app/langchain_rag.py

# ─────────────────────────────────────────────────────────────
# app/langchain_rag.py
# LangChain version of the RAG pipeline
# ─────────────────────────────────────────────────────────────

import os
import sys
import numpy as np
from dotenv import load_dotenv
load_dotenv()

sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))

from gradio_client import Client
from langchain.embeddings.base      import Embeddings
from langchain_community.vectorstores import FAISS
from langchain_community.llms        import HuggingFaceHub
from langchain.chains                import RetrievalQA
from langchain.schema                import Document


# ── Wrap your HF Gradio Space as LangChain Embeddings ────────
class GradioEmbeddings(Embeddings):
    """
    LangChain-compatible wrapper around your
    HF Gradio Space embedding API.
    """

    def __init__(self, space: str = None):
        self.space  = space or os.getenv("GRADIO_SPACE", "your-username/rag-embedder-app")
        self.client = Client(self.space)
        print(f"Connected to Gradio Space: {self.space}")

    def embed_documents(self, texts: list) -> list:
        return [self.client.predict(t, api_name="/predict") for t in texts]

    def embed_query(self, text: str) -> list:
        return self.client.predict(text, api_name="/predict")


# ── Load documents ────────────────────────────────────────────
def load_documents(path: str) -> list:
    with open(path) as f:
        lines = [line.strip() for line in f if line.strip()]
    return [Document(page_content=line) for line in lines]


# ── Build LangChain RAG chain ─────────────────────────────────
def build_rag_chain():
    docs_path = os.getenv("DOCS_PATH", "data/sample_docs.txt")
    hf_token  = os.getenv("HF_TOKEN",  "")
    llm_model = os.getenv("LLM_MODEL", "mistralai/Mistral-7B-Instruct-v0.1")

    print("Setting up LangChain RAG pipeline...")

    # Load docs
    documents  = load_documents(docs_path)
    print(f"Loaded {len(documents)} documents")

    # Embeddings via your HF Gradio Space
    embeddings = GradioEmbeddings()

    # Vector store
    vectorstore = FAISS.from_documents(documents, embeddings)
    retriever   = vectorstore.as_retriever(search_kwargs={"k": 3})

    # LLM via HF Hub
    llm = HuggingFaceHub(
        repo_id               = llm_model,
        huggingfacehub_api_token = hf_token,
        model_kwargs          = {"max_new_tokens": 200, "temperature": 0.3}
    )

    # Full RAG chain
    chain = RetrievalQA.from_chain_type(
        llm       = llm,
        retriever = retriever,
        chain_type= "stuff",
        return_source_documents = True
    )

    print("LangChain RAG chain ready!")
    return chain


# ── Run ───────────────────────────────────────────────────────
if __name__ == "__main__":
    chain = build_rag_chain()

    questions = [
        "What is the refund policy?",
        "How do I reset my password?",
        "When can I contact support?"
    ]

    print("\n" + "=" * 55)
    for q in questions:
        result  = chain({"query": q})
        answer  = result["result"]
        sources = [doc.page_content for doc in result["source_documents"]]
        print(f"Q: {q}")
        print(f"A: {answer}")
        print(f"Sources: {sources[:2]}")
        print("-" * 55)