A second brain: throw thoughts in, ask in your words

Notes pile up where they are convenient to write and cannot be found where they are needed: keyword search fails because a month later you remember the meaning, not the wording. Here thoughts land in a knowledge base, and you can search it the way you think.

What is inside

1. A thought or a question — the entry, one way in for both.
2. Is this a note or a question? — intent parsing.
3. The branch splits into writing and searching.
4. Write to the base stores the text as it is, Search what was written finds the relevant pieces, and Answer from the records phrases an answer from those only.

The base grows out of the conversation. Normally you fill one by uploading files; here it is enough to write the bot a thought, a link, a conclusion after a meeting — and a month later that is what is worth asking about.

The port beats the field, and here that decides everything

The write and search branches get plain execution edges. On a write node a value arriving on the text port wins over the template in the settings: wire the branch with an ordinary data edge and the parsed intent goes into the base instead of the thought. The bot keeps answering “saved”, the base keeps filling with internal objects, and you only notice when search stops finding anything.

What to set up for yourself

• The base id appears in three places: the write, the search and the base node. Change all three, or the bot writes into one base and searches another.
• Intent parsing. A question is currently told from a note by its shape. If you often write statement-questions, add an explicit trigger word.
• The answer prompt already carries the “records only” rule — keep it, or the bot starts supplementing your memory with its own.

What to connect

Create an empty knowledge base under Documents and paste its id. Nothing needs uploading: the scenario fills it. A Telegram bot is needed if you would rather write thoughts into a messenger than into a web chat.