monomind 2.7.12 → 2.7.13

package/packages/@monomind/cli/.claude/helpers/handlers/route-handler.cjs

@@ -1,770 +0,0 @@
-'use strict';
-// Extracted from hook-handler.cjs — receives hCtx from dispatcher.
-// Behavioral equivalence verified: 133 routing tests pass post-extraction.
-// hCtx (hook context) contains all shared state and utility functions:
-//   hCtx.hookInput, hCtx.toolInput, hCtx.toolName, hCtx.prompt, hCtx.args, hCtx.CWD
-//   hCtx.session, hCtx.router, hCtx.intelligence
-//   hCtx.isSimpleCommand — function defined in main(), passed via hCtx
-//   hCtx.getLearningService — async factory for LearningService singleton
-//   Utility fns: _recordRecentEdit, _findAffectedTests, _recordHookLatency,
-//     _getBudgetStatus, _injectCompactGraphMap, _maybeRebuildMonograph,
-//     _buildKnowledgeSearchFn, getMonographSuggestions, getMonographNeighbors,
-//     runWithTimeout, safeRequire, scanMicroAgentTriggers, _recordGraphTelemetry,
-//     _recordDecisionMarkers, _recordToolCall, _openMonographDb, fs, path
-//
-// NOTE: The 'route' handler has a local variable named 'ctx' (from intelligence.getContext).
-// The dispatcher passes the hook context as 'hCtx' to avoid collision.
-
-const path = require('path');
-const fs = require('fs');
-
-// ── Intelligence read-path bridge ───────────────────────────────────────────
-// route-handler.cjs runs as a fresh node subprocess per invocation, but a
-// single process may call handle() more than once (e.g. tests, or a daemon
-// wrapper) — cache the dynamic import of the CLI's compiled intelligence
-// bridge (hooks-embedding.js -> suggestAgentsFromIntelligence) so repeat
-// calls within the same process don't re-resolve/re-import the module.
-// suggestAgentsFromIntelligence() reads the SONA/ReasoningBank embedding
-// store populated by recordMemoryDecision() on every post-task — this is
-// the module that makes stored embedding patterns affect live routing.
-var _intelligenceModPromise = null;
-function _loadIntelligenceModule(CWD) {
-  if (!_intelligenceModPromise) {
-    _intelligenceModPromise = (async function() {
-      var candidates = [
-        path.resolve(CWD, 'packages', '@monomind', 'cli', 'dist', 'src', 'mcp-tools', 'hooks-embedding.js'),
-        path.resolve(CWD, 'packages', '@monomind', 'cli', 'node_modules', '@monomind', 'cli', 'dist', 'src', 'mcp-tools', 'hooks-embedding.js'),
-      ];
-      for (var i = 0; i < candidates.length; i++) {
-        if (fs.existsSync(candidates[i])) {
-          try {
-            return await import(candidates[i]);
-          } catch (e) {
-            return null;
-          }
-        }
-      }
-      return null;
-    })();
-  }
-  return _intelligenceModPromise;
-}
-
-module.exports = {
-  handle: async function(hCtx) {
-    var prompt = hCtx.prompt;
-    var hookInput = hCtx.hookInput;
-    var router = hCtx.router;
-    var intelligence = hCtx.intelligence;
-    var CWD = hCtx.CWD;
-
-    // For slash commands and single-action invocations: skip routing panel output
-    // but still write last-route.json so the statusline reflects the current action.
-    if (hCtx.isSimpleCommand(prompt)) {
-      try {
-        var cmdLabel = (typeof prompt === 'string' && prompt.trim().startsWith('/'))
-          ? prompt.trim().split(/\s+/)[0]          // e.g. "/ts"
-          : (hookInput.commandName || hookInput.command_name || 'command');
-        var routeDir = path.join(CWD, '.monomind');
-        fs.mkdirSync(routeDir, { recursive: true });
-        fs.writeFileSync(
-          path.join(routeDir, 'last-route.json'),
-          JSON.stringify({
-            agent: cmdLabel,
-            confidence: 1.0,
-            reason: 'predefined command — no routing needed',
-            semanticRouting: false,
-            updatedAt: new Date().toISOString(),
-          }),
-          'utf-8'
-        );
-      } catch (e) { /* non-fatal */ }
-      return;
-    }
-
-    if (intelligence && intelligence.getContext) {
-      try {
-        // Each hook event runs as a fresh node process, so the module-level
-        // _entries cache is always empty here — without init() getContext()
-        // returns null on every prompt and stored patterns are never recalled.
-        // init() reads one small JSON file (auto-memory-store.json), so the
-        // per-prompt cost is negligible.
-        if (intelligence.init) {
-          try { intelligence.init(); } catch (e) { /* non-fatal */ }
-        }
-        // Bootstrap intelligence from monograph on first prompt if store is sparse
-        if (intelligence.bootstrapFromDb) {
-          try {
-            var bDb = hCtx._openMonographDb();
-            if (bDb) {
-              var bootstrapped = intelligence.bootstrapFromDb(bDb);
-              if (bootstrapped > 0) console.log('[INTELLIGENCE] Bootstrapped ' + bootstrapped + ' hub nodes from knowledge graph');
-            }
-          } catch (e) { /* non-fatal */ }
-        }
-        const ctx = intelligence.getContext(prompt);
-        if (ctx) console.log(ctx);
-      } catch (e) { /* non-fatal */ }
-    }
-    if (router && (router.routeTaskSemantic || router.routeTask)) {
-      const routeFn = router.routeTaskSemantic || router.routeTask;
-      var result = await Promise.resolve(routeFn(prompt));
-
-      // ── Enrichment: when router.cjs falls to the broad "coder" catch-all,
-      //    try @monoes/routing's richer keyword pre-filter (30+ specialized
-      //    rules for Solidity, game engines, DevOps, embedded, etc.) for a
-      //    more specific agent match. This bridges the hooks layer (router.cjs)
-      //    with the CLI routing package without requiring ESM imports. ─────────
-      if (result && result.agentSlug === 'coder' && result.confidence <= 0.8) {
-        try {
-          var routingPkgPath = path.resolve(
-            CWD, 'packages', '@monomind', 'routing', 'dist', 'keyword-pre-filter.js'
-          );
-          // Also check CLI's node_modules (symlinked workspace)
-          if (!fs.existsSync(routingPkgPath)) {
-            routingPkgPath = path.resolve(
-              CWD, 'packages', '@monomind', 'cli', 'node_modules',
-              '@monomind', 'routing', 'dist', 'keyword-pre-filter.js'
-            );
-          }
-          if (fs.existsSync(routingPkgPath)) {
-            var routingMod = await import(routingPkgPath);
-            var enrichRules = routingMod.DEFAULT_KEYWORD_ROUTES;
-            if (enrichRules && enrichRules.length > 0) {
-              for (var eri = 0; eri < enrichRules.length; eri++) {
-                var erRule = enrichRules[eri];
-                if (erRule.pattern && erRule.pattern.test(prompt)) {
-                  result.agent = erRule.routeName || erRule.agentSlug;
-                  result.agentSlug = erRule.agentSlug;
-                  result.confidence = erRule.score != null ? Math.min(0.98, Math.max(0.70, erRule.score)) : 0.85;
-                  result.reason = 'Enriched: ' + (erRule.description || erRule.routeName);
-                  result.enrichedFrom = 'routing-keyword-pre-filter';
-                  break;
-                }
-              }
-            }
-          }
-        } catch (e) { /* non-fatal — routing package may not be available */ }
-      }
-
-      // ── Intelligence embedding suggestion (SONA/ReasoningBank read path) ──
-      // Wires the previously-write-only intelligence system into live routing:
-      // suggestAgentsFromIntelligence() runs an embedding similarity search
-      // over patterns stored by recordMemoryDecision() on every post-task.
-      // This ENHANCES the keyword route — it only overrides when the
-      // embedding match is meaningfully more confident, and never blocks or
-      // delays routing beyond a 2s budget (fails silently otherwise).
-      try {
-        var intelResult = await Promise.race([
-          (async function() {
-            var mod = await _loadIntelligenceModule(CWD);
-            if (!mod || !mod.suggestAgentsFromIntelligence) return null;
-            return await mod.suggestAgentsFromIntelligence(prompt);
-          })(),
-          new Promise(function(resolve) { setTimeout(function() { resolve(null); }, 2000); })
-        ]);
-        if (intelResult && intelResult.agents && intelResult.agents.length > 0) {
-          var topIntelAgent = intelResult.agents[0];
-          var intelConf = intelResult.confidence != null ? intelResult.confidence : 0;
-          result.intelligenceSuggestion = { agents: intelResult.agents, confidence: intelConf };
-
-          var curAgent = result.agentSlug || result.agent;
-          var curConf = result.confidence != null ? result.confidence : 0;
-          if (topIntelAgent !== curAgent) {
-            // Only override when the embedding match clearly beats the keyword
-            // route (0.1 margin) — otherwise just surface it as a signal.
-            if (intelConf > curConf + 0.1) {
-              console.log('[INTELLIGENCE] Embedding suggestion overrides keyword routing: ' + topIntelAgent + ' (confidence ' + intelConf.toFixed(2) + ') vs keyword ' + curAgent + ' (' + curConf.toFixed(2) + ')');
-              result.keywordAgent = curAgent;
-              result.agent = topIntelAgent;
-              result.agentSlug = topIntelAgent;
-              result.confidence = intelConf;
-              result.reason = 'Intelligence embedding match (SONA/ReasoningBank)' + (result.reason ? '; keyword route was: ' + result.reason : '');
-              result.enrichedFrom = 'intelligence-embedding';
-            } else {
-              console.log('[INTELLIGENCE] Embedding suggestion: ' + topIntelAgent + ' (confidence ' + intelConf.toFixed(2) + ') — kept keyword route ' + curAgent);
-            }
-          }
-        }
-      } catch (e) { /* non-fatal — intelligence system unavailable or timed out */ }
-
-      // ── Agent success pattern lookup ──────────────────────────
-      try {
-        var patternDb = hCtx._openMonographDb();
-        if (patternDb) {
-          var cutoff = Date.now() - 30 * 86400000;
-          var patterns = patternDb.prepare(
-            'SELECT agent_type, COUNT(*) as total, SUM(CASE WHEN success=1 THEN 1 ELSE 0 END) as wins FROM agent_interactions WHERE timestamp > @cutoff GROUP BY agent_type HAVING total >= 3'
-          ).all({ cutoff: cutoff });
-          if (patterns.length > 0) {
-            var patternMap = {};
-            for (var pi = 0; pi < patterns.length; pi++) {
-              patternMap[patterns[pi].agent_type] = {
-                total: patterns[pi].total,
-                successRate: Math.round(patterns[pi].wins / patterns[pi].total * 100)
-              };
-            }
-            var routedAgent = result.agent || 'coder';
-            var routedPattern = patternMap[routedAgent];
-            // Find if there's a clearly better agent
-            if (routedPattern && routedPattern.successRate < 50) {
-              var best = patterns.reduce(function(a, b) { return (b.wins/b.total) > (a.wins/a.total) ? b : a; });
-              if (best.wins / best.total > 0.8 && best.agent_type !== routedAgent) {
-                console.log('[AGENT_PATTERN] Historical: ' + best.agent_type + ' (' + Math.round(best.wins/best.total*100) + '% success, n=' + best.total + ') outperforms ' + routedAgent + ' (' + routedPattern.successRate + '%)');
-              }
-            }
-            result.historicalPattern = patternMap;
-          }
-        }
-      } catch (e) { /* non-fatal — agent_interactions table may not exist */ }
-
-      // ── Routing feedback accuracy check ──────────────────────
-      try {
-        var rfPath = path.join(CWD, '.monomind', 'routing-feedback.jsonl');
-        var MAX_RF = 256 * 1024;
-        if (fs.existsSync(rfPath) && fs.statSync(rfPath).size <= MAX_RF) {
-          var rfLines = fs.readFileSync(rfPath, 'utf-8').trim().split('\n').filter(Boolean);
-          // Check last 50 entries for this agent's feedback
-          var routedAgent = result.agent || 'coder';
-          var recentRf = rfLines.slice(-50);
-          var agentFeedback = [];
-          for (var ri = 0; ri < recentRf.length; ri++) {
-            try {
-              var rfEntry = JSON.parse(recentRf[ri]);
-              if (rfEntry.suggestedAgent === routedAgent && typeof rfEntry.intelligenceFeedback === 'boolean') {
-                agentFeedback.push(rfEntry.intelligenceFeedback);
-              }
-            } catch (e) {}
-          }
-          if (agentFeedback.length >= 3) {
-            var successes = agentFeedback.filter(function(f) { return f === true; }).length;
-            var accuracy = Math.round(successes / agentFeedback.length * 100);
-            if (accuracy < 60) {
-              console.log('[ROUTING] Warning: ' + routedAgent + ' routing accuracy ' + accuracy + '% over last ' + agentFeedback.length + ' sessions');
-            }
-            result.routingAccuracy = accuracy;
-          }
-        }
-      } catch (e) { /* non-fatal */ }
-
-      // monolean: graph-fallback override removed — it compensated for bad agent
-      // recommendations that are no longer injected into context. The statusline
-      // just shows the keyword router's pick; no need for 50 lines of overrides.
-
-      // ── Dispatch dedup: suppress re-recommending the same agent just dispatched ──
-      // agent-start-handler writes last-dispatch.json on SubagentStart.
-      // If the router picks the same agent within 60s, it's likely the parent re-routing
-      // the same prompt — log a note so the LLM can vary its approach.
-      try {
-        var dispatchPath = path.join(CWD, '.monomind', 'last-dispatch.json');
-        var MAX_DISPATCH = 4096;
-        if (fs.existsSync(dispatchPath) && fs.statSync(dispatchPath).size <= MAX_DISPATCH) {
-          var lastDispatch = JSON.parse(fs.readFileSync(dispatchPath, 'utf-8'));
-          var dispatchAge = Date.now() - new Date(lastDispatch.dispatchedAt || 0).getTime();
-          if (dispatchAge < 60000 && lastDispatch.agentType === (result.agentSlug || result.agent)) {
-            result.recentlyDispatched = true;
-            console.log('[DISPATCH_DEDUP] ' + lastDispatch.agentType + ' was dispatched ' + Math.round(dispatchAge / 1000) + 's ago — consider a different specialist or direct implementation');
-          }
-        }
-      } catch (e) { /* non-fatal */ }
-
-      var output = [];
-      var conf = result.confidence != null ? result.confidence : 0;
-
-      // ── Persist routing result for statusline display ─────────────
-      try {
-        var routeDir = path.join(CWD, '.monomind');
-        fs.mkdirSync(routeDir, { recursive: true });
-        fs.writeFileSync(
-          path.join(routeDir, 'last-route.json'),
-          JSON.stringify({
-            agent: result.agent || 'coder',
-            agentSlug: result.agentSlug || null,
-            confidence: result.confidence,
-            reason: result.reason,
-            prompt: (prompt || '').slice(0, 500),
-            historicalPattern: result.historicalPattern || null,
-            updatedAt: new Date().toISOString(),
-          }),
-          'utf-8'
-        );
-      } catch (e) { /* non-fatal */ }
-
-      // ── Skill auto-activation (the one thing the hook does better than Claude) ──
-      var matches = result.skillMatches || [];
-      if (matches.length > 0) {
-        var topMatch = matches[0];
-        // Auto-activate when one skill clearly dominates:
-        //   - score >= 2 with at most 2 matches (strong multi-keyword signal), or
-        //   - single match with score >= 2 and low agent confidence (skill is better fit)
-        var autoInvoke = (topMatch && topMatch.score >= 2 && matches.length <= 2) ||
-          (topMatch && topMatch.score >= 2 && matches.length === 1 && conf < 0.7);
-
-        if (autoInvoke) {
-          output.push('+======== SKILL AUTO-ACTIVATED ========+');
-          output.push('| ' + topMatch.invoke.padEnd(36) + ' |');
-          output.push('+======================================+');
-        } else {
-          output.push('+----------- Matching Skills (invoke via Skill tool) ----------+');
-          matches.forEach(function(m, i) {
-            output.push('| ' + (i + 1) + '. ' + m.skill.padEnd(28) + (m.description || '').substring(0, 30).padEnd(30) + ' |');
-            output.push('|   invoke: ' + m.invoke.substring(0, 51).padEnd(51) + '|');
-          });
-          output.push('+--------------------------------------------------------------+');
-        }
-      }
-
-      if (output.length > 0) console.log(output.join('\n'));
-
-      // ── Second Brain: per-request knowledge injection ──────────────────
-      // When the project has an indexed knowledge base, surface the most
-      // relevant excerpts for THIS prompt so Claude has them without having
-      // to decide to call knowledge_search. Semantic-first via the dashboard
-      // server's warm /api/knowledge/search (it holds the local embedding
-      // model hot — a hook subprocess can't afford a 1-3s model load per
-      // prompt); tokenized keyword scoring over chunks.jsonl as the fallback
-      // when the server is down or still warming.
-      try {
-        var sbPrompt = String(prompt || '');
-        // Skip slash commands and low-content prompts — injection would be
-        // noise. A prompt earns injection only when it carries at least two
-        // substantive terms ("what is next" / "ok go ahead" / "thanks" carry
-        // zero-to-one and must stay quiet; a real question about the project
-        // clears this easily).
-        var _SB_FILLER = new Set(['what', 'whats', 'is', 'are', 'was', 'were', 'the', 'this', 'that', 'these', 'those',
-          'next', 'now', 'then', 'and', 'but', 'for', 'not', 'you', 'your', 'can', 'could', 'should', 'would', 'will',
-          'lets', 'let', 'make', 'made', 'making', 'please', 'okay', 'yes', 'yeah', 'sure', 'thanks', 'thank',
-          'ahead', 'continue', 'proceed', 'more', 'again', 'how', 'why', 'when', 'where', 'who', 'which',
-          'with', 'about', 'from', 'into', 'onto', 'over', 'under', 'all', 'any', 'some', 'one', 'two', 'just',
-          'like', 'want', 'need', 'get', 'got', 'here', 'there', 'still', 'also', 'too', 'very', 'really',
-          'them', 'they', 'their', 'theirs', 'its', 'our', 'ours', 'him', 'her', 'his', 'hers',
-          'order', 'done', 'doing', 'did', 'does', 'goes', 'going', 'went', 'come', 'came', 'good', 'great', 'nice', 'fine']);
-        var _sbSubstantive = sbPrompt.toLowerCase().split(/[^a-z0-9]+/).filter(function(t) {
-          return t.length >= 3 && !_SB_FILLER.has(t);
-        });
-        if (_sbSubstantive.length >= 2 && sbPrompt.charAt(0) !== '/') {
-          var sbKnowledgeDir = path.join(CWD, '.monomind', 'knowledge');
-          if (fs.existsSync(path.join(sbKnowledgeDir, 'chunks.jsonl'))) {
-            var sbHits = null;
-            var sbMethod = 'keyword';
-            // Which corpus answered. The two paths are NOT the same corpus and
-            // deliberately so:
-            //   'second-brain'         — the full ingested document store
-            //                            (knowledge:shared in SQLite, superseded
-            //                            versions filtered out server-side).
-            //   'project-instructions' — the keyword fallback over
-            //                            .monomind/knowledge/chunks.jsonl, which
-            //                            _autoIndexKnowledge builds from CLAUDE.md,
-            //                            CLAUDE.local.md, docs/todo.md and monograph
-            //                            god-nodes ONLY. A hook subprocess cannot
-            //                            afford the 1-3s embedding-model load the
-            //                            real store needs, so when the warm server
-            //                            is down this narrower corpus is all that is
-            //                            reachable. It is labelled in the banner
-            //                            rather than passed off as the Second Brain.
-            var sbCorpus = 'project-instructions';
-
-            // Semantic path: warm control-server endpoint (strictly local traffic;
-            // sbAuth is the local dashboard session value from .monomind, same one
-            // every other hook event POST attaches — not a checked-in secret).
-            try {
-              var sbCtrlUrl = 'http://localhost:4242';
-              try {
-                var sbCtl = JSON.parse(fs.readFileSync(path.join(CWD, '.monomind', 'control.json'), 'utf-8'));
-                if (sbCtl.url) sbCtrlUrl = sbCtl.url;
-              } catch (_) {}
-              var sbAuth = '';
-              try { sbAuth = fs.readFileSync(path.join(CWD, '.monomind', 'dashboard-token'), 'utf-8').trim(); } catch (_) {}
-              var sbResp = await fetch(sbCtrlUrl + '/api/knowledge/search', {
-                method: 'POST',
-                headers: { 'Content-Type': 'application/json', 'x-monomind-token': sbAuth },
-                body: JSON.stringify({ query: sbPrompt, namespace: 'knowledge:shared', limit: 3 }),
-                signal: AbortSignal.timeout(900),
-              });
-              if (sbResp.ok) {
-                var sbData = await sbResp.json();
-                if (sbData && Array.isArray(sbData.results) && sbData.results.length > 0) {
-                  sbHits = sbData.results.map(function(r) {
-                    var src = (r.tags || []).find(function(t) { return t.indexOf('src:') === 0; });
-                    return { key: r.key, value: r.content, score: r.score, global: !!r.global, metadata: src ? { filePath: src.slice(4) } : {} };
-                  });
-                  sbMethod = sbData.method === 'semantic' ? 'semantic' : 'keyword';
-                  sbCorpus = 'second-brain';
-                }
-              }
-            } catch (_) { /* server down or warming — fall back below */ }
-
-            if (!sbHits) {
-              var sbSearch = hCtx._buildKnowledgeSearchFn(sbKnowledgeDir);
-              sbHits = await hCtx.runWithTimeout(
-                function() { return sbSearch(sbPrompt, { namespace: 'knowledge:shared', limit: 3, minScore: 0.45 }); },
-                'second-brain-inject'
-              );
-            }
-            // Relevance floor: injecting weak matches pollutes every prompt's
-            // context — configurable via .monomind/second-brain.json, default 0.35.
-            var sbFloor = 0.35;
-            var sbInjectionLimit = 5;
-            try {
-              var sbConf = JSON.parse(fs.readFileSync(path.join(CWD, '.monomind', 'second-brain.json'), 'utf-8'));
-              if (typeof sbConf.relevanceFloor === 'number') sbFloor = sbConf.relevanceFloor;
-              if (typeof sbConf.injectionLimit === 'number' && sbConf.injectionLimit > 0) sbInjectionLimit = sbConf.injectionLimit;
-            } catch (_) {}
-            if (sbHits) sbHits = sbHits.filter(function(h) { return (h.score || 0) >= sbFloor; });
-            if (sbHits && sbHits.length > sbInjectionLimit) sbHits = sbHits.slice(0, sbInjectionLimit);
-
-            // Telemetry: append one JSONL line per evaluated prompt so the
-            // thresholds above can be tuned from real usage (and misses can
-            // seed the golden-set eval). Prompt text is NOT logged.
-            try {
-              var sbMetricsDir = path.join(CWD, '.monomind', 'metrics');
-              fs.mkdirSync(sbMetricsDir, { recursive: true });
-              var sbMetricsFile = path.join(sbMetricsDir, 'second-brain.jsonl');
-              try {
-                if (fs.existsSync(sbMetricsFile) && fs.statSync(sbMetricsFile).size > 512 * 1024) {
-                  var sbOld = fs.readFileSync(sbMetricsFile, 'utf-8').trim().split('\n');
-                  fs.writeFileSync(sbMetricsFile, sbOld.slice(-500).join('\n') + '\n', 'utf-8');
-                }
-              } catch (_) {}
-              fs.appendFileSync(sbMetricsFile, JSON.stringify({
-                ts: Date.now(),
-                method: sbMethod,
-                hits: sbHits ? sbHits.length : 0,
-                topScore: sbHits && sbHits[0] ? Number((sbHits[0].score || 0).toFixed(3)) : null,
-                promptLen: sbPrompt.length,
-                terms: _sbSubstantive.length,
-                corpus: sbCorpus,
-                injected: !!(sbHits && sbHits.length > 0),
-              }) + '\n', 'utf-8');
-            } catch (_) { /* telemetry never blocks */ }
-
-            if (sbHits && sbHits.length > 0) {
-              var sbSource = sbCorpus === 'second-brain'
-                ? 'the Second Brain document store'
-                : 'project instructions only (CLAUDE.md/todo — Second Brain store unreachable)';
-              var sbLines = ['[SECOND_BRAIN] ' + sbHits.length + ' relevant excerpt(s) (' + sbMethod + ') from ' + sbSource + ':'];
-              for (var sbI = 0; sbI < sbHits.length; sbI++) {
-                var sbH = sbHits[sbI];
-                var sbSrc = (sbH.metadata && sbH.metadata.filePath) ? String(sbH.metadata.filePath).split('/').slice(-2).join('/') : sbH.key;
-                var sbText = String(sbH.value || '').replace(/\s+/g, ' ').slice(0, 240);
-                sbLines.push('  • [' + sbSrc + (sbH.global ? ' · global' : '') + '] ' + sbText);
-              }
-              sbLines.push('  (deeper lookup: mcp__monomind__knowledge_search or `monomind doc search -q "..."`)');
-              console.log(sbLines.join('\n'));
-            }
-          }
-        }
-      } catch (e) { /* non-fatal — knowledge injection must never block a prompt */ }
-
-      // Record any decision markers in this prompt (auto-ADR pipeline).
-      try { hCtx._recordDecisionMarkers(prompt); } catch (e) {}
-
-      // Cost budget — emit amber/red banner when approaching limit.
-      try {
-        var budget = hCtx._getBudgetStatus();
-        if (budget && budget.alert) {
-          var tunedNote = budget.autoTuned ? ' (auto-tuned)' : '';
-          if (budget.spike && !budget.breached) {
-            console.log('[BUDGET_SPIKE] Today $' + budget.todayCost.toFixed(2) + ' is >2x your rolling daily avg. Unusual spend — review .monomind/metrics/token-summary.json.');
-          } else if (budget.breached) {
-            console.log('[BUDGET_BREACHED] Daily $' + budget.todayCost.toFixed(2) + '/$' + budget.dailyLimit + ' (' + budget.dailyPct + '%) · Monthly $' + budget.monthCost.toFixed(2) + '/$' + budget.monthlyLimit + ' (' + budget.monthlyPct + '%)' + tunedNote + '. Switch to Haiku with /model haiku or edit .monomind/budget.json.');
-          } else {
-            console.log('[BUDGET_ALERT] Daily ' + budget.dailyPct + '% of $' + budget.dailyLimit + ' · Monthly ' + budget.monthlyPct + '% of $' + budget.monthlyLimit + tunedNote + '.');
-          }
-        }
-      } catch (e) {}
-
-      // ── Surface daemon metrics warnings (hook latency, token spikes) ──
-      try {
-        var metricsDir = path.join(CWD, '.monomind', 'metrics');
-        // Hook latency warnings
-        var latencyFile = path.join(metricsDir, 'hook-latency.json');
-        if (fs.existsSync(latencyFile) && fs.statSync(latencyFile).size < 32768) {
-          var latencyData = JSON.parse(fs.readFileSync(latencyFile, 'utf-8'));
-          if (latencyData && latencyData.avgMs > 500) {
-            console.log('[PERF] Hook latency avg ' + Math.round(latencyData.avgMs) + 'ms (target <500ms). Slowest: ' + (latencyData.slowest || 'unknown'));
-          }
-        }
-        // Token usage check — surface only when spend is high
-        var tokenFile = path.join(metricsDir, 'token-summary.json');
-        if (fs.existsSync(tokenFile) && fs.statSync(tokenFile).size < 32768) {
-          var tokenData = JSON.parse(fs.readFileSync(tokenFile, 'utf-8'));
-          if (tokenData && tokenData.todayCost > 50) {
-            console.log('[COST] Today: $' + (tokenData.todayCost || 0).toFixed(2) + ' (' + (tokenData.todayCalls || 0) + ' calls)');
-          }
-        }
-        // Security audit findings
-        var secAuditFile = path.join(metricsDir, 'security-audit.json');
-        if (fs.existsSync(secAuditFile) && fs.statSync(secAuditFile).size < 32768) {
-          var secData = JSON.parse(fs.readFileSync(secAuditFile, 'utf-8'));
-          if (secData && secData.findings && secData.findings.length > 0) {
-            var serious = secData.findings.filter(function (f) { return f && (f.severity === 'high' || f.severity === 'medium'); });
-            if (serious.length > 0) {
-              console.log('[SECURITY] ' + serious.length + ' finding(s) from background scan. Review .monomind/metrics/security-audit.json');
-            } else {
-              console.log('[AUDIT] ' + secData.findings.length + ' low-severity note(s) (architecture heuristics, not vulnerabilities) — .monomind/metrics/security-audit.json');
-            }
-          }
-        }
-        // Codebase map top files (high-centrality god nodes from monograph)
-        var mapFile = path.join(metricsDir, 'codebase-map.json');
-        if (fs.existsSync(mapFile) && fs.statSync(mapFile).size < 32768) {
-          var mapData = JSON.parse(fs.readFileSync(mapFile, 'utf-8'));
-          if (mapData && mapData.topFiles && mapData.topFiles.length > 0) {
-            console.log('[CODEBASE] ' + mapData.topFiles.length + ' high-centrality files. Top: ' + (mapData.topFiles[0].ref || 'unknown') + ' (degree ' + (mapData.topFiles[0].degree || '?') + ')');
-          }
-          if (mapData && mapData.graphStaleness && mapData.graphStaleness.commitsBehind > 10) {
-            console.log('[CODEBASE] Graph index ' + mapData.graphStaleness.commitsBehind + ' commits behind HEAD — run monograph build');
-          }
-        }
-        // Graph gate connectivity nudge — the pre-search/pre-bash gate
-        // (utils/monograph.cjs _graphGateShouldBlock) hard-blocks the first
-        // Grep/Glob/bash-grep-or-find call each session until a real
-        // monograph_query/monograph_suggest call fires. If that block was
-        // never followed by a real graph call, the monomind MCP server is
-        // most likely not connected this session (config present but
-        // unapproved/not started) — surface it once so the user can fix the
-        // actual cause instead of the gate silently degrading to a no-op.
-        var graphGateFile = path.join(CWD, '.monomind', 'graph-gate-state.json');
-        var mcpWarnFile = path.join(CWD, '.monomind', 'mcp-not-connected-warned.json');
-        if (fs.existsSync(graphGateFile) && fs.statSync(graphGateFile).size < 4096) {
-          var gateState = JSON.parse(fs.readFileSync(graphGateFile, 'utf-8'));
-          var gateSessId = String((hCtx.hookInput && (hCtx.hookInput.sessionId || hCtx.hookInput.session_id)) || '');
-          if (gateState && gateState.sessionId === gateSessId && gateState.blockedOnce && !gateState.queried) {
-            var alreadyWarnedMcp = false;
-            if (fs.existsSync(mcpWarnFile)) {
-              try {
-                var mcpWarnData = JSON.parse(fs.readFileSync(mcpWarnFile, 'utf-8'));
-                if (mcpWarnData && mcpWarnData.sessionId === gateSessId) alreadyWarnedMcp = true;
-              } catch (e) { /* corrupt — warn again to be safe */ }
-            }
-            if (!alreadyWarnedMcp) {
-              console.log('[MCP] The graph gate blocked a search but no monograph_query/monograph_suggest call followed — the monomind MCP server is likely not connected this session. Run `claude mcp add monomind -- npx monomind@latest mcp start` (then restart), or approve the .mcp.json trust prompt if one is pending.');
-              try {
-                fs.writeFileSync(mcpWarnFile, JSON.stringify({ sessionId: gateSessId, warnedAt: new Date().toISOString() }));
-              } catch (e) { /* non-fatal */ }
-            }
-          }
-        }
-        // Deep dive findings (god nodes, high-degree files from background analysis)
-        var deepdiveFile = path.join(metricsDir, 'deepdive.json');
-        if (fs.existsSync(deepdiveFile) && fs.statSync(deepdiveFile).size < 32768) {
-          var ddData = JSON.parse(fs.readFileSync(deepdiveFile, 'utf-8'));
-          if (ddData && ddData.findings && ddData.findings.length > 0) {
-            for (var di = 0; di < ddData.findings.length; di++) {
-              var finding = ddData.findings[di];
-              if (finding.category === 'god_nodes' && finding.items && finding.items.length > 0) {
-                var topGod = finding.items[0];
-                console.log('[DEEPDIVE] ' + finding.items.length + ' god nodes. Top: ' + (topGod.name || 'unknown') + ' (degree ' + (topGod.degree || '?') + ') in ' + (topGod.file || 'unknown'));
-              }
-            }
-          }
-        }
-        // Ultralearn insights (bridge nodes crossing community boundaries)
-        var ultralearnFile = path.join(metricsDir, 'ultralearn.json');
-        if (fs.existsSync(ultralearnFile) && fs.statSync(ultralearnFile).size < 32768) {
-          var ulData = JSON.parse(fs.readFileSync(ultralearnFile, 'utf-8'));
-          if (ulData && ulData.insightsGained && ulData.insightsGained.length > 0) {
-            for (var ui = 0; ui < ulData.insightsGained.length; ui++) {
-              var insight = ulData.insightsGained[ui];
-              if (insight.category === 'bridge_nodes' && insight.items && insight.items.length > 0) {
-                var topBridge = insight.items[0];
-                console.log('[ARCHITECTURE] ' + insight.items.length + ' bridge nodes crossing community boundaries. Top: ' + (topBridge.name || 'unknown') + ' (' + (topBridge.crossCommunityEdges || '?') + ' cross-edges) in ' + (topBridge.location || 'unknown'));
-              }
-            }
-          }
-        }
-        // Performance metrics (optimize worker — a one-shot process, not a daemon)
-        var perfFile = path.join(metricsDir, 'performance.json');
-        if (fs.existsSync(perfFile) && fs.statSync(perfFile).size < 32768) {
-          var perfData = JSON.parse(fs.readFileSync(perfFile, 'utf-8'));
-          if (perfData && perfData.workerProcessMemoryUsage) {
-            var rssBytes = perfData.workerProcessMemoryUsage.rss || 0;
-            var rssMB = Math.round(rssBytes / (1024 * 1024));
-            if (rssMB > 512) {
-              console.log('[PERF] optimize worker RSS ' + rssMB + 'MB (>512MB threshold) at last run — reflects that one-shot process, not a persistent daemon');
-            }
-          }
-        }
-        // Memory consolidation health
-        var consolFile = path.join(metricsDir, 'consolidation.json');
-        if (fs.existsSync(consolFile) && fs.statSync(consolFile).size < 32768) {
-          var consolData = JSON.parse(fs.readFileSync(consolFile, 'utf-8'));
-          if (consolData && consolData.patternsConsolidated > 0) {
-            console.log('[MEMORY] ' + consolData.patternsConsolidated + ' patterns consolidated into ' + consolData.clustersCreated + ' RAPTOR clusters');
-          }
-        }
-        // Benchmark worker RSS (manual-trigger `daemon trigger -w benchmark`)
-        var benchFile = path.join(metricsDir, 'benchmark.json');
-        if (fs.existsSync(benchFile) && fs.statSync(benchFile).size < 32768) {
-          var benchData = JSON.parse(fs.readFileSync(benchFile, 'utf-8'));
-          if (benchData && benchData.benchmarks && benchData.benchmarks.memoryUsage) {
-            var benchRssMB = Math.round((benchData.benchmarks.memoryUsage.rss || 0) / (1024 * 1024));
-            if (benchRssMB > 512) {
-              console.log('[PERF] Benchmark snapshot RSS ' + benchRssMB + 'MB (>512MB threshold) at last `daemon trigger -w benchmark` run');
-            }
-          }
-        }
-      } catch (e) { /* non-fatal */ }
-
-      // ── Memory: surface relevant past session context ──────────
-      try {
-        var epFile = path.join(CWD, '.monomind', 'episodic', 'episodes.jsonl');
-        var MAX_EP = 256 * 1024;
-        if (fs.existsSync(epFile) && fs.statSync(epFile).size <= MAX_EP) {
-          var epRaw = fs.readFileSync(epFile, 'utf-8').trim().split('\n').filter(Boolean);
-          if (epRaw.length > 0) {
-            // Simple keyword matching against episode summaries
-            var promptLower = (prompt || '').toLowerCase();
-            var promptTokens = promptLower.match(/[a-z][a-z0-9_-]{2,}/g) || [];
-            if (promptTokens.length > 0) {
-              var matches = [];
-              // Truncate at the last space before maxLen so fragments never
-              // end mid-word (falls back to a hard cut for space-less text).
-              var truncAtWord = function(s, maxLen) {
-                s = String(s || '');
-                if (s.length <= maxLen) return s;
-                var cut = s.slice(0, maxLen);
-                var sp = cut.lastIndexOf(' ');
-                return sp > 40 ? cut.slice(0, sp) : cut;
-              };
-              // Only search last 200 episodes
-              var searchEps = epRaw.slice(-200);
-              for (var ei = 0; ei < searchEps.length; ei++) {
-                try {
-                  var ep = JSON.parse(searchEps[ei]);
-                  var sumLower = (ep.summary || '').toLowerCase();
-                  // Skip if from current session
-                  var currentSessId = process.env.CLAUDE_SESSION_ID || '';
-                  if (ep.sessionId === currentSessId) continue;
-                  // Count keyword overlap
-                  var hits = 0;
-                  for (var ti = 0; ti < promptTokens.length; ti++) {
-                    if (sumLower.includes(promptTokens[ti])) hits++;
-                  }
-                  var relevance = promptTokens.length > 0 ? hits / promptTokens.length : 0;
-                  if (relevance >= 0.3 && hits >= 2) {
-                    matches.push({ summary: truncAtWord(ep.summary || '', 120), relevance: relevance, ts: ep.endedAt || 0 });
-                  }
-                } catch (e) {}
-              }
-              // Show top 2 most relevant matches — dedupe by fragment text so
-              // two episodes from the same conversation don't inject the same
-              // truncated snippet twice.
-              if (matches.length > 0) {
-                matches.sort(function(a, b) { return b.relevance - a.relevance; });
-                var topMatches = [];
-                var seenFrags = {};
-                for (var tmi = 0; tmi < matches.length && topMatches.length < 2; tmi++) {
-                  var frag = matches[tmi].summary.replace(/\n/g, ' ').trim();
-                  if (seenFrags[frag]) continue;
-                  seenFrags[frag] = true;
-                  topMatches.push(matches[tmi]);
-                }
-                var memLines = ['[MEMORY] Relevant past sessions:'];
-                for (var mi = 0; mi < topMatches.length; mi++) {
-                  var ago = topMatches[mi].ts ? Math.round((Date.now() - topMatches[mi].ts) / 3600000) + 'h ago' : '';
-                  memLines.push('  ' + topMatches[mi].summary.replace(/\n/g, ' ').trim() + (ago ? ' (' + ago + ')' : ''));
-                }
-                console.log(memLines.join('\n'));
-              }
-            }
-          }
-        }
-      } catch (e) { /* non-fatal */ }
-
-      // Inject monograph hint for complex tasks.
-      // Source of truth is .monomind/monograph.db (SQLite). Legacy stats.json
-      // is no longer written by the build, so it is checked only as a fallback.
-      try {
-        var legacyStats = path.join(CWD, '.monomind', 'graph', 'stats.json');
-        var nodeCount = 0;
-        try {
-          var hintDb = hCtx._openMonographDb();
-          if (hintDb) {
-            nodeCount = hintDb.prepare('SELECT COUNT(*) AS c FROM nodes').get().c;
-          }
-        } catch (e) { /* ignore — fall back to legacy */ }
-        if (nodeCount === 0 && fs.existsSync(legacyStats)) {
-          try {
-            var legacyStatsSt = fs.statSync(legacyStats);
-            if (legacyStatsSt.size <= 1024 * 1024) {
-              var gStats = JSON.parse(fs.readFileSync(legacyStats, 'utf-8'));
-              nodeCount = gStats.nodes || 0;
-            }
-          } catch (e) { /* ignore */ }
-        }
-        if (nodeCount > 100 && hCtx._isGraphFresh()) {
-          // Pre-resolve top-3 relevant files for the user's prompt — the LLM
-          // sees the answer inline instead of being told to call a tool.
-          // 3 is enough signal; more files inflate token cost on every prompt.
-          var suggestions = hCtx.getMonographSuggestions(prompt, 3);
-
-          // Boost recently-edited files to the top of pre-resolve suggestions.
-          // Even when the FTS index hasn't caught up to the latest edits, the
-          // LLM should see the files it just modified as the primary context.
-          try {
-            var recentEditsForRoute = hCtx._getRecentEdits();
-            if (recentEditsForRoute.length > 0) {
-              // Extract prompt keywords for relevance gating
-              var promptWords = (prompt || '').toLowerCase().match(/[a-z][a-z0-9_-]{2,}/g) || [];
-              var promptWordSet = {};
-              for (var pw = 0; pw < promptWords.length; pw++) promptWordSet[promptWords[pw]] = 1;
-
-              var existingFiles = {};
-              for (var se = 0; se < suggestions.length; se++) existingFiles[suggestions[se].file || ''] = 1;
-
-              var editBoosts = [];
-              for (var re = 0; re < recentEditsForRoute.length && editBoosts.length < 2; re++) {
-                var reFile = recentEditsForRoute[re].file;
-                // Skip if already in suggestions
-                if (existingFiles[reFile]) continue;
-                var reName = path.basename(reFile, path.extname(reFile)).toLowerCase();
-                // Only boost if filename shares a keyword with the prompt OR the edit is very recent (<3 min)
-                var veryRecent = (Date.now() - recentEditsForRoute[re].editedAt) < 3 * 60 * 1000;
-                var editMatches = promptWordSet[reName] || veryRecent;
-                if (editMatches) {
-                  editBoosts.push({ name: path.basename(reFile), label: 'File', file: reFile, deg: 0, _editBoost: true });
-                }
-              }
-              if (editBoosts.length > 0) {
-                suggestions = editBoosts.concat(suggestions).slice(0, 3);
-              }
-            }
-          } catch (e) { /* non-fatal */ }
-
-          if (suggestions.length > 0) {
-            // Compact single-line format: "[MONOGRAPH] N nodes. Top files: name [Label] — path:line, ..."
-            // Matches the agent-start-handler pattern — keeps per-prompt token cost minimal.
-            var hintParts = suggestions.map(function(s) {
-              var editTag = s._editBoost ? ' ✎' : '';
-              var fileLoc = (s.file || '');
-              if (fileLoc && s.startLine != null) fileLoc = fileLoc + ':' + s.startLine;
-              return s.name + ' [' + s.label + '] — ' + fileLoc + editTag;
-            });
-            console.log('[MONOGRAPH] ' + nodeCount + ' nodes. Top files: ' + hintParts.join(' · '));
-            hCtx._recordGraphTelemetry('preresolve_hit');
-          } else {
-            console.log('[MONOGRAPH] ' + nodeCount + ' nodes. Call mcp__monomind__monograph_suggest to find relevant files.');
-            hCtx._recordGraphTelemetry('preresolve_miss');
-          }
-        }
-      } catch(e) {}
-
-      // Swarm mode selection is available on-demand via /mastermind slash command.
-    } else {
-      console.log('[INFO] Router not available, using default routing');
-    }
-
-    // Task 22: TeamRoutingModes — only log when an explicit swarm config is present
-    try {
-      var swarmCfgPath = path.join(CWD, '.monomind', 'swarm-config.json');
-      if (fs.existsSync(swarmCfgPath)) {
-        var swarmCfgSt = fs.statSync(swarmCfgPath);
-        var topology22 = swarmCfgSt.size <= 1024 * 1024 ? (JSON.parse(fs.readFileSync(swarmCfgPath, 'utf-8')).topology || 'mesh') : 'mesh';
-        var mode22 = topology22 === 'hierarchical' ? 'route' : 'coordinate';
-        console.log('[ROUTING_MODE] topology=' + topology22 + ' → mode=' + mode22);
-      }
-    } catch (e) { /* non-fatal */ }
-  }
-};