improve

name	improve
description	Run one RL improvement tick on the workspace via POST /api/improve. Ships one verifiable change, scores it, writes the scorecard. The thing you pay for. Triggers on: improve, make this better, ship one thing, run a tick, get smarter.
version	1.0.0
tags	["rl","improve","reward","tick","autopilot"]

/improve

Runs one improvement tick on the workspace. Calls POST /api/improve on the backend, which plans one task, builds it, verifies it, and scores it. Returns what shipped + the reward. Writes the scorecard locally.

This is the product. The thing the user pays for. One call, one verifiable result.

How it works

/improve
  → POST /api/improve { workspace: ".", mode: "full" }
  → backend picks a task, plans, builds, reviews, verifies
  → returns { task, reward, files_changed, verify_pass, summary }
  → CLI writes scorecard to .atris/presidio/scorecards.md
  → CLI reports result to user

The inference is Claude Code (or whatever model the backend uses). The environment is the folder. The endpoint is the bridge.

On invoke

Run the CLI command — it does the whole tick (auth, the credit-metered call, scorecard, fallback):

atris improve            # one full tick: plan → build → verify → score (deducts credits)
atris improve plan       # show the plan only, change nothing
atris improve --json     # machine-readable result (this is what the member loop consumes)
atris improve --no-fallback   # fail loudly instead of running a local tick when the backend is down

Under the hood atris improve (commands/improve.js):

1. Loads the auth token via utils/auth.loadCredentials
2. POST /api/improve { workspace, mode, model } via utils/api.apiRequestJson
3. The backend plans, builds, runs the verify command, scores it, and deducts Atris credits per successful tick (bill_tick)
4. Writes a per-tick scorecard row to .atris/state/scorecards.jsonl (the receipt the brain ledger counts)
5. Falls back to a local autopilot tick only when you are not logged in or the backend is unreachable — a real error (insufficient credits, server error) is reported, never silently retried

The full-mode response does not echo credits_deducted (credits are still billed server-side), so the CLI shows "billed server-side" when the count is absent. To call the endpoint directly instead of the command, POST /api/improve with { workspace, mode, model }.

Modes

• full — plan, build, review, verify (default)
• plan — just pick the task and show what it would do
• dry_run — run everything but don't commit

Fallback

If the backend is unreachable (no auth, no network, localhost not running), fall back to local mode: run atris autopilot --auto --iterations=1 instead. Same loop, just local inference via claude -p subprocess. Report that it ran locally.

Output

improved.

  task:    fixed the stale wiki ref in auth-flow.md
  verify:  pass (npm test, 143/143)
  reward:  +4
  files:   atris/wiki/briefs/auth-flow.md
  time:    47s

  scorecard updated.

Rules

• One tick only. Never batch.
• Always verify. No reward without a check.
• Show what shipped, not what was attempted.
• Write the scorecard. This is the receipt.
• If verify fails, halt honestly and write a lesson.
• Fallback to local if backend is unreachable. Never error silently.
• The user pays because something real happened. Never fake it.