Upload 2 files

app.py

@@ -1,63 +1,98 @@
-import gradio as gr
-from huggingface_hub import InferenceClient
-
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
-)
-
-
-if __name__ == "__main__":
-    demo.launch()
+import os
+import pinecone
+import openai
+import gradio as gr
+import torch
+from dotenv import load_dotenv
+from pinecone import Pinecone
+from langchain_huggingface import HuggingFaceEmbeddings
+
+# Detect GPU availability and set device
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Running on device: {device}")
+
+# Suppress specific warning about clean_up_tokenization_spaces
+import warnings
+warnings.filterwarnings("ignore", category=FutureWarning, message="clean_up_tokenization_spaces was not set")
+
+# Load environment variables
+load_dotenv()
+
+# Access Pinecone and OpenAI API keys from environment variables
+pinecone_api_key = os.getenv("PINECONE_API_KEY")
+openai.api_key = os.getenv("OPENAI_API_KEY")
+index_name = "amtrak-rmm-image-text"
+
+# Initialize Pinecone using a class-based method
+pc = Pinecone(api_key=pinecone_api_key)
+
+# Check if the index exists, if not, create it
+def initialize_pinecone_index(index_name):
+    available_indexes = pc.list_indexes().names()
+    if index_name not in available_indexes:
+        print(f"Index '{index_name}' does not exist.")
+        # Create the index here if necessary for ZeroGPU usage
+    return pc.Index(index_name)
+
+index = initialize_pinecone_index(index_name)
+
+# Initialize HuggingFace embedding model
+embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/msmarco-distilbert-base-v4")
+
+# Function to interact with Pinecone and OpenAI GPT-4
+def get_model_response(human_input, chat_history=None):
+    try:
+        # Fetching and embedding query (move tensors to GPU)
+        query_embedding = torch.tensor(embedding_model.embed_query(human_input)).to(device)
+
+        # Convert NumPy array to list before passing it to Pinecone or any API that requires JSON-serializable data
+        query_embedding = query_embedding.cpu().numpy().tolist()
+
+        # Query Pinecone index using CPU or GPU for fast vector search (if supported)
+        search_results = index.query(vector=query_embedding, top_k=2, include_metadata=True)
+
+        context_list, images = [], []
+        for ind, result in enumerate(search_results['matches']):
+            document_content = result.get('metadata', {}).get('content', 'No content found')
+            image_url = result.get('metadata', {}).get('image_path', None)
+            figure_desc = result.get('metadata', {}).get('figure_description', '')
+            
+            context_list.append(f"Document {ind+1}: {document_content}")
+            
+            if image_url and figure_desc:
+                images.append((figure_desc, image_url))
+        
+        context_string = '\n\n'.join(context_list)
+        messages = [{"role": "system", "content": "You are a helpful assistant."},
+                    {"role": "user", "content": f"Here is some context:\n{context_string}\n\nUser's question: {human_input}"}]
+        
+        # Generate response using OpenAI GPT-3.5 Turbo for faster responses
+        response = openai.ChatCompletion.create(
+            model="gpt-3.5-turbo", 
+            messages=messages, 
+            max_tokens=500, 
+            temperature=0.5
+        )
+        output_text = response['choices'][0]['message']['content'].strip()
+        return output_text, images
+    except Exception as e:
+        return f"Error invoking model: {str(e)}", []
+
+# Function to format text and images for display
+def get_model_response_with_images(human_input, chat_history=None):
+    output_text, images = get_model_response(human_input, chat_history)
+    if images:
+        image_output = "".join([f"\n\n**{figure_desc}**\n![{figure_desc}]({image_path})" for figure_desc, image_path in images])
+        return output_text + image_output
+    return output_text
+
+# Set up Gradio interface
+gr_interface = gr.ChatInterface(
+    fn=get_model_response_with_images, 
+    title="Maintenance Assistant", 
+    description="Ask questions related to the RMMM documents."
+)
+
+# Ensure ZeroGPU or Hugging Face Spaces handles launching properly
+if __name__ == "__main__":
+    gr_interface.launch()

requirements.txt

@@ -1 +1,41 @@
-huggingface_hub==0.22.2
+# Core dependencies
+openai==0.28
+pinecone-client==5.0.1
+langchain==0.3.1  # Ensure compatibility with your current LangChain version
+langchain-huggingface==0.1.0  # Updated for HuggingFaceEmbeddings
+langchain-community==0.3.1  # Updated LangChain community version
+boto3==1.35.27
+gradio==4.36.1
+python-dotenv==1.0.1
+sentence-transformers==3.1.1
+transformers==4.45.0
+poppler-utils==0.1.0
+uvicorn==0.30.6
+fastapi==0.112.4
+
+# Additional dependencies for specific functionality
+altair==5.4.1
+authlib==1.3.2
+cryptography==43.0.1
+psutil==5.9.8
+spaces==0.30.2
+itsdangerous==2.2.0
+pydub==0.25.1
+matplotlib==3.9.2
+numpy==1.26.4
+scikit-learn==1.5.2
+scipy==1.14.1
+sqlalchemy==2.0.35
+tenacity==8.5.0
+requests==2.32.3
+torch==2.4.0
+pandas
+
+# Dependency for Pinecone plugin inference
+pinecone-plugin-inference==1.1.0
+
+# Update langchain-text-splitters for compatibility with langchain==0.3.1
+langchain-text-splitters==0.3.0
+
+# Updated for langchain-huggingface embedding handling
+langchain-huggingface==0.1.0