A memory workflow that loads only useful context to save tokens
A custom setup helps a pull only the knowledge needed for the current conversation. The knowledge is stored in Markdown files, and each file includes a short condensed story field.
A turns that condensed text into , then searches for the pieces that are closest to the new question. Only the most relevant pieces are added to the model conversation, instead of flooding the model with a large .
The goal is to work around limits, improve answer quality, and reduce token use. This is especially useful for scientific notes and other complex knowledge areas where the right small detail matters.
Key points
- Knowledge is kept in Markdown files with a short condensed story field.
- A creates and finds material similar to the current question.
- Only selected context is loaded into the model conversation.
- The approach aims to reduce token use and avoid overloading the .
- It fits complex knowledge work, such as scientific research notes.