Update app.py

app.py

@@ -17,7 +17,8 @@ import arxiv
 import scholarly
 import pymed
 import wikipedia
-from newspaper import Article
+#from newspaper import Article  # Removed direct import
+from newspaper3k import Article  # Import from newspaper3k
 import pickle
 import faiss
 import threading
@@ -33,9 +34,9 @@ if not HF_API_KEY:
 
 client = InferenceClient(provider="hf-inference", api_key=HF_API_KEY)
 
-MAIN_LLM_MODEL = "mistralai/Mistral-Nemo-Instruct-2407"
-REASONING_LLM_MODEL = "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B"
-CRITIC_LLM_MODEL = "Qwen/QwQ-32B-Preview"
+MAIN_LLM_MODEL = "meta-llama/Llama-3-70b-instruct"
+REASONING_LLM_MODEL = "anthropic/claude-3-opus-20240229"
+CRITIC_LLM_MODEL = "google/gemini-1.5-pro"
 ENSEMBLE_MODELS = [MAIN_LLM_MODEL, REASONING_LLM_MODEL, CRITIC_LLM_MODEL]
 
 MAX_ITERATIONS = 20
@@ -158,7 +159,7 @@ def tool_search_web(query: str, num_results: int = NUM_RESULTS, safesearch: str
 
             results = [r for r in ddgs.text(**kwargs)]
             if results:
-                return [{"title": r["title"], "snippet": r["body"], "url": r["href"]} for r in results] 
+                return [{"title": r["title"], "snippet": r["body"], "url": r["href"]} for r in results]
             else:
                 if time_filter and "time" in kwargs:
                     del kwargs["time"]
@@ -465,9 +466,8 @@ def tool_identify_focus_areas(prompt: str, insights: list = [],
     return []
 
 def add_to_faiss_index(text: str):
-    """Adds the embedding of the given text to the FAISS index."""
     embedding = document_similarity_model.encode(text, convert_to_tensor=True)
-    embedding_np = embedding.cpu().numpy().reshape(1, -1)  # Ensure 2D array
+    embedding_np = embedding.cpu().numpy().reshape(1, -1)
     if embedding_np.shape[1] != embedding_dim:
         logger.error(f"Embedding dimension mismatch: expected {embedding_dim}, got {embedding_np.shape[1]}")
         return
@@ -475,12 +475,11 @@ def add_to_faiss_index(text: str):
     index.add(embedding_np)
 
 def search_faiss_index(query: str, top_k: int = 5) -> List[str]:
-    """Searches the FAISS index for the most similar texts to the query."""
     query_embedding = document_similarity_model.encode(query, convert_to_tensor=True)
     query_embedding_np = query_embedding.cpu().numpy().reshape(1, -1)
     faiss.normalize_L2(query_embedding_np)
     distances, indices = index.search(query_embedding_np, top_k)
-    return indices[0].tolist() # Return indices
+    return indices[0].tolist()
 
 def filter_results(search_results, prompt, previous_snippets=None):
     if not main_similarity_model or not search_results:
@@ -507,7 +506,6 @@ def filter_results(search_results, prompt, previous_snippets=None):
                 result['relevance_score'] = cosine_score
                 filtered_results.append(result)
                 seen_snippets.add(result['snippet'])
-                # Add snippet to FAISS index
                 add_to_faiss_index(result['snippet'])