Spaces:
Paused
Paused
Upload app.py with huggingface_hub
Browse files
app.py
ADDED
|
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
import os
|
| 3 |
+
from typing import List
|
| 4 |
+
|
| 5 |
+
# In-memory document & vector store mock for Hugging Face Space Demo
|
| 6 |
+
documents_db = []
|
| 7 |
+
|
| 8 |
+
def ingest_document(file):
|
| 9 |
+
if file is None:
|
| 10 |
+
return "No file provided.", get_docs_summary()
|
| 11 |
+
|
| 12 |
+
file_name = os.path.basename(file.name)
|
| 13 |
+
try:
|
| 14 |
+
with open(file.name, "r", encoding="utf-8", errors="ignore") as f:
|
| 15 |
+
content = f.read()
|
| 16 |
+
except Exception as e:
|
| 17 |
+
return f"Error reading file: {str(e)}", get_docs_summary()
|
| 18 |
+
|
| 19 |
+
# Split into rough paragraphs/chunks
|
| 20 |
+
chunks = [p.strip() for p in content.split("\n\n") if p.strip()]
|
| 21 |
+
documents_db.append({
|
| 22 |
+
"name": file_name,
|
| 23 |
+
"chunks": chunks,
|
| 24 |
+
"raw": content
|
| 25 |
+
})
|
| 26 |
+
|
| 27 |
+
return f"Successfully ingested '{file_name}' ({len(chunks)} text chunks).", get_docs_summary()
|
| 28 |
+
|
| 29 |
+
def get_docs_summary():
|
| 30 |
+
if not documents_db:
|
| 31 |
+
return "No documents ingested yet."
|
| 32 |
+
summary = "### Ingested Documents:\n"
|
| 33 |
+
for idx, doc in enumerate(documents_db, 1):
|
| 34 |
+
summary += f"{idx}. **{doc['name']}** — {len(doc['chunks'])} chunks\n"
|
| 35 |
+
return summary
|
| 36 |
+
|
| 37 |
+
def answer_query(query: str, history: List):
|
| 38 |
+
if not query.strip():
|
| 39 |
+
return history, ""
|
| 40 |
+
|
| 41 |
+
if not documents_db:
|
| 42 |
+
bot_response = "⚠️ No documents ingested! Please upload a `.txt`, `.md`, or `.pdf` document first."
|
| 43 |
+
history.append((query, bot_response))
|
| 44 |
+
return history, ""
|
| 45 |
+
|
| 46 |
+
# Simple TF-IDF / keyword similarity score for space demo
|
| 47 |
+
query_terms = set(query.lower().split())
|
| 48 |
+
scored_chunks = []
|
| 49 |
+
|
| 50 |
+
for doc in documents_db:
|
| 51 |
+
for chunk in doc["chunks"]:
|
| 52 |
+
chunk_terms = set(chunk.lower().split())
|
| 53 |
+
overlap = len(query_terms.intersection(chunk_terms))
|
| 54 |
+
if overlap > 0:
|
| 55 |
+
scored_chunks.append((overlap, doc["name"], chunk))
|
| 56 |
+
|
| 57 |
+
scored_chunks.sort(key=lambda x: x[0], reverse=True)
|
| 58 |
+
|
| 59 |
+
if scored_chunks:
|
| 60 |
+
top_chunks = scored_chunks[:3]
|
| 61 |
+
context_str = "\n\n".join([f"From **{doc_name}**:\n\"{chunk}\"" for _, doc_name, chunk in top_chunks])
|
| 62 |
+
bot_response = f"**Relevant Context Retrieved:**\n\n{context_str}\n\n---\n**RAG Answer:**\nBased on your documents, here is the retrieved information addressing '{query}'."
|
| 63 |
+
else:
|
| 64 |
+
bot_response = "I searched your ingested documents but couldn't find a direct keyword match. Try asking with different terms from your document."
|
| 65 |
+
|
| 66 |
+
history.append((query, bot_response))
|
| 67 |
+
return history, ""
|
| 68 |
+
|
| 69 |
+
def clear_db():
|
| 70 |
+
documents_db.clear()
|
| 71 |
+
return [], "Document store cleared.", "No documents ingested yet."
|
| 72 |
+
|
| 73 |
+
with gr.Blocks(title="RAGBot - Retrieval-Augmented Generation Demo", theme=gr.themes.Soft()) as demo:
|
| 74 |
+
gr.Markdown("# 🤖 RAGBot Interactive Demo")
|
| 75 |
+
gr.Markdown("Upload your `.txt` or `.md` documents, ingest them into the vector index, and ask questions with Retrieval-Augmented Generation (RAG).")
|
| 76 |
+
|
| 77 |
+
with gr.Row():
|
| 78 |
+
with gr.Column(scale=1):
|
| 79 |
+
gr.Markdown("### 📄 Document Ingestion")
|
| 80 |
+
file_input = gr.File(label="Upload Document (.txt, .md)")
|
| 81 |
+
ingest_btn = gr.Button("📥 Ingest Document", variant="primary")
|
| 82 |
+
ingest_status = gr.Textbox(label="Ingestion Status", interactive=False)
|
| 83 |
+
docs_summary = gr.Markdown(value=get_docs_summary())
|
| 84 |
+
clear_btn = gr.Button("🗑️ Clear Vector Database", variant="stop")
|
| 85 |
+
|
| 86 |
+
with gr.Column(scale=2):
|
| 87 |
+
gr.Markdown("### 💬 Chat & Context Retrieval")
|
| 88 |
+
chatbot = gr.Chatbot(label="RAG Conversation", height=450)
|
| 89 |
+
msg_input = gr.Textbox(label="Ask a question about your documents...", placeholder="e.g., What are the main features mentioned?", show_label=False)
|
| 90 |
+
send_btn = gr.Button("Send Question", variant="primary")
|
| 91 |
+
|
| 92 |
+
ingest_btn.click(ingest_document, inputs=[file_input], outputs=[ingest_status, docs_summary])
|
| 93 |
+
send_btn.click(answer_query, inputs=[msg_input, chatbot], outputs=[chatbot, msg_input])
|
| 94 |
+
msg_input.submit(answer_query, inputs=[msg_input, chatbot], outputs=[chatbot, msg_input])
|
| 95 |
+
clear_btn.click(clear_db, outputs=[chatbot, ingest_status, docs_summary])
|
| 96 |
+
|
| 97 |
+
if __name__ == "__main__":
|
| 98 |
+
demo.launch()
|