Update app.py

upload cache and work without file

app.py

@@ -5,12 +5,16 @@ from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain.embeddings import HuggingFaceEmbeddings
 from langchain_community.embeddings import HuggingFaceEmbeddings
 import fitz  # PyMuPDF
+import os
+import hashlib
+
+# Directory to store cached files
+CACHE_DIR = "pdf_cache"
+os.makedirs(CACHE_DIR, exist_ok=True)
 
-# Function to get available models from Hugging Face
 def get_hf_models():
     return ["Qwen/Qwen2.5-3B-Instruct", "HuggingFaceH4/zephyr-7b-beta", "mistralai/Mistral-7B-Instruct-v0.1"]
 
-# Function to extract text from a PDF
 def extract_text_from_pdf(pdf_path):
     text = ""
     with fitz.open(pdf_path) as doc:
@@ -18,13 +22,11 @@ def extract_text_from_pdf(pdf_path):
             text += page.get_text()
     return text
 
-# Function for manual RAG
 def manual_rag(query, context, client):
     prompt = f"Context: {context}\n\nQuestion: {query}\n\nAnswer:"
     response = client.text_generation(prompt, max_new_tokens=512)
     return response
 
-# Function for classic RAG
 def classic_rag(query, pdf_path, client, embedder):
     text = extract_text_from_pdf(pdf_path)
     text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
@@ -36,26 +38,47 @@ def classic_rag(query, pdf_path, client, embedder):
     response = manual_rag(query, context, client)
     return response, context
 
-# Function for response without RAG
 def no_rag(query, client):
     response = client.text_generation(query, max_new_tokens=512)
     return response
 
-# Gradio interface function
-def process_query(query, pdf_path, llm_choice, embedder_choice):
+def cache_file(file):
+    if file is None:
+        return None
+    file_hash = hashlib.md5(file.read()).hexdigest()
+    cached_path = os.path.join(CACHE_DIR, f"{file_hash}.pdf")
+    if not os.path.exists(cached_path):
+        with open(cached_path, "wb") as f:
+            file.seek(0)
+            f.write(file.read())
+    return cached_path
+
+def get_cached_files():
+    return [f for f in os.listdir(CACHE_DIR) if f.endswith('.pdf')]
+
+def process_query(query, pdf_file, cached_file, llm_choice, embedder_choice):
     client = InferenceClient(llm_choice)
-    full_text = extract_text_from_pdf(pdf_path)
     no_rag_response = no_rag(query, client)
+
+    if pdf_file is not None:
+        pdf_path = cache_file(pdf_file)
+    elif cached_file:
+        pdf_path = os.path.join(CACHE_DIR, cached_file)
+    else:
+        return no_rag_response, "RAG non utilisé (pas de fichier PDF)", "RAG non utilisé (pas de fichier PDF)", "Pas de fichier PDF fourni", "Pas de contexte extrait"
+
+    full_text = extract_text_from_pdf(pdf_path)
     manual_rag_response = manual_rag(query, full_text, client)
     classic_rag_response, classic_rag_context = classic_rag(query, pdf_path, client, embedder_choice)
+
     return no_rag_response, manual_rag_response, classic_rag_response, full_text, classic_rag_context
 
-# Create Gradio interface
 iface = gr.Interface(
     fn=process_query,
     inputs=[
         gr.Textbox(label="Votre question"),
-        gr.File(label="Chargez votre PDF"),
+        gr.File(label="Chargez un nouveau PDF"),
+        gr.Dropdown(choices=get_cached_files, label="Ou choisissez un PDF déjà téléversé", interactive=True),
         gr.Dropdown(choices=get_hf_models(), label="Choisissez le LLM", value="Qwen/Qwen2.5-3B-Instruct"),
         gr.Dropdown(choices=["sentence-transformers/all-MiniLM-L6-v2", "nomic-ai/nomic-embed-text-v1.5"],
                     label="Choisissez l'Embedder", value="sentence-transformers/all-MiniLM-L6-v2")
@@ -72,6 +95,5 @@ iface = gr.Interface(
     theme="default"
 )
 
-# Launch the application
 if __name__ == "__main__":
     iface.launch()