Update app.py

app.py

@@ -3,7 +3,6 @@ import multiprocessing
 import concurrent.futures
 from langchain.document_loaders import TextLoader, DirectoryLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
-from langchain_community.vectorstores import FAISS
 from langchain_community.embeddings import HuggingFaceEmbeddings
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, BitsAndBytesConfig
 from datetime import datetime
@@ -14,15 +13,27 @@ from threading import Thread
 from transformers.agents import Tool, HfEngine, ReactJsonAgent
 from huggingface_hub import InferenceClient
 import logging
+import torch
 
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
+try:
+    from langchain_community.vectorstores import FAISS
+except ImportError:
+    logger.error("Failed to import FAISS. Make sure it's installed correctly.")
+    logger.info("You can try: pip install faiss-cpu --no-cache")
+    FAISS = None
+
 class DocumentRetrievalAndGeneration:
     def __init__(self, embedding_model_name, lm_model_id, data_folder):
         self.all_splits = self.load_documents(data_folder)
         self.embeddings = HuggingFaceEmbeddings(model_name=embedding_model_name)
-        self.vectordb = self.create_faiss_index()
+        if FAISS is not None:
+            self.vectordb = self.create_faiss_index()
+        else:
+            logger.warning("FAISS is not available. Vector search functionality will be limited.")
+            self.vectordb = None
         self.tokenizer, self.model = self.initialize_llm(lm_model_id)
         self.retriever_tool = self.create_retriever_tool()
         self.agent = self.create_agent()
@@ -37,6 +48,9 @@ class DocumentRetrievalAndGeneration:
         return all_splits
 
     def create_faiss_index(self):
+        if FAISS is None:
+            logger.error("FAISS is not available. Cannot create index.")
+            return None
         return FAISS.from_documents(self.all_splits, self.embeddings)
 
     def initialize_llm(self, model_id):
@@ -72,6 +86,8 @@ class DocumentRetrievalAndGeneration:
                 self.vectordb = vectordb
 
             def forward(self, query: str) -> str:
+                if self.vectordb is None:
+                    return "Vector database is not available. Cannot perform retrieval."
                 docs = self.vectordb.similarity_search(query, k=3)
                 return "\nRetrieved documents:\n" + "".join(
                     [f"===== Document {str(i)} =====\n" + doc.page_content for i, doc in enumerate(docs)]
@@ -126,53 +142,63 @@ Question:
         return response
 
     def save_index(self, path):
-        self.vectordb.save_local(path)
+        if self.vectordb is not None:
+            self.vectordb.save_local(path)
+        else:
+            logger.warning("Vector database is not available. Cannot save index.")
 
     def load_index(self, path):
-        self.vectordb = FAISS.load_local(path, self.embeddings)
+        if FAISS is not None:
+            self.vectordb = FAISS.load_local(path, self.embeddings)
+        else:
+            logger.warning("FAISS is not available. Cannot load index.")
 
 if __name__ == "__main__":
     embedding_model_name = 'thenlper/gte-small'
     lm_model_id = "meta-llama/Meta-Llama-3.1-8B-Instruct"
     data_folder = 'sample_embedding_folder2'
 
-    doc_retrieval_gen = DocumentRetrievalAndGeneration(embedding_model_name, lm_model_id, data_folder)
-
-    # Save the index for future use
-    doc_retrieval_gen.save_index("faiss_index")
-
-    def launch_interface():
-        css_code = """
-            .gradio-container {
-                background-color: #daccdb;
-            }
-            button {
-                background-color: #927fc7;
-                color: black;
-                border: 1px solid black;
-                padding: 10px;
-                margin-right: 10px;
-                font-size: 16px;
-                font-weight: bold;
-            }
-        """
-        EXAMPLES = [
-            "On which devices can the VIP and CSI2 modules operate simultaneously?", 
-            "I'm using Code Composer Studio 5.4.0.00091 and enabled FPv4SPD16 floating point support for CortexM4 in TDA2. However, after building the project, the .asm file shows --float_support=vfplib instead of FPv4SPD16. Why is this happening?", 
-            "Could you clarify the maximum number of cameras that can be connected simultaneously to the video input ports on the TDA2x SoC, considering it supports up to 10 multiplexed input ports and includes 3 dedicated video input modules?"
-        ]
-
-        interface = gr.Interface(
-            fn=doc_retrieval_gen.qa_infer_gradio,
-            inputs=[gr.Textbox(label="QUERY", placeholder="Enter your query here")],
-            allow_flagging='never',
-            examples=EXAMPLES,
-            cache_examples=False,
-            outputs=[gr.Textbox(label="RESPONSE"), gr.Textbox(label="RELATED QUERIES")],
-            css=css_code,
-            title="TI E2E FORUM Multi-Agent RAG"
-        )
+    try:
+        doc_retrieval_gen = DocumentRetrievalAndGeneration(embedding_model_name, lm_model_id, data_folder)
 
-        interface.launch(debug=True)
+        # Save the index for future use
+        doc_retrieval_gen.save_index("faiss_index")
 
-    launch_interface()
+        def launch_interface():
+            css_code = """
+                .gradio-container {
+                    background-color: #daccdb;
+                }
+                button {
+                    background-color: #927fc7;
+                    color: black;
+                    border: 1px solid black;
+                    padding: 10px;
+                    margin-right: 10px;
+                    font-size: 16px;
+                    font-weight: bold;
+                }
+            """
+            EXAMPLES = [
+                "On which devices can the VIP and CSI2 modules operate simultaneously?", 
+                "I'm using Code Composer Studio 5.4.0.00091 and enabled FPv4SPD16 floating point support for CortexM4 in TDA2. However, after building the project, the .asm file shows --float_support=vfplib instead of FPv4SPD16. Why is this happening?", 
+                "Could you clarify the maximum number of cameras that can be connected simultaneously to the video input ports on the TDA2x SoC, considering it supports up to 10 multiplexed input ports and includes 3 dedicated video input modules?"
+            ]
+
+            interface = gr.Interface(
+                fn=doc_retrieval_gen.qa_infer_gradio,
+                inputs=[gr.Textbox(label="QUERY", placeholder="Enter your query here")],
+                allow_flagging='never',
+                examples=EXAMPLES,
+                cache_examples=False,
+                outputs=[gr.Textbox(label="RESPONSE"), gr.Textbox(label="RELATED QUERIES")],
+                css=css_code,
+                title="TI E2E FORUM Multi-Agent RAG"
+            )
+
+            interface.launch(debug=True)
+
+        launch_interface()
+    except Exception as e:
+        logger.error(f"An error occurred: {str(e)}")
+        logger.info("Please check your environment setup and try again.")