Recall
AI-generatedSummary
Search a Hindsight memory bank for relevant memories matching a natural language query with optional tag filters and configurable recall budget and maximum token limits.
Inputs
- Bank ID (required) — The identifier of the Hindsight memory bank to search. If the bank does not exist, it will be created on first use.
- Query (required) — Natural-language query string used to search memories in the specified bank.
- Budget — Recall budget controlling how exhaustive the search is. Options are Low, Mid (default), and High.
- Max Tokens — Maximum number of tokens of memory content to return, defaulting to 4096.
- Tags Filter — Comma-separated list of tags to filter the recall search. Leave blank for no filtering.
Output shape
a single JSON object containing the recall results from the memory bank
The operation sends a POST request to the endpoint /v1/default/banks/{bank_id}/memories/recall with the query, budget, max_tokens, and optional tags filter. The returned JSON includes the memories relevant to the query. Errors are reported per item if continueOnFail is enabled.
Examples
Example 1: Recall relevant memories from a user's memory bank with default mid budget and no tags filter. Returns matched memories up to 4096 tokens.
Bank ID: user-123; Query: "recent project updates"; Budget: Mid (default); Max Tokens: 4096 (default); Tags Filter: none