LangChain & AI Agent untuk Developer

from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from langchain_community.vectorstores import Chroma
from langchain_community.document_loaders import PyPDFLoader, TextLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnablePassthrough
from langchain_core.output_parsers import StrOutputParser

# ===== STEP 1: Load Documents =====
# Dari file PDF
loader = PyPDFLoader("dokumen_perusahaan.pdf")
docs = loader.load()

# Dari file teks
loader = TextLoader("faq.txt", encoding="utf-8")
docs = loader.load()

print(f"Jumlah halaman/dokumen: {len(docs)}")

# ===== STEP 2: Split Documents =====
text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=1000,      # Ukuran setiap chunk (karakter)
    chunk_overlap=200,    # Overlap antar chunk
    separators=["\n\n", "\n", ". ", " ", ""]
)
chunks = text_splitter.split_documents(docs)
print(f"Jumlah chunks: {len(chunks)}")

# ===== STEP 3: Create Embeddings & Vector Store =====
embeddings = OpenAIEmbeddings(model="text-embedding-3-small")

# Simpan ke ChromaDB (in-memory)
vectorstore = Chroma.from_documents(
    documents=chunks,
    embedding=embeddings,
    persist_directory="./chroma_db"  # Persist ke disk
)

# Buat retriever
retriever = vectorstore.as_retriever(
    search_type="similarity",
    search_kwargs={"k": 3}  # Ambil 3 dokumen paling relevan
)

# ===== STEP 4: Buat RAG Chain =====
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)

rag_prompt = ChatPromptTemplate.from_template("""
Gunakan konteks berikut untuk menjawab pertanyaan. 
Jika jawaban tidak ada dalam konteks, katakan bahwa Anda tidak tahu.

Konteks:
{context}

Pertanyaan: {question}

Jawaban:""")

def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)

rag_chain = (
    {"context": retriever | format_docs, "question": RunnablePassthrough()}
    | rag_prompt
    | llm
    | StrOutputParser()
)

# ===== STEP 5: Jalankan RAG =====
response = rag_chain.invoke("Apa kebijakan cuti karyawan?")
print(response)

response = rag_chain.invoke("Bagaimana cara klaim asuransi?")
print(response)