You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit c599e14
Browse filesBrowse the repository at this point in the historyBrowse files
- automatic promotion requires a correction, review or CI signal; reflections
backed only by captured tool failures are held
- auto marks are tied to the hash of the approved set the last automatic
promotion published, so any manual promote, rollback, pin or unpin voids
them without an extra write
- `Store.promote` publishes a supplied text and keeps the staged candidate,
so `candidate.json` is never overwritten and survives a crash
- a publish that landed before a later failure is reported as promoted with
a warning, not held
- cancellation and the run deadline are rechecked under the learn lock, and
the lock wait is bounded by the remaining time
- status shows only completed promotions; reservations still count toward
the daily limit; held reasons recorded for more paths
- tests isolate the limit env overrides; help and docs state the capture and
auto-reflect prerequisite and the env precedence over `learn disable`
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Copy file name to clipboardExpand all lines: docs/docs/configure/learn.md
+7-5Lines changed: 7 additions & 5 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -124,7 +124,9 @@ Off by default. With `learn.auto_promote` on, a candidate that automatic reflect
124
124
altimate-code learn enable --auto-promote
125
125
```
126
126
127
-
This writes `learn.capture`, `learn.auto_reflect` and `learn.auto_promote` to the project config. `ALTIMATE_LEARN_AUTO_PROMOTE=1` or `0` overrides the config for one process. Automatic promotion needs capture and automatic reflection, and `learn.enabled: false` or `ALTIMATE_LEARN=0` turns it off with the rest of learning. `learn disable` turns it off too.
127
+
This writes `learn.capture`, `learn.auto_reflect` and `learn.auto_promote` to the project config. Automatic promotion needs all three: `learn.auto_promote: true` or `ALTIMATE_LEARN_AUTO_PROMOTE=1` alone does nothing while capture or automatic reflection is off. `ALTIMATE_LEARN_AUTO_PROMOTE=1` or `0` overrides the config for one process.
128
+
129
+
`learn disable` writes `false` for all three, but environment variables override config: with `ALTIMATE_LEARN_CAPTURE`, `ALTIMATE_LEARN_AUTO` and `ALTIMATE_LEARN_AUTO_PROMOTE` set to `1`, automatic promotion stays on (`learn disable` reports which setting keeps learning active). `learn.enabled: false` or `ALTIMATE_LEARN=0` turns everything off, whatever the other settings say.
128
130
129
131
It runs only after automatic reflection (end of `run`, threshold, idle, startup recovery) on the default store, `team-playbook`. `learn reflect`, `learn bootstrap` and `learn import-reviews` always stage a candidate for review.
130
132
@@ -135,9 +137,9 @@ It runs only after automatic reflection (end of `run`, threshold, idle, startup
135
137
| Same candidate | The candidate is not exactly the one this reflection staged (it was edited, replaced, promoted, or rejected in the meantime). |
136
138
| Only this reflection | The candidate already had lesson changes waiting for review before this reflection: from `learn reflect`, `bootstrap`, `import-reviews`, or an earlier candidate that was held back. |
137
139
| Validation | It fails the checks `learn promote` runs: lint, hidden characters, coexistence, and overlaps. |
138
-
| Feedback |The reflection had no signal behind it, or its feedback was flagged as an instruction to the model. |
140
+
| Feedback |No signal from a person or an external check is behind the reflection: it needs at least one correction (`user_correction`, including `learn signal add --kind user`), `review`or `ci` signal. Repeated tool failures (`tool_retry`) are captured automatically, so on their own they only stage a candidate. Also held when the feedback was flagged as an instruction to the model. |
139
141
| Verification flags | A lesson it adds or changes mentions skipping or disabling verification. Those always need a person. |
140
-
| Your approvals | It would edit or remove a lesson a person approved. Only lessons that were promoted automatically, with their text unchanged and never pinned or unpinned, can be edited or removed automatically. Lessons from another checkout count as person-approved. |
142
+
| Your approvals | It would edit or remove a lesson a person approved. A lesson counts as automatic only while the approved set is exactly what the last automatic promotion published and the lesson is not pinned. Any `learn promote`, `rollback`, `pin` or `unpin`, or a change to `approved.json` from a pull, makes every approved lesson person-approved. Lessons from another checkout count as person-approved. |
141
143
| Size | It adds, edits, or removes more than `learn.auto_promote_max_changes` lessons (default 3), or changes only counters. |
142
144
| Rate | The store already had `learn.auto_promote_daily` automatic promotions (default 5) in the last 24 hours. |
143
145
@@ -152,7 +154,7 @@ learn: 1 signal -> +1 added; auto-promoted (previous lessons archived as v3). Un
152
154
learn: 1 signal -> +1 added; staged .altimate-code/learn/team-playbook/candidate.json for review (not auto-promoted: daily limit reached (5 automatic promotions in 24 hours; learn.auto_promote_daily=5)), review with `altimate-code learn show`
153
155
```
154
156
155
-
The TUI has no outcome line. `learn status` shows whether automatic promotion is on, the last automatic promotion with the lessons it added or changed and the ones it removed, and the last reason a candidate was held back (cleared by the next successful automatic promotion). `learn show` marks automatically promoted lessons `(auto-promoted)`. History records each one as an `auto-promote` entry with the lesson ids, removed ids, and the number of signals.
157
+
The TUI has no outcome line. `learn status` shows whether automatic promotion is on, the last automatic promotion with the lessons it added or changed and the ones it removed, and the last reason a candidate was held back (cleared by the next successful automatic promotion). A promotion that could not finish is not listed. When another process held the learn lock for the whole wait, the reason is not recorded; the outcome line still has it. `learn show` marks automatically promoted lessons `(auto-promoted)`. History records each one as an `auto-promote` entry with the lesson ids, removed ids, and the number of signals.
156
158
157
159
The state behind this is `.altimate-code/learn/<name>/auto-promote.json`. If it cannot be read, automatic promotion stops and `learn status` says so. Deleting the file restarts it, but it also resets the daily limit and makes every lesson count as person-approved.
158
160
@@ -216,7 +218,7 @@ altimate-code learn disable
216
218
|---|---|
217
219
|`--auto-promote`| Also turn on [fully automatic mode](#fully-automatic-mode). Without the flag, `enable` leaves an existing `learn.auto_promote` setting as it is. |
218
220
219
-
`disable` stops capture, automatic reflection, and automatic promotion. It does not delete anything, and approved lessons keep being delivered.
221
+
`disable` stops capture, automatic reflection, and automatic promotion. It does not delete anything, and approved lessons keep being delivered. Environment variables (`ALTIMATE_LEARN_CAPTURE`, `ALTIMATE_LEARN_AUTO`, `ALTIMATE_LEARN_AUTO_PROMOTE`) override the written config; `ALTIMATE_LEARN=0` turns all learning off.
0 commit comments