wehub-resource-sync
38 lines
990 B
Python
38 lines
990 B
Python
from beam import Image, Pod
|
|
|
|
def load_model():
|
|
import subprocess
|
|
import time
|
|
|
|
subprocess.Popen(["ollama", "serve"])
|
|
time.sleep(5)
|
|
subprocess.run(["ollama", "pull", "gemma3:4b"], check=True)
|
|
|
|
|
|
image = Image().add_python_packages([
|
|
"streamlit",
|
|
"nest-asyncio",
|
|
"load_dotenv",
|
|
"llama-index==0.12.33",
|
|
"llama-index-embeddings-fastembed==0.3.1",
|
|
"llama-index-vector-stores-milvus==0.8.4",
|
|
"llama-index-llms-ollama==0.6.2",
|
|
"llama-index-llms-openai==0.3.12",
|
|
"llama-index-readers-web==0.3.9",
|
|
"IPython",
|
|
"ollama",
|
|
"llama-index-core==0.12.33.post1",
|
|
])
|
|
|
|
streamlit_server = Pod(
|
|
image=image,
|
|
ports=[8501], # Default port for streamlit
|
|
gpu="T4",
|
|
memory="2Gi",
|
|
entrypoint=["streamlit", "run", "app.py"],
|
|
env={"OPENAI_API_KEY": "...",
|
|
"FIRECRAWL_API_KEY": "..."}
|
|
)
|
|
res = streamlit_server.create()
|
|
|
|
print("✨ Streamlit server hosted at:", res.url) |