Hosting LLM APIs and Apps

In this example we will be using the streamlit. Streamlit script should be in same folder with this notebook. If you inspect streamlit code look under 'Supplementary Files'.

import practicuscore as prt

# When you finish test, stop this cell. If you dont stop cell always be open.
prt.apps.test_app()

After testing our application we can set our configurations and start the deployment process.

Define params from region

import practicuscore as prt
region = prt.get_region()

We receive the necessary information for deploy from the region

my_app_list = region.app_list
display(my_app_list.to_pandas())
app_name = my_app_list[0].name
print("Using first app name:", app_name)

my_app_prefix_list = region.app_prefix_list
display(my_app_prefix_list.to_pandas())
app_prefix = my_app_prefix_list[0].prefix
print("Using first app prefix", app_prefix)

my_app_settings = region.app_deployment_setting_list
display(my_app_settings.to_pandas())
deployment_setting_key = my_app_settings[0].key
print("Using first setting with key:", deployment_setting_key)

prt.apps.deploy(
    deployment_setting_key=deployment_setting_key, # Deployment Key, ask admin for deployment key
    prefix=app_prefix, # Apphost deployment extension
    app_name=app_name, 
    app_dir=None # Directory of files that will be deployed ('None' for current directory)
)

Supplementary Files

streamlit_app.py

# The below is official Streamlit + Langchain demo.

import streamlit as st
import practicuscore as prt
from practicuscore.gen_ai import PrtLangMessage, PrtLangRequest
import requests


prt.apps.secure_page(
    page_title="🦜🔗 Quickstart App" # Give page title
)


st.title("🦜🔗 Quickstart App v2") # Give app title


# This function use our 'api_token' and 'endpoint_url' and return the response.
def generate_response(messages, model):

    api_url = "Enter your model api"
    token ="Enter your model token"

    practicus_llm_req = PrtLangRequest( # This class need message and model and if u want to stream u should change streaming value false to true
        messages=messages, # Our contest
        lang_model= model, #"gpt-4o", # Select model
        streaming=True, # Streaming mode
        llm_kwargs={"kw1": 123, "kw2": "k2"} # If we have a extra parameters at model.py we can add them here 
    )

    headers = {
    'authorization': f'Bearer {token}',
    'content-type': 'application/json'
    }

    data_js = practicus_llm_req.model_dump_json(indent=2, exclude_unset=True) # Convert our returned parameter to json

    with requests.post(api_url, headers=headers, data=data_js, stream=True) as r: 
        for word in r.iter_content(1024):
            yield word.decode("utf-8")

def reset_chat():
    st.session_state["messages"] = []


# Save chat history
if "messages" not in st.session_state:
    st.session_state["messages"] = []

# Input chat
user_message = st.chat_input("Write message")
if user_message:
    # Add user message to history
    st.session_state.messages.append({"role": "user", "content": user_message})

    human_msg = PrtLangMessage(
        content=user_message,
        role = "human"
    )

    # Show messages
    for msg in st.session_state.messages:
        st.chat_message(msg["role"]).write(msg["content"])

    with st.chat_message("assistant"):
        response = st.write_stream(generate_response([human_msg], "gpt-4o"))

    # Show answer
    st.session_state.messages.append({"role": "assistant", "content": response})

# Create a container to hold the button at the bottom
with st.container():
    st.write("")  # Add some empty space to push the button to the bottom
    if st.button("Clear history"):
        reset_chat()
        st.success("History has been cleaned")

---

Previous: Sdk Streamlit Hosting | Next: Langflow LLM Apphost > Langflow Streamlit Hosting