diff options
Diffstat (limited to 'help.html')
| -rw-r--r-- | help.html | 34 |
1 files changed, 31 insertions, 3 deletions
@@ -537,6 +537,17 @@ diff a.txt b.txt</pre> <tr><td><code>reset-mem</code></td><td>Clear AI internal logs without touching VFS/config</td><td><code>reset-mem</code></td></tr> </table></div> + <h3>Debug Supervisor</h3> + <div class="table-wrap"><table> + <tr><th>Command</th><th>Description</th><th>Example</th></tr> + <tr><td><code>debug-stats</code></td><td>Error summary (total, unique, clusters, top issues)</td><td><code>debug-stats</code></td></tr> + <tr><td><code>debug-history</code></td><td>Show last N errors with hash, type, severity</td><td><code>debug-history 10</code></td></tr> + <tr><td><code>debug-analyze</code></td><td>Deep analysis of a specific error (root cause, fix suggestions)</td><td><code>debug-analyze a1b2c3d4</code></td></tr> + <tr><td><code>debug-trends</code></td><td>Error frequency over time, grouped by day and type</td><td><code>debug-trends 7</code></td></tr> + <tr><td><code>debug-fix</code></td><td>Suggest and optionally auto-apply fix for an error</td><td><code>debug-fix a1b2c3d4</code></td></tr> + <tr><td><code>debug-health</code></td><td>Check DebugSupervisorModule health status</td><td><code>debug-health</code></td></tr> + </table></div> + <h3>Skills</h3> @@ -1076,7 +1087,7 @@ cat /sys/modules/calc/info cat /sys/dom/node-count cat /sys/config/ai cat /sys/config/version</pre> - <p>Also available: <code>mem-debug</code> (one-shot text output) and <code>reset-mem</code> (clear AI internal logs without touching VFS or config).</p> + <p>Also available: <code>mem-debug</code> (one-shot text output), <code>reset-mem</code> (clear AI internal logs without touching VFS or config), and the <strong>DebugSupervisorModule</strong> (<code>debug-stats</code>, <code>debug-history</code>, <code>debug-analyze</code>, <code>debug-trends</code>, <code>debug-fix</code>, <code>debug-health</code>) for persistent error tracking and root-cause analysis.</p> </section> <!-- ── Shell ── --> @@ -1564,6 +1575,23 @@ listmods</pre> <h2>📋 Changelog</h2> <div class="version-item"> + <div class="v-header"><span class="v-num">v4.5.86 — July 2026</span><span class="v-date">2026-07-04</span></div> + <ul> + <li><strong>Cold module activation</strong> — Three modules (Memory Graph, Predictor, World Model) were loaded but produced zero actionable data. All now wired into the AI agent loop.</li> + <li><strong>Predictor feedback loop</strong> — <code>predictor.recordResult()</code> now called after multi-turn AI tasks (toolTurns ≥ 2). Predictor accumulates historical data for duration/turn predictions. Previously had zero call sites — module was dead.</li> + <li><strong>New AI tools</strong> — <code>predict_task</code> (estimate task duration before starting) and <code>world_model_query</code> (query system self-awareness: modules, skills, files, agents, health, changes, projects). AI now has 51 tools (was 49).</li> + <li><strong>Memory graph noise suppression</strong> — <code>getContextForPrompt()</code> returns empty string when graph has ≤2 nodes and 0 edges. Eliminates useless <code>[Memory Graph Context] 1 nodes, 0 edges</code> token waste on every AI turn.</li> + <li><strong>Memory graph task recording</strong> — <code>recordTask()</code> now called for direct AI conversations, not just orchestrator tasks. Graph learns about all AI interactions.</li> + <li><strong><code>memgraph-learn</code> command</strong> — Documented in help.html but never implemented. Now creates concept+fact nodes with weighted edge. <code>memgraph-learn <entity> <fact></code>.</li> + <li><strong><code>getStats()</code> alias</strong> — Orchestrator's <code>_pipelineVerify()</code> called <code>memgraph.getStats()</code> but the method was <code>showStats()</code>. Added <code>getStats()</code> alias returning <code>{nodes, edges, types, totalAccesses}</code>.</li> + <li><strong>World Model summary fix</strong> — <code>_getTotalModules()</code> returned registry count (7) instead of <code>max(loaded, registry)</code> (43). Summary now shows <code>43/43</code> not <code>43/7</code>.</li> + <li><strong>Dashboard World Model panel</strong> — New section showing version, last synthesized time, module/skill/file/agent counts, memory graph stats, health, active projects, recent changes.</li> + <li><strong>System prompt update</strong> — Now mentions <code>predict_task</code> and <code>world_model_query</code> capabilities.</li> + <li>Cache: <code>2026-07-04-v4.5.86</code> — hard refresh required</li> + </ul> + </div> + + <div class="version-item"> <div class="v-header"><span class="v-num">v4.5.64–v4.5.67 — June 2026</span><span class="v-date">2026-06-18</span></div> <ul> <li><strong>Orphan removal</strong> — Removed 5 dead modules from core-bundle: <code>MyModule</code>, <code>ModuleIOModule</code>, <code>DynamicPromptsModule</code>, <code>SmartContextModule</code>, <code>DuckDuckGoModule</code>. 454 lines, 23KB saved. 12+ ghost commands cleaned from help.html. <code>modules-index.json</code> 50→45 entries. <code>rebuild-bundles.py</code> updated to prevent resurrection.</li> @@ -1785,10 +1813,10 @@ listmods</pre> <div class="v-header"><span class="v-num">v4.5.12 — June 2026</span><span class="v-date">2026-06-06</span></div> <ul> <li><strong>Tiered Indexing toggle</strong> — <code>nvfs-config deep-index on|off</code> switches between Fast (metadata-only, instant) and Deep (LLM-assisted, semantic) modes. Off by default. <code>nvfs-index --deep</code> or <code>nvfs-sync --deep</code> forces deep mode per-run.</li> - <li><strong>Configurable deep model</strong> — <code>nvfs-config deep-model <name></code> selects the LLM for deep indexing. Default: <code>ministral-3:3b-cloud</code> for speed. Example: <code>nvfs-config deep-model gemma4:31b-cloud</code> for deeper analysis.</li> + <li><strong>Configurable deep model</strong> — <code>nvfs-config deep-model <name></code> selects the LLM for deep indexing. Default: <code>gemma4:31b-cloud</code> for speed. Example: <code>nvfs-config deep-model gemma4:31b-cloud</code> for deeper analysis.</li> <li><strong>Status bar indicator</strong> — Real-time 🧠 status in terminal bar: concept count + 🌊 wave for deep mode, 🟡 amber for pending auto-sync, 🧠 Deep Indexing progress percentage during LLM processing.</li> <li><strong>nvfs-sync command</strong> — <code>nvfs-sync [--deep]</code> for incremental background sync of dirty files.</li> - <li><strong>Deep concept extraction</strong> — LLM generates 2-5 semantic concepts per file (configurable model via <code>nvfs-config deep-model</code>, default <code>ministral-3:3b-cloud</code>, 2000 chars, 3-file batches with UI yield). Fast filename-concepts stored immediately, LLM concepts augment in background queue.</li> + <li><strong>Deep concept extraction</strong> — LLM generates 2-5 semantic concepts per file (configurable model via <code>nvfs-config deep-model</code>, default <code>gemma4:31b-cloud</code>, 2000 chars, 3-file batches with UI yield). Fast filename-concepts stored immediately, LLM concepts augment in background queue.</li> <li><strong>nvfs-status mode display</strong> — Shows 🌊 Deep or ⚡ Fast mode.</li> <li><strong>LLM self-healing</strong> — If Ollama disconnected, deep mode silently falls back to fast metadata-only.</li> <li>Cache: <code>2026-06-06-v4.5.12</code> — hard refresh required</li> |