Skip to content

knowledge-load

Loads and prepares knowledge base context for downstream agents.

Synopsis

/knowledge-load

/rp1-base/knowledge-load

Description

The knowledge-load command ingests the knowledge base from .rp1/context/ and prepares it for use by other agents. It builds an internal knowledge graph, extracts entities and relationships, and optimizes the context for memory constraints.

This command is typically called internally by other KB-aware agents, but can be invoked directly to verify the KB is loadable.

Parameters

Parameter Default Description
RP1_ROOT .rp1/ Root directory for KB artifacts
PROJECT_PATH . Project path (for monorepo subprojects)
FOCUS_MODE balanced Context allocation strategy
MEMORY_BUDGET Auto Memory limit for loaded context

Focus Modes

Mode Allocation Use Case
project 80% project / 20% system Feature development
system 50% project / 50% system Architecture analysis
balanced 60% project / 40% system General use

Loading Strategy

The command adapts its loading strategy based on repository type:

Single Project: - Loads: index.md, concept_map.md, architecture.md, modules.md, patterns.md

Monorepo Root: - Loads: index.md, architecture.md, dependencies.md, patterns.md - Optional: Project summaries

Monorepo Subproject: - Loads: dependencies.md, patterns.md, projects/{name}/*.md - Context: System architecture and shared components

Examples

Basic Usage

/knowledge-load

/rp1-base/knowledge-load

Expected output (single project): 

READY

Expected output (monorepo): 

READY [monorepo: 2 projects - rp1-base, rp1-dev]

Error Response

If the KB doesn't exist or is corrupted:

ERROR: Required documentation files not found. Run /knowledge-build first.

Output

The command returns a simple status:

Response Meaning
READY Single project KB loaded successfully
READY [system: N projects] Monorepo root KB loaded
READY [project: name] Monorepo subproject KB loaded
ERROR: message Loading failed with reason

Memory Budget Management

If the KB exceeds memory budget, compression is applied in this order:

  1. Remove tertiary project details (keep interfaces only)
  2. Compress unused shared code
  3. Summarize verbose descriptions
  4. Remove historical information
  5. Compress target project details (last resort)

See Also