load-context

/load-context Command

Load large files into Recall's RLM (Recursive Language Model) system for processing content that exceeds context window limits.

Usage

/load-context <file_path> [task_description]
/load-context <glob_pattern> [task_description]

Examples

/load-context /var/log/app.log "Find all errors and their causes"
/load-context src/**/*.ts "Analyze code patterns and potential issues"
/load-context ./large-document.pdf "Summarize key points"

Behavior

When invoked, this command will:

1. Read the specified content:

   • Single file: Read the entire file
   • Glob pattern: Find matching files and concatenate content

2. Analyze content size:

   • If <25K tokens: Suggest direct analysis (RLM not needed)
   • If >25K tokens: Proceed with RLM

3. Create execution context:

   • Store content externally in Recall
   • Generate chain_id for tracking
   • Detect optimal decomposition strategy

4. Return setup information:

   • Chain ID for subsequent commands
   • Estimated token count
   • Recommended strategy
   • Suggested next steps

Output

After loading, you'll receive:

RLM Context Created
-------------------
Chain ID: 01HXYZ...
Tokens: ~125,000
Strategy: filter (recommended)

Next steps:
1. Run /decompose 01HXYZ... to break down the analysis
2. Or manually call decompose_task with a custom strategy

Notes

• Large files (>1MB) may take time to process
• The task description helps optimize the decomposition strategy
• Use glob patterns to analyze multiple related files together
• Chain IDs are needed for all subsequent RLM operations