rlm

RLM - Recursive Language Model Protocol

Based on MIT's RLM paper (arXiv:2512.24601) and DSPy's structured REPL pattern. Explore data programmatically — only printed results enter context.

Tokens are CPU, not storage. Never dump raw data into context. Write code to extract what matters, print only the summary.

When to Use

• File/data too large for context window (logs, databases, binaries)
• Codebase-wide analysis (grep across 100+ files, dependency graphs)
• Multi-step data extraction where each step depends on prior results
• Any task where raw data would burn tokens without adding value

Decision Logic

Size	Protocol
< 5KB	Read directly — no RLM needed
5KB-500KB	Steps 1-3 (METADATA, PEEK, SEARCH)
500KB+	Full 6-step protocol with sub-agent decomposition

6-Step Protocol

Execute IN ORDER. Each step uses python3 -c (or python -c on Windows). Raw data never enters context. Prefer python scripts over PowerShell for data processing.

Step 1: METADATA

Assess file type, size, line count, preview (200 chars). Use glob tool for multi-file discovery (never find via Bash). WebFetch is blocked — download via python3 urllib, save locally, then process.

Step 2: PEEK

Sample head (20 lines), tail (10 lines), random slices to understand structure.

Step 3: SEARCH

Targeted extraction: regex, AST parsing, JSON key traversal based on PEEK findings.

Step 4: ANALYZE (500KB+ only)

Sub-agent decomposition — up to 15 sub-queries. Use @explore sub-agents for parallel chunk analysis. Extract each chunk via python3 -c, pass to sub-agent:

with open('/path/to/file') as f:
    lines = f.readlines()
chunk = lines[START:END]
print(f'=== Chunk N ({len(chunk)} lines) ===')
for l in chunk: print(l.rstrip())

Sub-query types: chunk analysis, cross-reference, semantic filter, recursive drill.

Step 5: SYNTHESIZE

Combine findings, cross-reference, resolve conflicts.

Step 6: SUBMIT