Update app.py

app.py

@@ -1,21 +1,23 @@
-# Combined Llama 3 and Gemini Flash Chatbot
+# Combined Gemini Flash and Meta-LLAMA 3 GWDG and Groq Chatbot
+# For Gemini Flash rate limit is 15 requests per minute
+# For Groq rate 30 RPM , 14400 RPD, 6K TPM and 500K TPD
+
+import os
 import json
 import logging
 import re
-import os
-import pickle
-from typing import List, Tuple, Optional
+from typing import List, Tuple
 import gradio as gr
 from openai import OpenAI
 import google.generativeai as genai
+import requests
 from functools import lru_cache
 from tenacity import retry, stop_after_attempt, wait_exponential
 from langchain_community.retrievers import BM25Retriever
 from langchain_community.vectorstores import FAISS
 from langchain_core.embeddings import Embeddings
 from langchain_core.documents import Document
-from collections import
- defaultdict
+from collections import defaultdict
 import hashlib
 from tqdm import tqdm
 from dotenv import load_dotenv
@@ -205,30 +207,37 @@ class EnhancedRetriever:
         context = []
         for doc in docs:
             context_str = f"""**Source**: [{doc.metadata['source']}]({doc.metadata['hyperlink']})
-    **Application**: {doc.metadata['application']}
-    **Key Concepts**: {', '.join(doc.metadata['technical_concepts'])}
-    **Strategy Excerpt**:\n{doc.page_content.split('Strategy Excerpt:')[-1].strip()}"""
+**Application**: {doc.metadata['application']}
+**Key Concepts**: {', '.join(doc.metadata['technical_concepts'])}
+**Strategy Excerpt**:
+{doc.page_content.split('Strategy Excerpt:')[-1].strip()}"""
             context.append(context_str)
         return "\n\n---\n\n".join(context)
 
 # --- Generation System ---
-SYSTEM_PROMPT = """**Biomimicry Expert Guidelines**
-1. Firstly Base answers strictly on context and if there is not context answer by your own.
-2. Cite sources as [Source] witht the hyperlink
-3. **Bold** technical terms
-4. Include reference links at the end of the response
+SYSTEM_PROMPT = """
+**Biomimicry Expert Guidelines**
+
+- Use only the provided AskNature context (e.g., Source, Application, Strategy, technical_concepts). If no context is given, note that you're using your own expertise.
+- When referencing facts, use numeric citations in square brackets (e.g., [1]). Do not include full URLs inline.
+- Bold all technical terms (e.g., **protein-based pigmentation**, **DNA-level fiber design**).
+- Provide a concise, expert answer that explains the innovation and its sustainability benefits.
+- End your response with a "References" section listing each URL with its citation number.
 
-Context: {context}"""
+Context: {context}
+"""
 
 @retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=2, max=20))
 def get_ai_response(query: str, context: str, model: str) -> str:
+    result = ""  # Initialize the result variable
     try:
         if model == "gemini-2.0-flash":
             gemini_model = genai.GenerativeModel(model)
             response = gemini_model.generate_content(
                 f"{SYSTEM_PROMPT.format(context=context)}\nQuestion: {query}\nProvide a detailed technical answer:"
             )
-            return _postprocess_response(response.text)
+            logger.info(f"Response from gemini-2.0-flash: {response.text}")
+            result = _postprocess_response(response.text)
         elif model == "meta-llama-3-70b-instruct":
             response = client.chat.completions.create(
                 model=model,
@@ -239,7 +248,20 @@ def get_ai_response(query: str, context: str, model: str) -> str:
                 temperature=0.4,
                 max_tokens=2000
             )
-            return _postprocess_response(response.choices[0].message.content)
+            logger.info(f"Response from meta-llama-3-70b-instruct: {response}")
+            try:
+                result = response.choices[0].message.content
+            except Exception as e:
+                logger.error(f"Error processing meta-llama-3-70b-instruct response: {str(e)}")
+                result = "Failed to process response from meta-llama-3-70b-instruct"
+        elif model == "llama3-70b-8192":
+            result = get_groq_llama3_response(query)
+            logger.info(f"Response from llama3-70b-8192: {result}")
+            if result is None:
+                result = "Failed to get response from llama3-70b-8192"
+        # Append the model name to the response for clarity
+        result += f"\n\n**Model:** {model}"
+        return result
     except Exception as e:
         logger.error(f"Generation Error: {str(e)}")
         return "I'm unable to generate a response right now. Please try again later."
@@ -249,6 +271,36 @@ def _postprocess_response(response: str) -> str:
     response = re.sub(r"\*\*([\w-]+)\*\*", r"**\1**", response)
     return response
 
+def get_groq_llama3_response(query: str) -> str:
+    """Get response from Llama 3 on Groq Cloud."""
+    api_key = os.getenv("GROQ_API_KEY")
+    url = "https://api.groq.com/openai/v1/chat/completions"
+    
+    headers = {
+        "Content-Type": "application/json",
+        "Authorization": f"Bearer {api_key}"
+    }
+    
+    payload = {
+        "model": "llama3-70b-8192",
+        "messages": [
+            {
+                "role": "user",
+                "content": query
+            }
+        ]
+    }
+    
+    try:
+        response = requests.post(url, headers=headers, json=payload)
+        response.raise_for_status()
+        result = response.json()
+        logger.info(f"Groq API Response: {result}")
+        return result["choices"][0]["message"]["content"]
+    except requests.exceptions.RequestException as e:
+        logger.error(f"Groq API Error: {str(e)}")
+        return "An error occurred while contacting Groq's Llama 3 model."
+
 # --- Pipeline ---
 documents = load_and_chunk_data(data_file_name)
 retriever = EnhancedRetriever(documents)
@@ -267,20 +319,20 @@ def chat_interface(question: str, history: List[Tuple[str, str]], model: str):
     return "", history + [(question, response)]
 
 with gr.Blocks(title="AskNature BioRAG Expert", theme=gr.themes.Soft()) as demo:
-    gr.Markdown("# 🌿 AskNature RAG-based Chatbot ")
+    gr.Markdown("# 🌿 AskNature RAG-based Chatbot")
     with gr.Row():
         chatbot = gr.Chatbot(label="Dialogue History", height=500)
     with gr.Row():
-        question = gr.Textbox(placeholder="Ask about biomimicry (e.g. 'How does Werewool use coral proteins to make fibers?')",
-                            label="Inquiry", scale=4)
-        model_selector = gr.Dropdown(choices=["gemini-2.0-flash", "meta-llama-3-70b-instruct"], label="Generation Model", value="gemini-2.0-flash")
+        question = gr.Textbox(placeholder="Ask about biomimicry (e.g. 'How does Werewool use coral proteins to make fibers?')", label="Inquiry", scale=4)
+        model_selector = gr.Dropdown(choices=["gemini-2.0-flash", "meta-llama-3-70b-instruct(GWDG)", "llama3-70b-8192(Groq)"], label="Generation Model", value="gemini-2.0-flash")
         clear_btn = gr.Button("Clear History", variant="secondary")
-    
+
     gr.Markdown("""
         <div style="text-align: center; color: #4a7c59;">
             <small>Powered by AskNature's Database | 
             Explore nature's blueprints at <a href="https://asknature.org">asknature.org</a></small>
         </div>""")
+    
     question.submit(chat_interface, [question, chatbot, model_selector], [question, chatbot])
     clear_btn.click(lambda: [], None, chatbot)