# 1) Instal
# pip install llama-index llama-index-embeddings-openai llama-index-llms-openai
import os
from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.llms.openai import OpenAI
# 2) Konfigurasi model + embeddings Anda
os.environ["OPENAI_API_KEY"] = "YOUR_KEY" # atau gunakan penyedia LLM/embedding yang didukung
llm = OpenAI(model="gpt-4o-mini")
embed_model = OpenAIEmbedding(model="text-embedding-3-small")
# 3) Muat dokumen (mis., ./data/*.pdf, .md, .txt)
docs = SimpleDirectoryReader("./data").load_data
# 4) Bangun indeks
index = VectorStoreIndex.from_documents(docs, embed_model=embed_model)
# 5) Buat query engine dan ajukan pertanyaan
query_engine = index.as_query_engine(llm=llm)
response = query_engine.query("What are the key security practices mentioned in the docs?")
print(response)

from llama_index.core import SimpleDirectoryReader
from llama_index.readers.web import SimpleWebPageReader
file_docs = SimpleDirectoryReader("./policies").load_data
web_docs = SimpleWebPageReader(html_to_text=True).load_data
all_docs = file_docs + web_docs

from llama_index.core.node_parser import SentenceSplitter
from llama_index.core import Document
parser = SentenceSplitter(chunk_size=800, chunk_overlap=100)
nodes = []
for d in all_docs:
nodes.extend(parser.get_nodes_from_documents([Document(text=d.text, metadata=d.metadata)]))

from llama_index.core import VectorStoreIndex, SummaryIndex
from llama_index.core.retrievers import BM25Retriever
from llama_index.core.query_engine import RetrieverQueryEngine
# Indeks vektor dari nodes yang telah di-parse sebelumnya
v_index = VectorStoreIndex(nodes)
# BM25 keyword retriever
bm25_retriever = BM25Retriever.from_defaults(nodes=nodes, similarity_top_k=6)
# Hybrid: gabungkan kandidat, lalu rerank
from llama_index.core.retrievers import RouterRetriever
from llama_index.retrievers.merge import MergerRetriever
v_retriever = v_index.as_retriever(similarity_top_k=6)
hybrid = MergerRetriever(retrievers=[v_retriever, bm25_retriever], top_k=8)
query_engine = RetrieverQueryEngine.from_args(retriever=hybrid)

from llama_index.postprocessor.flag_embedding_reranker import FlagEmbeddingReranker
from llama_index.core.query_engine import RetrieverQueryEngine
reranker = FlagEmbeddingReranker(top_n=5, model="BAAI/bge-reranker-base")
query_engine = v_index.as_query_engine(
similarity_top_k=12,
node_postprocessors=[reranker]
)

from llama_index.core.response_synthesizers import get_response_synthesizer
from llama_index.core import ServiceContext
synth = get_response_synthesizer(response_mode="tree_summarize")
query_engine = v_index.as_query_engine(response_synthesizer=synth)
ans = query_engine.query("Summarize the onboarding steps and cite sources.")
print(ans)

from llama_index.core.prompts import PromptTemplate
qa_tmpl = PromptTemplate(
"""
Anda adalah asisten ringkas yang mengutamakan bukti. Gunakan hanya konteks yang disediakan.
Jika tidak yakin, katakan Anda tidak tahu. Kembalikan JSON dengan kunci: answer, sources.
Pertanyaan: {query_str}
Konteks: {context_str}
"""
)
query_engine = v_index.as_query_engine(text_qa_template=qa_tmpl)

from llama_index.core.agent import ReActAgent
from llama_index.tools.sql import SQLQueryEngineTool
from sqlalchemy import create_engine
engine = create_engine("sqlite:///analytics.db")
sql_tool = SQLQueryEngineTool.from_engine(engine)
agent = ReActAgent.from_tools([sql_tool], llm=llm, verbose=True)
agent.chat("What was monthly churn in Q2 2025? If needed, query the DB.")

from llama_index.core.evaluation import FaithfulnessEvaluator, RelevancyEvaluator
faith = FaithfulnessEvaluator(llm=llm)
rel = RelevancyEvaluator(llm=llm)
pred = query_engine.query("List SOC 2 control families in our policy.")
print("faithful?", faith.evaluate_response(pred))
print("relevant?", rel.evaluate_response(pred))

# Contoh: Chroma
# pip install chromadb llama-index-vector-stores-chroma
from llama_index.vector_stores.chroma import ChromaVectorStore
from llama_index.core import StorageContext
import chromadb
chroma_client = chromadb.PersistentClient(path="./chroma_store")
collection = chroma_client.get_or_create_collection("company_knowledge")
vector_store = ChromaVectorStore(chroma_collection=collection)
storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(all_docs, storage_context=storage_context)

# Contoh: filtering berbasis metadata pada waktu kueri
retriever = index.as_retriever(similarity_top_k=8)
retriever.metadata_filters = {"department": ["legal", "security"], "published": [True]}

# Pembungkus FastAPI minimal
# pip install fastapi uvicorn
from fastapi import FastAPI
app = FastAPI
qe = index.as_query_engine(llm=llm)
@app.post("/query")
async def query(payload: dict):
q = payload.get("q", "")
resp = qe.query(q)
return {"answer": str(resp), "sources": [s.node.metadata for s in resp.source_nodes]}