drayGet started

Understand AI memory 2 min read

AI memory vs. chat history

Chat history records the conversation. Memory selects information for later use. Use history when exact wording or the sequence of decisions matters; use a memory record as a starting point for finding relevant context.

The answer can depend on what happened next

A summary might say that a team chose a monthly plan. The next message could have made that choice conditional on a customer interview. If you only read the summary, a provisional choice can look final.

Open the surrounding messages when you are checking a quote, a commitment, a price, or the reason for a decision. A short extracted fact is useful for discovery, but it cannot carry every qualification from a long exchange.

Choose the smallest source that answers the question

For a tone preference, a short note with an example may be enough. For a disagreement about what was approved, read the original exchange. You do not have to move an entire archive into the next assistant just to keep one settled requirement.

  • Need exact wording? Find the original message.
  • Need the current position? Check later corrections too.
  • Need a starting brief? Use a summary, then verify the important details.

TAKE THIS WITH YOU

Ask for the record, then the conclusion

An illustrative template. Replace the details with material you are allowed to share.

Find the conversation where we discussed the monthly plan. Show the decision and any conditions attached to it. Check whether a later message changed that decision. Separate quoted text from your summary. If you cannot access the original chat, say so.

USING THIS WITH DRAY

Keep the context. Check the source.

Dray keeps a local chat archive and processes conversation content for memory. What is available depends on capture scope and sync; a connected account does not imply a complete history.

Dray is in beta. Available sources and connections depend on your setup. Read the data notes before granting access.

Sources & next reads

Product details are grounded in the guides below. Example projects and prompts are illustrative, not customer results.