Saltar a contenido

Salud de Sesión

RaiSE v2.4 adds two features that maintain session hygiene automatically: Session Doctor detects and cleans stale session state, and Workstream Monitor provides productivity insights at session close.

Session Doctor

What It Does

The Session Doctor runs automatically on rai session start. It checks for stale or orphaned session state in .raise/rai/personal/ and cleans up what's safe to remove.

What It Detects

Finding Severity What It Means Auto-Clean?
Zombie session warning Active session pointer >48h old, no content to preserve Yes
Zombie with content warning Active session pointer >48h old, has narrative/state No — needs your review
Stale output info session-output.yaml >24h old Yes
Retention exceeded info >20 session directories accumulated No — batch decision

How It Works

The doctor follows a 3-phase pattern:

  1. Diagnose — scan for issues, no side effects
  2. Classify — separate auto-safe from needs-consent items
  3. Execute — clean only authorized items

Items classified as "auto-safe" (empty zombies, stale output files) are cleaned without prompting. Items that might contain work you haven't saved require explicit consent.

Example Output

Session Doctor — 2 finding(s):

  [!] Zombie session: S-E-260401-0900 (72h old)
      No content to preserve
      Action: Clear stale pointer

  [i] Stale session-output.yaml (26h old)
      26.1 KB — safe to remove
      Action: Remove stale output file

Auto-cleaned:
  - Cleared zombie pointer: S-E-260401-0900
  - Removed stale output: session-output.yaml

Running Manually

To inspect without modifying anything:

rai session doctor

Skipping (CI/Automation)

For CI pipelines or scripts that don't need health checks:

rai session start --no-doctor

Configuration

Default thresholds (not yet user-configurable):

Threshold Default Purpose
Zombie age 48 hours When an active session pointer is considered stale
Stale output age 24 hours When output files are considered safe to remove
Directory retention 20 Maximum session directories before suggesting cleanup

Workstream Monitor

What It Does

The Workstream Monitor analyzes your git history at rai session close and provides brief productivity insights. It's purely advisory — it suggests improvements but never blocks your work.

What It Measures

Metric Source What It Tells You
Commit count git log Session productivity
Test commit ratio Commit subjects starting with test TDD compliance
Revert count Commit subjects starting with Revert Stability of approach
Duration First to last commit timestamps Session time span

Insights

The monitor suggests improvements when patterns indicate room for growth:

Low TDD compliance (test ratio <30%):

Test commit ratio is 2/14 (14%). Consider writing test commits before implementation (RED-GREEN-REFACTOR).

High revert frequency (>2 reverts):

3 reverts detected. Consider smaller, more focused commits to reduce revert frequency.

Example Output

Session S-E-260403-1000 closed.
  Patterns added: 2
  Corrections recorded: 1

  Session insights: 14 commits | Test ratio: 28% | Reverts: 1 | Duration: 95m

Future (v3.0)

The current monitor uses simple git heuristics. In v3.0, it will support:

  • Cross-session trend detection
  • Team-level pattern recognition
  • Deeper analysis via async agents