Upload 3 files

RAG_Solution_FIXED.py

@@ -0,0 +1,92 @@
+import argparse
+from pathlib import Path
+from typing import List
+
+from langchain_text_splitters import RecursiveCharacterTextSplitter
+from langchain_community.vectorstores import FAISS
+from langchain_community.embeddings import SentenceTransformerEmbeddings
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
+
+DEFAULT_EMBED_MODEL = "sentence-transformers/all-MiniLM-L6-v2"
+DEFAULT_GEN_MODEL = "google/flan-t5-base"
+
+def load_text_files(data_dir: Path, files: List[Path] = None) -> List[str]:
+    texts = []
+    if files:
+        paths = files
+    else:
+        paths = sorted(Path(data_dir).glob("*.txt"))
+    for p in paths:
+        if p.exists() and p.is_file():
+            texts.append(p.read_text(encoding="utf-8", errors="ignore"))
+    if not texts:
+        raise FileNotFoundError(f"No text files found in {Path(data_dir).resolve()} or provided via --files")
+    return texts
+
+def build_chunks(texts: List[str], chunk_size: int = 800, chunk_overlap: int = 120):
+    splitter = RecursiveCharacterTextSplitter(
+        chunk_size=chunk_size, chunk_overlap=chunk_overlap,
+        separators=["\n\n", "\n", " ", ""]
+    )
+    docs = splitter.create_documents(texts)
+    return docs
+
+def build_vectorstore(docs):
+    embeddings = SentenceTransformerEmbeddings(model_name=DEFAULT_EMBED_MODEL)
+    vs = FAISS.from_documents(docs, embeddings)
+    return vs
+
+def make_generator(model_name: str = DEFAULT_GEN_MODEL, device: int = -1):
+    tok = AutoTokenizer.from_pretrained(model_name)
+    mdl = AutoModelForSeq2SeqLM.from_pretrained(model_name)
+    gen = pipeline("text2text-generation", model=mdl, tokenizer=tok, device=device)
+    return gen
+
+def format_prompt(question: str, contexts):
+    context_block = "\n\n".join([d.page_content for d in contexts])
+    return (
+        "You are an expert assistant. Use ONLY the context to answer.\n"
+        "If the answer can't be found in the context, say you don't know.\n\n"
+        f"Context:\n{context_block}\n\n"
+        f"Question: {question}\n"
+        "Answer:"
+    )
+
+def answer_question(vs, generator, question: str, k: int = 4, max_new_tokens: int = 256):
+    contexts = vs.similarity_search(question, k=k)
+    prompt = format_prompt(question, contexts)
+    out = generator(prompt, max_new_tokens=max_new_tokens, do_sample=False)
+    text = out[0]["generated_text"]
+    sources = [d.metadata.get("source", "") for d in contexts]
+    return text, contexts, sources
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--question", required=True, help="Question to ask the RAG system.")
+    ap.add_argument("--data_dir", default="./data", help="Folder with .txt files.")
+    ap.add_argument("--files", nargs="*", help="Specific files to use (overrides data_dir)")
+    ap.add_argument("--k", type=int, default=4)
+    ap.add_argument("--chunk_size", type=int, default=800)
+    ap.add_argument("--chunk_overlap", type=int, default=120)
+    ap.add_argument("--max_new_tokens", type=int, default=256)
+    args = ap.parse_args()
+
+    data_dir = Path(args.data_dir)
+    file_paths = [Path(f) for f in args.files] if args.files else None
+
+    texts = load_text_files(data_dir, file_paths)
+    docs = build_chunks(texts, chunk_size=args.chunk_size, chunk_overlap=args.chunk_overlap)
+    vs = build_vectorstore(docs)
+    generator = make_generator()
+
+    answer, contexts, sources = answer_question(
+        vs, generator, args.question, k=args.k, max_new_tokens=args.max_new_tokens
+    )
+    print("\n=== Answer ===\n", answer.strip())
+    print("\n=== Top Sources (chunk previews) ===")
+    for i, d in enumerate(contexts, 1):
+        preview = d.page_content[:200].replace("\n", " ")
+        print(f"[{i}] {preview}...")
+
+if __name__ == "__main__":
+    main()

app.py

@@ -0,0 +1,55 @@
+from flask import Flask, request, jsonify
+from flask_cors import CORS
+from pathlib import Path
+from waitress import serve
+import RAG_Solution_FIXED as rag # Import your existing RAG code
+
+app = Flask(__name__)
+CORS(app)
+
+# --- Load your RAG model components ---
+print("--- RAG Model Initializing ---")
+
+# Define the data directory
+data_dir = Path("./data")
+
+# 1. Load text files
+print("[1/4] Loading text files...")
+texts = rag.load_text_files(data_dir)
+print("      ...Done.")
+
+# 2. Build chunks
+print("[2/4] Building text chunks...")
+docs = rag.build_chunks(texts)
+print("      ...Done.")
+
+# 3. Build vector store (This can be slow)
+print("[3/4] Building vector store with embeddings. This may take a moment...")
+vs = rag.build_vectorstore(docs)
+print("      ...Done.")
+
+# 4. Load the generative model (This is the slowest part)
+print("[4/4] Loading generative model (e.g., Flan-T5). This can be very slow...")
+generator = rag.make_generator()
+print("      ...Done.")
+
+print("--- Model loading complete. Server is now starting. ---")
+# ------------------------------------
+
+@app.route('/ask', methods=['POST'])
+def ask_question():
+    """ This is the API endpoint that will receive questions from the frontend. """
+    # ... (rest of your function is the same)
+    if not request.json or 'question' not in request.json:
+        return jsonify({'error': 'Missing question in request body'}), 400
+    question = request.json['question']
+    try:
+        answer, _, _ = rag.answer_question(vs, generator, question)
+        return jsonify({'answer': answer.strip()})
+    except Exception as e:
+        return jsonify({'error': str(e)}), 500
+
+if __name__ == '__main__':
+    # This line will only run AFTER all the models above are loaded
+    serve(app, host="127.0.0.1", port=8765)
+    print("Server has started on http://127.0.0.1:3000 and is ready for requests.")

requirements.txt

@@ -0,0 +1,17 @@
+# Core AI and Language Model Libraries
+langchain
+langchain-community
+transformers
+sentence-transformers
+torch
+
+# Vector Store Library (CPU version is best for cloud deployment)
+faiss-cpu
+
+# Web Server Libraries
+flask
+flask-cors
+waitress
+
+# Additional Utilities
+langchain-text-splitters