Rhippo
What We Do
We believe you should be able to communicate with LLMs
in a collaborative setting the same you communicate with other people.
LLM's currently cannot remember, reason about, combine,
or innovate from multiple people’s ideas and preferences
in a team or classroom, rendering them oftentimes unhelpful.
We use SOTA tech to retrieve only relevant information and compose contextualized prompts
under the hood while your chatbot experience remains completely uninterrupted.
Our larger vision is that in order to effectively scale AI models in the future,
the intermediate representations themselves must be compressed. Our novel
compression based embedding models allow us to do more fine-grained and therefore
more effective memory retrieval at scale.
Asychronous productivity
Rhippo agents are continuously working, even after you log-off. During idle time, every Rhippo agent will analyze and combine ideas from different conversations in the team. It can then send a message to a user, detailing a new idea it thought of, refine a previously insufficient answer with updated information, or simply store novel ideas to better answer a question in the future.
Remember personal preferences
Add personal preferences by writing, or by filling out a form. These personal preferences will be composed by our embddding models and language models into more organized forms and stored to be referenced during a later query. Your data (conversations, preferences, etc.) are all stored through Google Drive on your own account - we store zero user data on our servers.
Knowledge injection
Inject LLMs with prior knowledge using any documents you would like: pdf documents, messages, powerpoints, screenshot, you name it. Our fine-grained embedding-based retrieval technology will make sure only relevant prior knowledge is leveraged to answer your queries.
Zero activation energy
Users simply install our free chrome extension - they can continue using their favorite LLM chatbots, completely uninterrupted.
DEMO
How it Works
Rhippo turns any LLM into a productive team member, study buddy, or teaching assistant by maintaining, updating, and referencing a digital brain.
Start by creating a profile through our web interface to store your personal preferences, past conversations, and anything else you would like your digital brain to know. Your profile links directly to your Google Drive for secure storage. Our embedding and language models will retrieve and compose relevant memories from these digital brains into a contextualized prompt under the hood. User preferences, teammate's ideas, past conversations, and teammate's unique expertises are all considered to construct the most useful answers, no matter the chatbot.