diff --git a/.env.example b/.env.example
index 683573e..8918e6a 100644
--- a/.env.example
+++ b/.env.example
@@ -1,8 +1,8 @@
 TELEGRAM_BOT_TOKEN=isi_token_dari_botfather
-OPENAI_API_KEY=isi_api_key_openai
-OPENAI_MODEL=gpt-4o
-OPENAI_EMBEDDING_MODEL=text-embedding-3-small
+DEEPSEEK_API_KEY=isi_api_key_deepseek
+DEEPSEEK_MODEL=deepseek-chat
 ADMIN_TELEGRAM_ID=isi_telegram_id_anda
 DB_PATH=/app/data/second_brain.db
 CHROMA_PATH=/app/data/chromadb
 UPLOAD_PATH=/app/data/uploads
+EMBEDDING_MODEL=all-MiniLM-L6-v2
diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..dc80642
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1,7 @@
+.env
+data/*
+!data/.gitkeep
+__pycache__/
+*.pyc
+*.pyo
+.DS_Store
diff --git a/Dockerfile b/Dockerfile
index f8afa71..536746c 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -10,6 +10,9 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 COPY requirements.txt .
 RUN pip install --no-cache-dir -r requirements.txt
 
+# Pre-download embedding model agar tidak diunduh ulang saat runtime
+RUN python -c "from sentence_transformers import SentenceTransformer; SentenceTransformer('all-MiniLM-L6-v2')"
+
 COPY src/ ./src/
 
 CMD ["python", "src/main.py"]
diff --git a/requirements.txt b/requirements.txt
index 31a6872..60a533a 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -4,3 +4,4 @@ python-dotenv==1.0.1
 chromadb==0.5.3
 pdfplumber==0.11.0
 python-docx==1.1.2
+sentence-transformers==3.0.1
diff --git a/src/config.py b/src/config.py
index 510d2a8..685641a 100644
--- a/src/config.py
+++ b/src/config.py
@@ -4,10 +4,11 @@ from dotenv import load_dotenv
 load_dotenv()
 
 TELEGRAM_BOT_TOKEN = os.environ["TELEGRAM_BOT_TOKEN"]
-OPENAI_API_KEY = os.environ["OPENAI_API_KEY"]
-OPENAI_MODEL = os.getenv("OPENAI_MODEL", "gpt-4o")
-OPENAI_EMBEDDING_MODEL = os.getenv("OPENAI_EMBEDDING_MODEL", "text-embedding-3-small")
+DEEPSEEK_API_KEY = os.environ["DEEPSEEK_API_KEY"]
+DEEPSEEK_MODEL = os.getenv("DEEPSEEK_MODEL", "deepseek-chat")
+DEEPSEEK_BASE_URL = "https://api.deepseek.com"
 ADMIN_TELEGRAM_ID = int(os.environ["ADMIN_TELEGRAM_ID"])
 DB_PATH = os.getenv("DB_PATH", "/app/data/second_brain.db")
 CHROMA_PATH = os.getenv("CHROMA_PATH", "/app/data/chromadb")
 UPLOAD_PATH = os.getenv("UPLOAD_PATH", "/app/data/uploads")
+EMBEDDING_MODEL = os.getenv("EMBEDDING_MODEL", "all-MiniLM-L6-v2")
diff --git a/src/handlers/user.py b/src/handlers/user.py
index 378fc44..1346775 100644
--- a/src/handlers/user.py
+++ b/src/handlers/user.py
@@ -2,13 +2,13 @@ from pathlib import Path
 from telegram import Update
 from telegram.ext import ContextTypes
 from openai import AsyncOpenAI
-from config import OPENAI_API_KEY, OPENAI_MODEL, ADMIN_TELEGRAM_ID
+from config import DEEPSEEK_API_KEY, DEEPSEEK_BASE_URL, DEEPSEEK_MODEL, ADMIN_TELEGRAM_ID
 import database as db
 import rag
 import document as doc_processor
 from guard import is_academic_topic, REJECT_MESSAGE
 
-_client = AsyncOpenAI(api_key=OPENAI_API_KEY)
+_client = AsyncOpenAI(api_key=DEEPSEEK_API_KEY, base_url=DEEPSEEK_BASE_URL)
 
 ACADEMIC_SYSTEM_PROMPT = (
     "Anda adalah asisten akademik personal untuk dosen di perguruan tinggi Indonesia. "
@@ -128,7 +128,7 @@ async def ask(update: Update, context: ContextTypes.DEFAULT_TYPE):
     await update.message.chat.send_action("typing")
 
     # Guard: hanya izinkan topik akademik
-    if not await is_academic_topic(question, _client, OPENAI_MODEL):
+    if not await is_academic_topic(question, _client, DEEPSEEK_MODEL):
         await update.message.reply_text(REJECT_MESSAGE)
         return
 
@@ -154,7 +154,7 @@ async def ask(update: Update, context: ContextTypes.DEFAULT_TYPE):
 
     try:
         response = await _client.chat.completions.create(
-            model=OPENAI_MODEL,
+            model=DEEPSEEK_MODEL,
             messages=messages,
             max_tokens=1500,
         )
diff --git a/src/rag.py b/src/rag.py
index e929954..91c185b 100644
--- a/src/rag.py
+++ b/src/rag.py
@@ -1,10 +1,10 @@
 import asyncio
 import chromadb
-from openai import AsyncOpenAI
-from config import CHROMA_PATH, OPENAI_API_KEY, OPENAI_EMBEDDING_MODEL
+from sentence_transformers import SentenceTransformer
+from config import CHROMA_PATH, EMBEDDING_MODEL
 
-_client = AsyncOpenAI(api_key=OPENAI_API_KEY)
 _chroma = chromadb.PersistentClient(path=CHROMA_PATH)
+_embedding_model = SentenceTransformer(EMBEDDING_MODEL)
 
 
 def _get_collection(name: str):
@@ -12,8 +12,8 @@ def _get_collection(name: str):
 
 
 async def _embed(text: str) -> list:
-    resp = await _client.embeddings.create(model=OPENAI_EMBEDDING_MODEL, input=text)
-    return resp.data[0].embedding
+    embedding = await asyncio.to_thread(_embedding_model.encode, text)
+    return embedding.tolist()
 
 
 async def add_memory(user_id: int, memory_id: int, content: str, scope: str = "private"):