aboutsummaryrefslogtreecommitdiff
path: root/help.html
diff options
context:
space:
mode:
Diffstat (limited to 'help.html')
-rw-r--r--help.html10
1 files changed, 10 insertions, 0 deletions
diff --git a/help.html b/help.html
index 18a03ef..9a9a363 100644
--- a/help.html
+++ b/help.html
@@ -1660,6 +1660,16 @@ listmods</pre>
<h2>📋 Changelog</h2>
<div class="version-item">
+ <div class="v-header"><span class="v-num">v4.6.56 — October 2026</span><span class="v-date">2026-10-23</span></div>
+ <ul>
+ <li><strong>Predictor accuracy was structurally 0%</strong> — the denominator counted every observed run while the numerator only moved when a <em>stored prediction</em> matched, and predictions were stored only by the manual <code>predict</code> command. Unpaired runs no longer enter the accuracy denominator; the AI loop now registers a prediction per turn (<code>notePrediction</code>) and <code>predict_task</code> returns a tracked <code>predictionId</code> instead of discarding its estimate.</li>
+ <li><strong>Paired scoring</strong> — a prediction is consumed by threaded id → task hash → cli-only feature similarity ≥ 0.5, within a 15-minute TTL (<code>predictor-config pair-ttl &lt;seconds&gt;</code>). A run counts as correct within 30% <em>or</em> 3 seconds of the predicted duration; <code>predictor-status</code> shows paired accuracy, mean error, turns error, pending predictions and unpaired runs.</li>
+ <li><strong>Schema v2</strong> — stored payloads from v1 are migrated on load: observed runs and the prediction counter are kept, the meaningless <code>correct/total</code> counters are dropped. Task text is stored as an FNV-1a hash, never in plaintext. Tuning (<code>confidence-threshold</code>, <code>min-data-points</code>, <code>pair-ttl</code>) now persists.</li>
+ <li><strong>Knobs that do something</strong> — <code>min-data-points</code> now scales prediction confidence (it was declared, displayed and settable but never read), durations exclude chat-lock queue wait, and the confidence percentages on <code>predict</code>/<code>predict_task</code> will move for the same history (fewer points ⇒ lower confidence).</li>
+ </ul>
+ </div>
+
+ <div class="version-item">
<div class="v-header"><span class="v-num">v4.6.52 — October 2026</span><span class="v-date">2026-10-22</span></div>
<ul>
<li><strong>Documentation backlog cleared</strong> — 74 commands that the code registered but this page never documented are now in their own sections: <code>dashboard</code> (with <code>--watch</code>/<code>--stop</code>), <code>usage</code>, <code>ps</code>/<code>top</code>/<code>df</code>/<code>free</code>, the six <code>a2a-*</code> mailbox commands, <code>webxai</code>/<code>webx</code>/<code>wx</code>, <code>context</code>/<code>contexts</code>/<code>context-new</code>/<code>context-delete</code>/<code>context-auto</code>/<code>smart-ask</code>, the <code>commit-*</code> approval trio, <code>module-load|unload|enable|reset</code>, <code>memgraph-agent|session|decay-config</code>, <code>orchestrator-schedule|schedules|unschedule</code>, <code>skill-validate|audit|restore|pack-core</code>, <code>learn-skill</code>, <code>encrypt</code>/<code>decrypt</code>, <code>gzip</code>/<code>wget</code>, <code>kanban</code>, <code>three</code>, <code>course</code>, <code>videos</code>, <code>speak</code>/<code>voice</code>/<code>analyze</code>, <code>gui-run|text|clear</code>, the <code>vaultmind-*</code> set, <code>debug-sync</code>/<code>debug-sync-url</code>, <code>setfont</code>, <code>bug</code>, <code>session-new</code>, <code>files</code>, <code>node</code>, <code>sed</code>, <code>nvfs-index-graph</code>, <code>wallet-lookup</code>, <code>ai-model</code>, <code>ai-ignore</code>, <code>webdav-login</code>, <code>view_module</code>, <code>run</code>, <code>sysmsg</code>, <code>history</code>.</li>