Actions46
- Bank Actions
- Directive Actions
- Document Actions
- Entity Actions
- Memory Actions
- Mental Model Actions
- Operation Actions
- Webhook Actions
Memory → Reflect
AI-generatedSummary
Generate an AI response reflecting on stored memories within a specified bank based on a given question or prompt.
Inputs
- Bank ID (required) — ID of the bank to reflect on
- Query (required) — Question or prompt to reflect on using stored memories
- Budget — Search thoroughness before generating response; options are Low, Mid, High
- Include Facts — Whether to include evidence (memories, mental models, directives) used to generate the response
- Include Tool Calls — Whether to include execution trace of tool and LLM calls
- Max Tokens — Maximum tokens for the generated response
- Response Schema — JSON Schema for structured output, ensuring response conforms to this schema if provided
- Tags — Comma-separated tags to filter accessible memories
- Tags Match — How to match multiple tags; options are Any, All, Any Strict, All Strict
Output shape
a single generated reflection response object as returned by the Hindsight API
The response includes the AI-generated content reflecting on stored memories filtered by provided inputs. Optional fields like evidence and execution traces are included based on input flags. Response respects the provided maximum token limit and, if a response schema is specified, conforms to that schema.