Update app.py

app.py

@@ -1 +1,113 @@
-# Main application logic will go here
+# Correcting the CSV_PATH to reflect the actual location after copying
+CSV_PATH = "memes.csv"
+
+# Re-running the app logic with the corrected path
+import os
+import sys
+import pandas as pd
+import gradio as gr
+from openai import OpenAI
+from langchain.docstore.document import Document
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.vectorstores import FAISS
+
+# --- CONFIG ---
+# CSV_PATH = "/data/Memes and descriptions - Sheet1.csv" # Corrected above
+MODEL_NAME = "qwen/qwen3-32b:free"
+
+# Use environment variable for API key in Hugging Face Spaces
+API_KEY = os.getenv("OPENROUTER_API_KEY")
+#from google.colab import userdata
+#API_KEY = userdata.get('open_router')
+if not API_KEY:
+    # Fallback for local testing if needed, but prefer env var
+    # sys.exit("❌ Missing OpenRouter/OpenAI API key.")
+    print("⚠️ OPENROUTER_API_KEY not set. Using dummy key.")
+    API_KEY = "dummy_key"
+
+
+try:
+    client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=API_KEY)
+except Exception as e:
+    print(f"❌ Failed to initialize OpenRouter client: {e}")
+    client = None # Handle case where client initialization fails
+
+# --- LOAD DATA ---
+try:
+    df = pd.read_csv(CSV_PATH).fillna({"Description": "", "Link": ""})
+    documents = [
+        Document(
+            page_content=row["Description"],
+            metadata={"url": str(row["link"]).strip()}
+        )
+        for _, row in df.iterrows()
+    ]
+
+    # --- FAISS ---
+    embedding_model = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-large-instruct")
+    vectorstore = FAISS.from_documents(documents, embedding_model)
+    retriever = vectorstore.as_retriever(search_kwargs={"k": 10})
+
+except FileNotFoundError:
+    print(f"❌ Data file not found at {CSV_PATH}")
+    documents = []
+    vectorstore = None
+    retriever = None
+except Exception as e:
+    print(f"❌ Error loading data or creating vectorstore: {e}")
+    documents = []
+    vectorstore = None
+    retriever = None
+
+
+# --- LLM ---
+def ask_llm(question: str, docs: list) -> str:
+    if client is None:
+        return "❌ LLM client not initialized."
+    context = "\n\n".join(
+        f"Meme {i+1}:\nDescription: {doc.page_content}\nLink: {doc.metadata.get('url', 'N/A')}"
+        for i, doc in enumerate(docs)
+    )
+    messages = [
+        {"role": "system", "content": f"You're a meme expert. the user will say something and the goal is to find the accurate meme out of the following choices : \n{context}"},
+        {"role": "user", "content": f"{question}"}
+    ]
+    try:
+        response = client.chat.completions.create(
+            model=MODEL_NAME,
+            messages=messages,
+            extra_headers={"HTTP-Referer": "https://your-site.com", "X-Title": "MemeRAG"}
+        )
+        return response.choices[0].message.content
+    except Exception as e:
+        return f"❌ LLM Error: {e}"
+
+# --- MAIN QUERY ---
+def query_memes(user_input: str):
+    if retriever is None:
+        return "❌ RAG system not initialized due to errors."
+
+    src_docs = retriever.invoke(user_input)
+    answer = ask_llm(user_input, src_docs)
+
+    output_text = f"💡 Answer:\n{answer}\n\n📂 Top Matching Memes:"
+    for i, doc in enumerate(src_docs, 1):
+        raw = doc.metadata.get("url", "").strip()
+        url = raw if raw.startswith("http") else f"https://drive.google.com/search?q={raw.replace(' ', '%20')}"
+        output_text += f"\n\n{i}. {doc.page_content}\n   Link: {url}"
+
+    return output_text
+
+# --- GRADIO INTERFACE ---
+if __name__ == "__main__":
+    if retriever is None or client is None:
+        print("Gradio interface will not run due to RAG/LLM initialization errors.")
+    else:
+        interface = gr.Interface(
+            fn=query_memes,
+            inputs=gr.Textbox(label="Ask something about memes"),
+            outputs=gr.Textbox(label="Results"),
+            title="Memes lharba 🎬",
+            description="Ask me to find the perfect meme!"
+        )
+        interface.launch()