Best Approach for Text Chunking in AI Chat Agent using React and Django (Langchain) without Sockets?
I’m building an AI chat agent that needs to handle text chunking and display responses word by word in real-time. My setup involves a React (frontend) and Django Langchain (backend) integration. I want to achieve this without relying on sockets to stream data and instead use an approach that can fetch and display responses as they are generated, ensuring the user sees a continuous flow of words.
Here's a breakdown of my current setup:
Frontend: React (with or without Redux) Backend: Django integrated with Langchain for handling the AI logic What I’m Looking For: Text chunking: Best practices for chunking AI responses (word by word or partial sentences) and updating the UI in React. Real-time updates: How to fetch and display this data incrementally (e.g., fetching a stream or chunks) without using sockets. Efficient handling: Strategies for making the experience smooth for users, especially on the frontend with React.
I prefer not to use WebSockets and want to stick with traditional HTTP requests if possible. Looking for an approach that handles data/response chunking seamlessly so users get real-time updates without long waits. If anyone has experience with Langchain in this context or can share ideas on how to achieve this using REST APIs or another efficient method, I’d love to hear your suggestions.
I presume, of you wrote a const array ready to be called, you could 'push' a text chunk, but as a common case, library limitation could conflict, and might not allow conversion, particularly on code attributes that may be common to other languages. For respondance however, your 'look' may not be replied to, but on a word-to-word basis you may get a partial sentence or another word, and even better of you can define that in your array. You can display of the same way by either calling directly by input or calling an array variable name by input, if you can fix that back to the language. Web sockets needn't be intertwined with, and you can let the browser do its best compatible for you, if you use api keys, or in case of open resources like wikipedia, just use the searcharea code.