Keymate.AI is becoming a human competent AI search engine model for AI: Human intent guided all in one AI search for any AI.

Keymate.AI is doing this by fine-tuning a model for RAG. (Retrieval Augmentative Generation)
​

We are happy to announce that we are implementing a neural search engine model by integrating our current tech with another tech with a vendor and together we will be able to support these below
​

​1. File uploads below will be supported

​2. It grabs the correct data and optimizes ChatGPT and GPT-4 context window from unlimited data (it will use Keymate.AI ONE Search Model)
Your knowledge base will instantly store and index search results retrieved during your chats. Your knowledge base will automatically grow and your responses will be grounded by information from the internet (via Google Search), your uploaded files, your past interaction history, your past search results, your ready to be used knowledge bases and more.
​
​3. New Human Competent Search engine for AI
The search infrastructure we are implementing is the best of it's kind because we signed an NDA and I can't share the details. It allowed me to talk with the whole book I didn't have time to read. It gets the book's related information or human POV insights about the ideas from the internet. It constantly fills in information from your past so your chat will be always personalized.
​
​4. API is having a new baby and it will have memory!
When you pass Keymate API user API token you will be able to access user's personalized knowledge base instantly searching anything ( optionally improving results with google assisted search)
​
​5. Moving to a future to empower tech product builders
Keymate's Search will transform the way tech products are built and deployed. Through Keymate's APIs and new Keymate.AI ONE search you will get instant answers about the feature you want to implement. Keymate will increase knowledge base sources by adding github, jira connector bringing developers and tech product managers to the same page where they will share and get help while building tech products. You will be able to work on product requirements, product features, directly code reading and editing, grabbing the latest code examples from the latest libraries with the help of Keymate's ALL IN ONE search.
​
6. Transforming how people work: To revolutionize the workday for knowledge workers, our vision is to harness innovative tools and methodologies to effectively double productivity, enabling the completion of an 8-hour workload in just 4 hours. This transformation aims not only to boost efficiency but also to enhance work-life balance, foster creativity, and cultivate a more fulfilling and sustainable work environment for tech workers.

7. Future AI Agents will be built via Keymate API with the help of Keymate Ultimate:
You will use Keymate Ultimate to build with Keymate API we will provide a ready to use knowledge base for you to load and you will be able to load any web powered data to knowledge base by using AI Agents you will build with Keymate API. Let keymate grab recent data from all sources and feed your AI automation agent latest data specially optimized for each prompt.
​
​8. Keymate AI Alerts ( an AI powered Google Alerts ) will be also available for you to build with Keymate API. Set the alert based on your prompt on Google Search + LLM (GPT-4) and build any product you dream of:

Alert yourself if your public facing website gives any error with the help of AI

9. Business automations powered with Keymate AI

Use zapier-like no code tools to integrate Keymate's web search+browsing+scraping+Claude2+GPT-4 API to automate business cases.

Don't miss any data from the internet and knowledge base by leveraging Keymate.AI ONE Search and integrate LLM prompting to format input or output.
​
​10. Data recall to inject data to the prompt based on easy to remember human metadata tags
​When you store your files or data to your knowledge base you will be able to recall them with metadata tags this will allow you to use only required parts of your knowledge base. Your knowledge base will be further filtered based on the tags you use.
​
A prompt like:

​will allow Keymate.AI ONE model to curate your input window based on these focuses and it will filter out unnecessary input to be fed to language model completely. It will also answer under 45 seconds ( a requirement comes from being a ChatGPT plus plugin)
​
​How does Keymate.AI ONE search model work?

It uses vector and boolean direct text search ( sparse and dense) indexes together and reranks results based on diversity controlled by the Google search+Knowledge base results by another fine tuned model. So it understands your intents based on your ChatGPT conversation topic or prompt and learns how it should conduct search on your multiple knowledge bases. If intent is very specific it brings the most related results conversely if the intent is exploration it brings diverse results, allowing you to still get relevant results from your knowledge bases but this time exploration focused.
​
We started to use Claude 2 in as one of the PDF Readers in ChatGPT Plus, this technology will be an implementation of a fine tuned model or AI tech that is not belong to OpenAI or Anthropic or any other foundational model provider.

We will still continue using our old knowledge base search with OpenAI Ada embedding+Pinecone to allow you to use your older knowledge base but after Keymate.AI ONE launch new data inserted to Keymate will work better.
​
​Keymate.AI ONE (a human competent search engine model for other LLMs) will be launched for Alpha preview soon! Stay tuned!
​
Keymate.AI ONE will optimize search outputs for 3 different langauge models it will be used at Keymate Ultimate for Claude 2 100K and GPT-4 128K and it will be available to use right inside ChatGPT plus plugin.
​
It will also serve as a Keymate API so you can use it as the RAG layer to pass information to any LLM.

​
​
​
​