A Reflex custom component chat.
pip install reflex-chat
See the chat_demo
folder for an example app.
import reflex as rx
from reflex_chat import chat, api
@rx.page()
def index() -> rx.Component:
return rx.container(
rx.box(
chat(process=api.openai()),
height="100vh",
),
size="2",
)
app = rx.App()
- Import the
chat
component to your code.
from reflex_chat import chat
- Specify the
process
function that will be called every time the user submits a question on the chat box. Theprocess
function should be an async function that yields after appending parts of the streamed response.
We have a default process
function that uses the OpenAI API to get the response. You can use it by importing the api
module. Over time we will add more process
functions into the library.
To use the OpenAI API, you need to set the OPENAI_API_KEY
environment variable. You can specify the mdoel with the OPENAI_MODEL
environment variable or pass it as an argument to the api.openai()
function.
chat(process=api.openai(model="gpt-3.5-turbo")),
See below on how to specify your own process
function.
3. Add the `chat` component to your page.
By default the component takes up the full width and height of the parent container. You can specify the width and height of the component by passing the `width` and `height` arguments to the `chat` component.
```python
@rx.page()
def index() -> rx.Component:
return rx.container(
rx.box(
chat(process=api.openai(model="gpt-3.5-turbo")),
height="100vh",
),
size="2",
)
Once you create a chat component, you can access its state through the chat.State
object.
Get the messages from the chat state.
chat1 = chat()
@rx.page()
def index() -> rx.Component:
return rx.container(
# Get the messages through chat1.State.messages.
rx.text("Total Messages: ", chat1.State.messages.length()),
# Get the last user message through chat1.State.last_user_message.
rx.text(chat1.State.last_user_message),
rx.hstack(
chat1,
height="100vh",
),
# Get the processing state through chat1.State.processing.
background_color=rx.cond(chat1.State.processing, "gray", "white"),
size="4",
)
You can specify your own process
function that will be called every time the user submits a question on the chat box. The process
function should be an async function that takes in the current chat state and yields after appending parts of the streamed response.
The OpenAI process
function is defined as below:
async def process(chat: Chat):
# Start a new session to answer the question.
session = client.chat.completions.create(
model=model,
# Use chat.get_messages() to get the messages when processing.
messages=chat.get_messages(),
stream=True,
)
# Stream the results, yielding after every word.
for item in session:
delta = item.choices[0].delta.content
# Append to the last bot message (which defaults as an empty string).
chat.append_to_response(delta)
yield
return process