Skip to content
Reflex

Pulse

Pulse turns Reflex from a query tool into one that proactively surfaces information about your codebase. It generates a self-contained static site with three browsable surfaces grounded in index data.

Quick start

Section titled “Quick start”
Terminal window
# Generate the Pulse site
rfx pulse generate --output ./pulse-site


# Serve it locally
rfx pulse serve

That’s it — open your browser and explore your codebase’s structure, history, and architecture.

What gets generated

Section titled “What gets generated”

The generated site includes three surfaces:

Digest

Section titled “Digest”

A periodic summary of structural changes — not a git log, but an analysis of how your code’s architecture is evolving. The digest reports on:

  • Module activity and growth
  • New symbols and removed symbols
  • Dependency changes
  • Hotspot movements
  • Convention drift

Wiki

Section titled “Wiki”

One auto-generated documentation page per module. Each page includes:

  • What the module does
  • Its dependencies and dependents
  • Key symbols
  • Recent structural activity

Architecture map

Section titled “Architecture map”

A dependency graph at varying zoom levels, rendered as interactive diagrams.

Generation options

Section titled “Generation options”
Terminal window
rfx pulse generate \
  --output ./pulse-site \
  --title "My Project" \
  --theme dark \
  --no-llm           # Structural data only, no LLM narration

Snapshots

Section titled “Snapshots”

Pulse is built on snapshots — point-in-time captures of your codebase’s structural state. Comparing snapshots produces the diffs that power the digest.

Terminal window
# Take a snapshot
rfx snapshot


# Compare snapshots
rfx snapshot diff


# List snapshots
rfx snapshot list


# Clean old snapshots
rfx snapshot gc

Retention policy

Section titled “Retention policy”

By default, Pulse keeps 7 daily, 4 weekly, and 12 monthly snapshots (~23 total). Configure retention in your project config:

[pulse.retention]
daily = 7
weekly = 4
monthly = 12

LLM narration

Section titled “LLM narration”

By default, Pulse uses structural data only (--no-llm equivalent). When an LLM provider is configured (via rfx llm config), Pulse can add narrative explanations grounded in the structural data.

Every narrative claim is backed by structural evidence — the LLM adds readability, not speculation.

Design philosophy

Section titled “Design philosophy”

Pulse follows a “structure first, prose only when grounded” principle:

  • Structural claims come from the index (zero LLM)
  • Narrative claims require grounding evidence
  • Every assertion is traceable to index data

Next steps

Section titled “Next steps”