thesys-genui-sdk 0.1.2

Thesys GenUI utilities

Thesys GenUI SDK

⚡ Building application with generative UI ⚡

Quick Install

pip install thesys_genui_sdk

🤔 What is this?

Thesys is a developer platform for building generative UI applications. There are 2 major components of building your own generative UI application:

• API - Thesys C1 API is our flagship Generative UI API that leverages the power of LLMs to generate UI on the fly.

• Frontend - GenUI SDK is our React framework that integrates well with the C1 API and allows you to present your end user with a live app that they can interact with.

This library is designed to help you build your own generative UI applications by providing you with the tools to build your own application on top of Thesys C1 API.

Read more abount C1 api here and what is genui here

🤖 Chatbots

• Documentation
• Fast api template
• Examples

📖 Documentation

Streaming response with fast-api:

1. wrap your route with the with_c1_response decorator:
2. use write_content, write_think_item, write_custom_markdown from thesys_genui_sdk.context to write to the response stream
3. use get_assistant_message from thesys_genui_sdk.context to get the assistant message for the history

from thesys_genui_sdk.fast_api import with_c1_response
from thesys_genui_sdk.context import write_content, get_assistant_message, write_think_item

@app.post("/generate_ui")
# decorator to wrap the route with the c1 response instance
# this will create a c1 response instance and make it available in the context
# this expects an async function which will use write_content, write_think_item, write_custom_markdown
# to write to the response stream
@with_c1_response()
async def generate_ui(request: Request):
    await generate_llm_response(request)


async def generate_llm_response(request: Request):
    stream = openai_client.chat. ....

    # write a think item to the response stream
    await write_think_item("Thinking", "I am thinking about the user's request")

    for chunk in stream:
        delta = chunk.choices[0].delta
        finish_reason = chunk.choices[0].finish_reason

        if delta and delta.content:
            # write a content chunk to the response stream
            await write_content(delta.content)

        if finish_reason:
            # get the assistant message for the history
            assistant_message_for_history = get_assistant_message()
            # store the assistant message for the history

Framework independent response streaming:

1. create a c1 response instance
2. create an async task and use write_content, write_think_item, write_custom_markdown methods on the c1 response instance to write to the response stream
3. use c1_response.get_assistant_message to get the assistant message for the history

from thesys_genui_sdk.context import C1Response

c1_response = C1Response()

async def generate_llm_response():
    # write a think item to the response stream
    await c1_response.write_think_item("Thinking", "I am thinking about the user's request")

    # write a markdown chunk to the response stream
    await c1_response.write_custom_markdown("## Hello, world!")

    for chunk in stream:
            delta = chunk.choices[0].delta
            finish_reason = chunk.choices[0].finish_reason

            if delta and delta.content:
                # write a content chunk to the response stream
                await c1_response.write_content(delta.content)

            if finish_reason:
                # get the assistant message for the history
                assistant_message_for_history = c1_response.get_assistant_message()
                # store the assistant message for the history

    # end the response stream
    await c1_response.end()

asyncio.create_task(generate_llm_response())

# use the stream from c1_response to send to the client
stream = c1_response.stream()