Documentation Index
Fetch the complete documentation index at: https://docs.selftune.dev/llms.txt
Use this file to discover all available pages before exploring further.
Usage
Live action feed
The dashboard includes a live action feed that shows real-time output from selftune commands — whether triggered from the dashboard UI or run directly in your terminal. The simplified lifecycle aliases are the primary teaching surface in the feed. Older stage-level commands still appear there as advanced actions when you run them directly. When you run any of the following commands in your terminal, their output automatically appears in the dashboard feed:| Command | Action shown |
|---|---|
selftune verify | Verify draft |
selftune publish --no-watch | Ship candidate |
selftune publish | Monitor live |
selftune improve | Ship candidate |
selftune improve --dry-run | Replay dry-run |
selftune run | Run loop |
selftune eval generate | Generating evals |
selftune eval unit-test --generate | Generating unit tests |
selftune create check | Verify draft (advanced alias) |
selftune create replay | Replaying draft package |
selftune create baseline | Measuring draft baseline |
selftune create publish | Publishing draft package (advanced alias) |
selftune grade baseline | Measuring baseline |
selftune evolve | Deploying candidate |
selftune evolve --dry-run | Dry-run replay |
selftune watch | Monitor live |
selftune orchestrate | Run loop (advanced alias) |
selftune evolve rollback | Rolling back |
- Status — started, running, or finished
- Skill name — which skill the action targets
- Live output — stdout and stderr streamed in real time
- Result — success or failure with exit code
verify, create report, and publish runs, the live-run
screen also shows package candidate lineage when it is available:
- Candidate ID — the durable key for the evaluated package fingerprint
- Parent — the previous evaluated draft for the same skill, or
root - Generation — how far this draft is from the root candidate in the lineage
~/.selftune.
Dashboard actions
You can also trigger actions directly from the dashboard UI. The same live feed shows their output as they run — no need to switch to a terminal. Supported dashboard actions:| Action | What it does |
|---|---|
generate-evals | Generates evaluation cases for a skill |
generate-unit-tests | Generates unit tests from existing evals |
measure-baseline | Measures a draft package baseline or grades the current skill version |
replay-dry-run | Runs draft-package replay or an evolve dry-run replay |
deploy-candidate | Ships a validated draft package or deploys an improved candidate |
watch | Monitors live behavior, or publishes and immediately monitors a draft package |
rollback | Rolls back to a previous skill version |