prompt.txt

Create a simple memory system for AI Agents. It stores the memory in the form of Question & Answer format. 
You can think of it as semantic version of key-value store.

The developers can use it in their projects. This project requires LLM for Embedding, Generating response and vector store.

get(question, topk, thresold=default) method returns the closest topk questions (and answer tuples) to the question based on the embeddings of question.

save(question, answer) method saves it after deduplication. It uses the LLM for deduplication. It first looksup() for the question and then if there are already existing questions, it fetches them and asks LLM to come up with final list. It then deletes the old questions and adds new questions. This ensures that the question/answers do not explode. It also ensure that the older information is removed.