Contextual

Home screen
Selecting context
Using context

Inspiration

Nobody likes saying things twice.

If you're anything like us, your ChatGPT history is probably a mess: hundreds of chats that you won't use, can't delete, and are too scared to search through. Some LLMs remember nothing, and some remember everything, but wouldn't it be great if you could tell it what to remember?

Introducing Contextual, a dynamic AI context-manager. Our mascot is a parrot, because you'll never have to say anything twice again. 🦜

What it does

Contextual is a web app built on top of your favourite LLMs, allowing you to select, delete, and combine your chat history into a concise context package. It does this by:

Parsing your entire LLM history and sorting it into topics
Automatically selecting your top chats that relate to your new prompt
Allowing you to include or exclude whichever chat sections you want

How it makes your life easier:

You never have to search through your GPT history again
You can choose which parts of the chat to exclude (in case it ever went off-topic)
You can emphasize which things you want the LLM to remember more
It works identically to your favourite chat client by leveraging Groq's ultra low-latency responses, so you don't need to change your workflow

How we built it

Your entire GPT history is downloaded as a .json, which is then parsed into a .txt
It's then sectioned into topics with Groq for on-the-spot decision making and Claude for response generation while preserving the key ideas
When a user starts a new chat, they're presented with their 5 most similar past chats. They can add or delete chats, or even parts of chats.
All of this can be processed by the LLM in seconds - It's like a GPT that can read your mind!