_nowait(current_time)
-    last_groq_request_time = time.time()
-    groq_token_count = num_tokens if (elapsed_time > 60) else (groq_token_count + num_tokens)
-
 
 # --- Chat Logic with Groq ---
 async def chat_with_groq(user_input, chat_history):
     start_time = time.time()
     try:
         # Prepare chat history for the prompt
-        formatted_history = "\n".join([f"User: {msg[0]}\nAI: {msg[1]}" for msg in chat_history[-5:]])
-         # Check for web scraping command
-        if user_input.lower().startswith("scrape"):
-            parts = user_input.split(maxsplit=1)
-            if len(parts) > 1:
-                url = parts[1].strip()
-                if is_valid_url(url):
-                  html_content = fetch_webpage(url)
-                  if not html_content.startswith("Error"):
-                    webpage_text = extract_text_from_html(html_content)
-                    user_input =  f"The content from the webpage: {webpage_text}. {user_input}"
-                  else:
-                      user_input =  f"{html_content}. {user_input}"
-                else:
-                     user_input = "Invalid URL provided. " + user_input
+        formatted_history = "\n".join([f"User: {msg[0]}\nAI: {msg[1]}" for msg in chat_history[-10:]])
 
         messages = [
-                {"role": "system", "content": f"""You are IntellijMind, an advanced AI designed to assist users with detailed insights, problem-solving, and chain-of-thought reasoning. You have access to various tools to help the user, and can initiate actions when needed. Be creative and inject humor when appropriate. You can use tools to browse the web when instructed with a 'scrape' command followed by a URL.  If there is a request for model info, use the get_model_info function. Current conversation: {formatted_history} Available actions: take_action: 'scrape', parameters: url. Example action: Action: take_action, Parameters: {{"action":"scrape", "url":"https://example.com"}}  """},
-                {"role": "user", "content": user_input}
-            ]
+            {"role": "system", "content": f"""You are IntellijMind, a highly advanced and proactive AI agent. You are designed to assist users in achieving their goals through detailed insights, creative problem-solving, and the use of various tools.  Your objective is to understand the user's intentions, break them into logical steps, and use available tools when needed to achieve the best outcome. Available tools: scrape with a URL, and search_internet with a query. Be creative and inject humor when appropriate. You have access to multiple tools to help the user with their requests. Available actions: take_action: 'scrape', parameters: url, take_action: 'search_internet', parameters: query. Example action: Action: take_action, Parameters: {{"action":"scrape", "url":"https://example.com"}} or  Action: take_action, Parameters: {{"action":"search_internet", "query":"latest news on AI"}} . Current conversation: {formatted_history}"""},
+            {"role": "user", "content": user_input}
+        ]
+
         if user_input.lower() == "model info":
              response = get_model_info()
              return response, "", f"Compute Time: {time.time() - start_time:.2f} seconds", f"Tokens used: {len(user_input.split()) + len(response.split())}"
 
-
-        num_tokens = len(user_input.split())
-        await check_groq_rate_limit(num_tokens)
-
         completion = client_groq.chat.completions.create(
             model="llama3-groq-70b-8192-tool-use-preview",
             messages=messages,
             temperature=1,
-            max_tokens=1024,
+            max_tokens=2048,
             top_p=1,
             stream=True,
             stop=None,
@@ -180,31 +115,37 @@ async def chat_with_groq(user_input, chat_history):
 
         response = ""
         chain_of_thought = ""
+        tool_execution_count = 0
         for chunk in completion:
             if chunk.choices[0].delta and chunk.choices[0].delta.content:
                 content = chunk.choices[0].delta.content
                 response += content
                 if "Chain of Thought:" in content:
                     chain_of_thought += content.split("Chain of Thought:", 1)[-1]
-                # Check if action needs to be taken
+
                 if "Action:" in content:
                     action_match = re.search(r"Action: (\w+), Parameters: (\{.*\})", content)
-                    if action_match:
+                    if action_match and tool_execution_count < 3:  # Limit tool use to avoid infinite loops
+                        tool_execution_count +=1
                         action = action_match.group(1)
                         parameters = json.loads(action_match.group(2))
                         if action == "take_action":
-                           if parameters.get("action") == "scrape":
+                            if parameters.get("action") == "scrape":
                                 url = parameters.get("url")
                                 if is_valid_url(url):
-                                  html_content = fetch_webpage(url)
-                                  if not html_content.startswith("Error"):
-                                    webpage_text = extract_text_from_html(html_content)
-                                    response += f"\nWebpage Content: {webpage_text}\n"
-                                  else:
-                                     response += f"\nError scraping webpage: {html_content}\n"
+                                    html_content = fetch_webpage(url)
+                                    if not html_content.startswith("Error"):
+                                        webpage_text = extract_text_from_html(html_content)
+                                        response += f"\nWebpage Content: {webpage_text}\n"
+                                    else:
+                                       response += f"\nError scraping webpage: {html_content}\n"
                                 else:
                                    response +=  "\nInvalid URL provided.\n"
-
+                            elif parameters.get("action") == "search_internet":
+                                query = parameters.get("query")
+                                response += f"\n Search query: {query}. Note: Search is simulated in this environment. Results may vary. \n"
+                                # Replace the line with a real internet search if you have a search api
+                                response += f"\n Search Results: Mock Results for query: {query} \n"
 
         compute_time = time.time() - start_time
         token_usage = len(user_input.split()) + len(response.split())
@@ -217,7 +158,7 @@ async def chat_with_groq(user_input, chat_history):
 # --- Gradio Interface ---
 def gradio_ui():
     with gr.Blocks() as demo:
-        gr.Markdown("""# 🚀 IntellijMind: The Crazy Agent Chatbot\nExperience the most advanced chatbot for deep insights, chain-of-thought reasoning, and unmatched clarity! Get ready for some proactive action!""")
+        gr.Markdown("""# 🚀 IntellijMind: The Autonomous AI Agent\nExperience the forefront of AI capabilities, where the agent proactively achieves your goals!""")
 
         with gr.Row():
             with gr.Column(scale=6):
@@ -263,7 +204,7 @@ def gradio_ui():
 
         user_input.submit(handle_chat, inputs=[chat_history, user_input], outputs=[chat_history, chain_of_thought_display, compute_time, token_usage_display])
 
-        gr.Markdown("""---\n### 🌟 Features:\n- **Advanced Reasoning**: Chain-of-thought explanations for complex queries.\n- **Proactive Actions**: The agent will take actions without being explicitly asked.\n- **Web Scraping**: The agent will use the scrape command if needed\n- **Humor and Creativity**: Enjoy a more engaging and creative experience.\n- **Real-Time Performance Metrics**: Measure response compute time instantly.\n- **Token Usage Tracking**: Monitor token usage per response for transparency.\n- **Export Chat History**: Save your conversation as a text file for future reference.\n- **User-Friendly Design**: Intuitive chatbot interface with powerful features.\n- **Insightful Chain of Thought**: See the reasoning process behind AI decisions.\n- **Submit on Enter**: Seamless interaction with keyboard support.\n""")
+        gr.Markdown("""---\n### 🌟 Features:\n- **Autonomous Agent**: Proactively pursues your goals.\n- **Advanced Tool Use**: Utilizes multiple tools like web scraping and search.\n- **Dynamic and Creative**: Engages with humor and creative responses.\n- **Enhanced Chat History**: Maintains better context of the conversation.\n- **Real-Time Performance Metrics**: Measure response compute time instantly.\n- **Token Usage Tracking**: Monitor token usage per response for transparency.\n- **Export Chat History**: Save your conversation as a text file for future reference.\n- **User-Friendly Design**: Intuitive chatbot interface with powerful features.\n- **Insightful Chain of Thought**: See the reasoning process behind AI decisions.\n- **Submit on Enter**: Seamless interaction with keyboard support.\n""")
 
     return demo