build_context
Retrieves and compresses relevant memories from a knowledge graph to build a compact context pack for the current task, resolving contradictions and staying within a token budget.
Instructions
Recursively retrieves and compresses relevant Waggle memory for the current task, using graph, hybrid, transcript, update, and conflict-aware retrieval. Decomposes the query into targeted subqueries, expands the graph around key nodes, resolves contradictions and superseded memories, and returns a compact context pack under a configurable token budget. Aliases: recursive_context, assemble_context, rlm_context.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| mode | No | Retrieval depth mode: 'fast' runs fewer subqueries for low latency; 'balanced' is the default; 'deep' adds extra subqueries for thorough coverage. | balanced |
| depth | No | Graph expansion depth around retrieved nodes. | |
| query | Yes | Current user task or question to build context for. | |
| project | No | Optional project or workspace name used to partition memory. | |
| agent_id | No | Optional agent or client identifier used to partition memory. | |
| session_id | No | Optional conversation or run identifier used to partition memory. | |
| token_budget | No | Maximum token budget for the context pack (approximate). | |
| max_subqueries | No | Maximum number of decomposed subqueries to run. | |
| include_evidence | No | Whether to include verbatim transcript evidence in the context pack. |