gitnexus 1.4.8 → 1.4.10

package/dist/mcp/local/local-backend.js

@@ -7,7 +7,8 @@
  */
 import fs from 'fs/promises';
 import path from 'path';
-import { initLbug, executeQuery, executeParameterized, closeLbug, isLbugReady } from '../core/lbug-adapter.js';
+import { initLbug, executeQuery, executeParameterized, closeLbug, isLbugReady, isWriteQuery } from '../core/lbug-adapter.js';
+export { isWriteQuery };
 // Embedding imports are lazy (dynamic import) to avoid loading onnxruntime-node
 // at MCP server startup — crashes on unsupported Node ABI versions (#89)
 // git utilities available if needed
@@ -35,9 +36,11 @@ export const VALID_NODE_LABELS = new Set([
     'Community', 'Process', 'Struct', 'Enum', 'Macro', 'Typedef', 'Union',
     'Namespace', 'Trait', 'Impl', 'TypeAlias', 'Const', 'Static', 'Property',
     'Record', 'Delegate', 'Annotation', 'Constructor', 'Template', 'Module',
+    'Route',
+    'Tool',
 ]);
 /** Valid relation types for impact analysis filtering */
-export const VALID_RELATION_TYPES = new Set(['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'HAS_METHOD', 'HAS_PROPERTY', 'OVERRIDES', 'ACCESSES']);
+export const VALID_RELATION_TYPES = new Set(['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'HAS_METHOD', 'HAS_PROPERTY', 'OVERRIDES', 'ACCESSES', 'HANDLES_ROUTE', 'FETCHES', 'HANDLES_TOOL', 'ENTRY_POINT_OF', 'WRAPS']);
 /**
  * Per-relation-type confidence floor for impact analysis.
  *
@@ -73,12 +76,6 @@ export const IMPACT_RELATION_CONFIDENCE = {
  * Falls back to 0.5 for unknown types so they are not silently elevated.
  */
 const confidenceForRelType = (relType) => IMPACT_RELATION_CONFIDENCE[relType ?? ''] ?? 0.5;
-/** Regex to detect write operations in user-supplied Cypher queries */
-export const CYPHER_WRITE_RE = /\b(CREATE|DELETE|SET|MERGE|REMOVE|DROP|ALTER|COPY|DETACH)\b/i;
-/** Check if a Cypher query contains write operations */
-export function isWriteQuery(query) {
-    return CYPHER_WRITE_RE.test(query);
-}
 /** Structured error logging for query failures — replaces empty catch blocks */
 function logQueryError(context, err) {
     const msg = err instanceof Error ? err.message : String(err);
@@ -346,6 +343,14 @@ export class LocalBackend {
                 return this.context(repo, { name: params?.name, ...params });
             case 'overview':
                 return this.overview(repo, params);
+            case 'route_map':
+                return this.routeMap(repo, params);
+            case 'shape_check':
+                return this.shapeCheck(repo, params);
+            case 'tool_map':
+                return this.toolMap(repo, params);
+            case 'api_impact':
+                return this.apiImpact(repo, params);
             default:
                 throw new Error(`Unknown tool: ${method}`);
         }
@@ -370,10 +375,12 @@ export class LocalBackend {
         const searchQuery = params.query.trim();
         // Step 1: Run hybrid search to get matching symbols
         const searchLimit = processLimit * maxSymbolsPerProcess; // fetch enough raw results
-        const [bm25Results, semanticResults] = await Promise.all([
+        const [bm25SearchResult, semanticResults] = await Promise.all([
             this.bm25Search(repo, searchQuery, searchLimit),
             this.semanticSearch(repo, searchQuery, searchLimit),
         ]);
+        const bm25Results = bm25SearchResult.results;
+        const ftsUsed = bm25SearchResult.ftsUsed;
         // Merge via reciprocal rank fusion
         const scoreMap = new Map();
         for (let i = 0; i < bm25Results.length; i++) {
@@ -540,6 +547,7 @@ export class LocalBackend {
             processes,
             process_symbols: dedupedSymbols,
             definitions: definitions.slice(0, 20), // cap standalone definitions
+            ...(!ftsUsed && { warning: 'FTS extension unavailable - keyword search degraded. Run: gitnexus analyze --force to rebuild indexes.' }),
         };
     }
     /**
@@ -553,8 +561,9 @@ export class LocalBackend {
         }
         catch (err) {
             console.error('GitNexus: BM25/FTS search failed (FTS indexes may not exist) -', err.message);
-            return [];
+            return { results: [], ftsUsed: false };
         }
+        const ftsUsed = bm25Results.length === 0 || (bm25Results[0]?.ftsUsed !== false);
         const results = [];
         for (const bm25Result of bm25Results) {
             const fullPath = bm25Result.filePath;
@@ -598,7 +607,7 @@ export class LocalBackend {
                 });
             }
         }
-        return results;
+        return { results, ftsUsed };
     }
     /**
      * Semantic vector search helper
@@ -671,7 +680,7 @@ export class LocalBackend {
             return { error: 'LadybugDB not ready. Index may be corrupted.' };
         }
         // Block write operations (defense-in-depth — DB is already read-only)
-        if (CYPHER_WRITE_RE.test(params.query)) {
+        if (isWriteQuery(params.query)) {
             return { error: 'Write operations (CREATE, DELETE, SET, MERGE, REMOVE, DROP, ALTER, COPY, DETACH) are not allowed. The knowledge graph is read-only.' };
         }
         try {
@@ -847,6 +856,47 @@ export class LocalBackend {
             return { error: `Symbol '${name || uid}' not found` };
         }
         // Step 2: Disambiguation
+        // When multiple nodes share the same name (e.g. a Java Class and its
+        // Constructor both named 'SessionTracker'), prefer the Class node so
+        // context() returns the semantically meaningful result rather than
+        // triggering ambiguous disambiguation (#480).
+        // labels(n)[0] returns empty string in LadybugDB, so we resolve the
+        // preferred node by re-querying with explicit label filters, scoped to
+        // the candidate IDs already in symbols.
+        //
+        // Guard: only attempt Class-preference when at least one candidate has an
+        // empty/unknown type (LadybugDB limitation) or is a Constructor — meaning
+        // the ambiguity may be a Class/Constructor name collision rather than two
+        // genuinely distinct symbols (e.g. two Functions in different files).
+        //
+        // resolvedLabel is set here and threaded to Step 3 to avoid a redundant
+        // classCheck round-trip later.
+        let resolvedLabel = '';
+        if (symbols.length > 1 && !uid) {
+            const hasAmbiguousType = symbols.some((s) => {
+                const t = s.type || s[2] || '';
+                return t === '' || t === 'Constructor';
+            });
+            if (hasAmbiguousType) {
+                const candidateIds = symbols.map((s) => s.id || s[0]).filter(Boolean);
+                const PREFER_LABELS = ['Class', 'Interface'];
+                let preferred = null;
+                for (const label of PREFER_LABELS) {
+                    const match = await executeParameterized(repo.id, `
+            MATCH (n:\`${label}\`) WHERE n.id IN $candidateIds RETURN n.id AS id LIMIT 1
+          `, { candidateIds }).catch(() => []);
+                    if (match.length > 0) {
+                        preferred = symbols.find((s) => (s.id || s[0]) === (match[0].id || match[0][0]));
+                        if (preferred) {
+                            resolvedLabel = label;
+                            break;
+                        }
+                    }
+                }
+                if (preferred)
+                    symbols = [preferred];
+            }
+        }
         if (symbols.length > 1 && !uid) {
             return {
                 status: 'ambiguous',
@@ -864,12 +914,74 @@ export class LocalBackend {
         const sym = symbols[0];
         const symId = sym.id || sym[0];
         // Categorized incoming refs
-        const incomingRows = await executeParameterized(repo.id, `
+        let incomingRows = await executeParameterized(repo.id, `
       MATCH (caller)-[r:CodeRelation]->(n {id: $symId})
       WHERE r.type IN ['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'HAS_METHOD', 'HAS_PROPERTY', 'OVERRIDES', 'ACCESSES']
       RETURN r.type AS relType, caller.id AS uid, caller.name AS name, caller.filePath AS filePath, labels(caller)[0] AS kind
       LIMIT 30
     `, { symId });
+        // Fix #480: Class/Interface nodes have no direct CALLS/IMPORTS edges —
+        // those point to Constructor and File nodes respectively. Fetch those
+        // extra incoming refs and merge them in so context() shows real callers.
+        //
+        // Determine if this is a Class/Interface node. If resolvedLabel was set
+        // during disambiguation (Step 2), use it directly — no extra round-trip.
+        // Otherwise fall back to a single label check only when the type field is
+        // empty (LadybugDB labels(n)[0] limitation).
+        const symRawType = sym.type || sym[2] || '';
+        let isClassLike = resolvedLabel === 'Class' || resolvedLabel === 'Interface';
+        if (!isClassLike && symRawType === '') {
+            try {
+                // Single UNION query instead of two serial round-trips.
+                const typeCheck = await executeParameterized(repo.id, `
+          MATCH (n:Class) WHERE n.id = $symId RETURN 'Class' AS label LIMIT 1
+          UNION ALL
+          MATCH (n:Interface) WHERE n.id = $symId RETURN 'Interface' AS label LIMIT 1
+        `, { symId });
+                isClassLike = typeCheck.length > 0;
+            }
+            catch { /* not a Class/Interface node */ }
+        }
+        else if (!isClassLike) {
+            isClassLike = symRawType === 'Class' || symRawType === 'Interface';
+        }
+        if (isClassLike) {
+            try {
+                // Run both incoming-ref queries in parallel — they are independent.
+                const [ctorIncoming, fileIncoming] = await Promise.all([
+                    executeParameterized(repo.id, `
+            MATCH (n)-[hm:CodeRelation]->(ctor:Constructor)
+            WHERE n.id = $symId AND hm.type = 'HAS_METHOD'
+            MATCH (caller)-[r:CodeRelation]->(ctor)
+            WHERE r.type IN ['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'ACCESSES']
+            RETURN r.type AS relType, caller.id AS uid, caller.name AS name, caller.filePath AS filePath, labels(caller)[0] AS kind
+            LIMIT 30
+          `, { symId }),
+                    executeParameterized(repo.id, `
+            MATCH (f:File)-[rel:CodeRelation]->(n)
+            WHERE n.id = $symId AND rel.type = 'DEFINES'
+            MATCH (caller)-[r:CodeRelation]->(f)
+            WHERE r.type IN ['CALLS', 'IMPORTS']
+            RETURN r.type AS relType, caller.id AS uid, caller.name AS name, caller.filePath AS filePath, labels(caller)[0] AS kind
+            LIMIT 30
+          `, { symId }),
+                ]);
+                // Deduplicate by (relType, uid) — a caller can have multiple relation
+                // types to the same target (e.g. both IMPORTS and CALLS), and each
+                // must be preserved so every category appears in the output.
+                const seenKeys = new Set(incomingRows.map((r) => `${r.relType || r[0]}:${r.uid || r[1]}`));
+                for (const r of [...ctorIncoming, ...fileIncoming]) {
+                    const key = `${r.relType || r[0]}:${r.uid || r[1]}`;
+                    if (!seenKeys.has(key)) {
+                        seenKeys.add(key);
+                        incomingRows.push(r);
+                    }
+                }
+            }
+            catch (e) {
+                logQueryError('context:class-incoming-expansion', e);
+            }
+        }
         // Categorized outgoing refs
         const outgoingRows = await executeParameterized(repo.id, `
       MATCH (n {id: $symId})-[r:CodeRelation]->(target)
@@ -910,7 +1022,7 @@ export class LocalBackend {
             symbol: {
                 uid: sym.id || sym[0],
                 name: sym.name || sym[1],
-                kind: sym.type || sym[2],
+                kind: isClassLike ? (resolvedLabel || 'Class') : (sym.type || sym[2]),
                 filePath: sym.filePath || sym[3],
                 startLine: sym.startLine || sym[4],
                 endLine: sym.endLine || sym[5],
@@ -1294,20 +1406,100 @@ export class LocalBackend {
         const minConfidence = params.minConfidence ?? 0;
         const relTypeFilter = relationTypes.map(t => `'${t}'`).join(', ');
         const confidenceFilter = minConfidence > 0 ? ` AND r.confidence >= ${minConfidence}` : '';
-        const targets = await executeParameterized(repo.id, `
-      MATCH (n)
-      WHERE n.name = $targetName
-      RETURN n.id AS id, n.name AS name, labels(n)[0] AS type, n.filePath AS filePath
-      LIMIT 1
-    `, { targetName: target });
-        if (targets.length === 0)
+        // Resolve target by name, preferring Class/Interface over Constructor
+        // (fix #480: Java class and constructor share the same name).
+        // labels(n)[0] returns empty string in LadybugDB, so we use explicit
+        // label-typed sub-queries in a single UNION ordered by priority to avoid
+        // up to 6 serial round-trips for non-Class targets.
+        let sym = null;
+        let symType = '';
+        try {
+            const rows = await executeParameterized(repo.id, `
+        MATCH (n:\`Class\`) WHERE n.name = $targetName
+        RETURN n.id AS id, n.name AS name, n.filePath AS filePath, 0 AS priority LIMIT 1
+        UNION ALL
+        MATCH (n:\`Interface\`) WHERE n.name = $targetName
+        RETURN n.id AS id, n.name AS name, n.filePath AS filePath, 1 AS priority LIMIT 1
+        UNION ALL
+        MATCH (n:\`Function\`) WHERE n.name = $targetName
+        RETURN n.id AS id, n.name AS name, n.filePath AS filePath, 2 AS priority LIMIT 1
+        UNION ALL
+        MATCH (n:\`Method\`) WHERE n.name = $targetName
+        RETURN n.id AS id, n.name AS name, n.filePath AS filePath, 3 AS priority LIMIT 1
+        UNION ALL
+        MATCH (n:\`Constructor\`) WHERE n.name = $targetName
+        RETURN n.id AS id, n.name AS name, n.filePath AS filePath, 4 AS priority LIMIT 1
+      `, { targetName: target }).catch(() => []);
+            if (rows.length > 0) {
+                // Pick the row with the lowest priority value (Class wins over Constructor)
+                const best = rows.reduce((a, b) => (a.priority ?? a[3] ?? 99) <= (b.priority ?? b[3] ?? 99) ? a : b);
+                sym = best;
+                const priorityToLabel = ['Class', 'Interface', 'Function', 'Method', 'Constructor'];
+                symType = priorityToLabel[best.priority ?? best[3]] ?? '';
+            }
+        }
+        catch { /* fall through to unlabeled match */ }
+        // Fall back to unlabeled match for any other node type
+        if (!sym) {
+            const rows = await executeParameterized(repo.id, `
+        MATCH (n)
+        WHERE n.name = $targetName
+        RETURN n.id AS id, n.name AS name, n.filePath AS filePath
+        LIMIT 1
+      `, { targetName: target });
+            if (rows.length > 0)
+                sym = rows[0];
+        }
+        if (!sym)
             return { error: `Target '${target}' not found` };
-        const sym = targets[0];
         const symId = sym.id || sym[0];
         const impacted = [];
         const visited = new Set([symId]);
         let frontier = [symId];
         let traversalComplete = true;
+        // Fix #480: For Java (and other JVM) Class/Interface nodes, CALLS edges
+        // point to Constructor nodes and IMPORTS edges point to File nodes — not
+        // the Class/Interface itself. Seed the frontier with the Constructor(s)
+        // and owning File so the BFS traversal finds those edges naturally.
+        // The owning File is kept only as an internal seed (frontier/visited) and
+        // is NOT added to impacted — it is the definition container, not an
+        // upstream dependent. The BFS will discover IMPORTS edges on it naturally.
+        if (symType === 'Class' || symType === 'Interface') {
+            try {
+                // Run both seed queries in parallel — they are independent.
+                const [ctorRows, fileRows] = await Promise.all([
+                    executeParameterized(repo.id, `
+            MATCH (n)-[hm:CodeRelation]->(c:Constructor)
+            WHERE n.id = $symId AND hm.type = 'HAS_METHOD'
+            RETURN c.id AS id, c.name AS name, labels(c)[0] AS type, c.filePath AS filePath
+          `, { symId }),
+                    // Restrict to DEFINES edges only — other File->Class edge types (if
+                    // any) should not be treated as the owning file relationship.
+                    executeParameterized(repo.id, `
+            MATCH (f:File)-[rel:CodeRelation]->(n)
+            WHERE n.id = $symId AND rel.type = 'DEFINES'
+            RETURN f.id AS id, f.name AS name, labels(f)[0] AS type, f.filePath AS filePath
+          `, { symId }),
+                ]);
+                for (const r of ctorRows) {
+                    const rid = r.id || r[0];
+                    if (rid && !visited.has(rid)) {
+                        visited.add(rid);
+                        frontier.push(rid);
+                    }
+                }
+                for (const r of fileRows) {
+                    const rid = r.id || r[0];
+                    if (rid && !visited.has(rid)) {
+                        visited.add(rid);
+                        frontier.push(rid);
+                    }
+                }
+            }
+            catch (e) {
+                logQueryError('impact:class-node-expansion', e);
+            }
+        }
         for (let depth = 1; depth <= maxDepth && frontier.length > 0; depth++) {
             const nextFrontier = [];
             // Batch frontier nodes into a single Cypher query per depth level
@@ -1364,62 +1556,201 @@ export class LocalBackend {
         let affectedProcesses = [];
         let affectedModules = [];
         if (impacted.length > 0) {
-            // Cap IN-clause to 100 IDs to prevent oversized queries that crash
-            // the native DB engine on arm64 macOS (#292)
-            const cappedImpacted = impacted.slice(0, 100);
-            const allIds = cappedImpacted.map(i => `'${String(i.id ?? '').replace(/'/g, "''")}'`).join(', ');
-            const d1Items = (grouped[1] || []).slice(0, 100);
-            const d1Ids = d1Items.map((i) => `'${String(i.id ?? '').replace(/'/g, "''")}'`).join(', ');
-            // Enrichment queries: sequential on arm64 macOS to avoid SIGSEGV from
-            // concurrent native DB access (#285, #290, #292); parallel elsewhere
-            // to preserve performance on unaffected platforms.
-            const isArm64Mac = process.platform === 'darwin' && process.arch === 'arm64';
-            const processQuery = executeQuery(repo.id, `
-        MATCH (s)-[r:CodeRelation {type: 'STEP_IN_PROCESS'}]->(p:Process)
-        WHERE s.id IN [${allIds}]
-        RETURN p.heuristicLabel AS name, COUNT(DISTINCT s.id) AS hits, MIN(r.step) AS minStep, p.stepCount AS stepCount
-        ORDER BY hits DESC
-        LIMIT 20
-      `).catch(() => []);
-            const moduleQuery = () => executeQuery(repo.id, `
-        MATCH (s)-[:CodeRelation {type: 'MEMBER_OF'}]->(c:Community)
-        WHERE s.id IN [${allIds}]
-        RETURN c.heuristicLabel AS name, COUNT(DISTINCT s.id) AS hits
-        ORDER BY hits DESC
-        LIMIT 20
-      `).catch(() => []);
-            const directModuleQuery = () => d1Ids
-                ? executeQuery(repo.id, `
+            const CHUNK_SIZE = 100;
+            // Max number of chunks to process to avoid unbounded DB round-trips.
+            // Configurable via env IMPACT_MAX_CHUNKS, default 10 => max items = 1000
+            const MAX_CHUNKS = parseInt(process.env.IMPACT_MAX_CHUNKS || '10', 10);
+            // ── Process enrichment: batched chunking (bounded by MAX_CHUNKS) ─
+            // Uses merged Cypher query (WITH + OPTIONAL MATCH) to fetch
+            // process + entry point info in 1 round-trip per chunk. Converted to
+            // parameterized queries to avoid manual string escaping and long query strings.
+            const entryPointMap = new Map();
+            // Map process id -> entryPointId to allow fixing missing minStep values later
+            const processToEntryPoint = new Map();
+            // Collect process ids where MIN(r.step) returned null so we can retry in batch
+            const processesMissingMinStep = new Set();
+            let chunksProcessed = 0;
+            for (let i = 0; i < impacted.length && chunksProcessed < MAX_CHUNKS; i += CHUNK_SIZE, chunksProcessed++) {
+                const chunk = impacted.slice(i, i + CHUNK_SIZE);
+                const ids = chunk.map(item => String(item.id ?? ''));
+                try {
+                    // Use parameterized list to avoid building long query strings
+                    const rows = await executeParameterized(repo.id, `
+            MATCH (s)-[r:CodeRelation {type: 'STEP_IN_PROCESS'}]->(p:Process)
+            WHERE s.id IN $ids
+            WITH p, COUNT(DISTINCT s.id) AS hits, MIN(r.step) AS minStep
+            OPTIONAL MATCH (ep {id: p.entryPointId})
+            RETURN p.id AS pId, p.heuristicLabel AS name, p.processType AS processType,
+                   p.entryPointId AS entryPointId, hits, minStep, p.stepCount AS stepCount,
+                   ep.name AS epName, labels(ep)[0] AS epType, ep.filePath AS epFilePath
+          `, { ids }).catch(() => []);
+                    for (const row of rows) {
+                        const pId = row.pId ?? row[0];
+                        const epId = row.entryPointId ?? row[3] ?? row.pId ?? row[0];
+                        // Track mapping from process -> entryPoint so we can backfill missing minStep
+                        if (pId)
+                            processToEntryPoint.set(String(pId), String(epId));
+                        // Normalize epName: prefer epName, fall back to other columns, and
+                        // ensure we don't keep an empty string (labels(...) can return "").
+                        const epNameRaw = row.epName ?? row[7] ?? row.name ?? row[1] ?? 'unknown';
+                        const epName = (typeof epNameRaw === 'string' && epNameRaw.trim().length > 0) ? epNameRaw.trim() : 'unknown';
+                        // Normalize epType: labels(ep)[0] can return an empty string in
+                        // some DBs (LadybugDB). Using nullish coalescing (??) preserves
+                        // empty strings, which results in empty `type` values being
+                        // propagated. Treat empty-string labels as missing and fall back
+                        // to the next candidate or a sensible default.
+                        const epTypeRaw = row.epType ?? row[8] ?? '';
+                        const epType = (typeof epTypeRaw === 'string' && epTypeRaw.trim().length > 0)
+                            ? epTypeRaw.trim()
+                            : 'Function';
+                        const epFilePath = row.epFilePath ?? row[9] ?? '';
+                        const hits = row.hits ?? row[4] ?? 0;
+                        const minStep = row.minStep ?? row[5];
+                        // If the DB returned null for minStep, note the process id so we
+                        // can run a follow-up query using a different aggregation strategy.
+                        if (minStep === null || minStep === undefined) {
+                            if (pId)
+                                processesMissingMinStep.add(String(pId));
+                        }
+                        if (!entryPointMap.has(epId)) {
+                            entryPointMap.set(epId, {
+                                name: epName,
+                                type: epType,
+                                filePath: epFilePath,
+                                affected_process_count: 0,
+                                total_hits: 0,
+                                earliest_broken_step: Infinity,
+                            });
+                        }
+                        const ep = entryPointMap.get(epId);
+                        ep.affected_process_count += 1;
+                        ep.total_hits += hits;
+                        ep.earliest_broken_step = Math.min(ep.earliest_broken_step, minStep ?? Infinity);
+                    }
+                }
+                catch (e) {
+                    logQueryError('impact:process-chunk', e);
+                }
+            }
+            // If some processes returned null minStep, try a batched follow-up query
+            // using the full impacted id set. This handles older indexes or DBs
+            // where MIN(r.step) can come back null even when step properties exist.
+            if (processesMissingMinStep.size > 0) {
+                try {
+                    const pIds = Array.from(processesMissingMinStep);
+                    const allImpactedIds = impacted.map(it => String(it.id ?? ''));
+                    const missingRows = await executeParameterized(repo.id, `
+            MATCH (s)-[r:CodeRelation {type: 'STEP_IN_PROCESS'}]->(p:Process)
+            WHERE p.id IN $pIds AND s.id IN $ids
+            RETURN p.id AS pid, MIN(r.step) AS minStep
+          `, { pIds, ids: allImpactedIds }).catch(() => []);
+                    for (const mr of missingRows) {
+                        const pid = mr.pid ?? mr[0];
+                        const minStep = mr.minStep ?? mr[1];
+                        const epId = processToEntryPoint.get(String(pid));
+                        if (!epId)
+                            continue;
+                        const ep = entryPointMap.get(epId);
+                        if (!ep)
+                            continue;
+                        if (typeof minStep === 'number') {
+                            ep.earliest_broken_step = Math.min(ep.earliest_broken_step, minStep);
+                        }
+                    }
+                }
+                catch (e) {
+                    logQueryError('impact:process-chunk-backfill', e);
+                }
+            }
+            // If we capped chunks, mark traversal incomplete so caller knows results are partial
+            if (chunksProcessed * CHUNK_SIZE < impacted.length) {
+                traversalComplete = false;
+            }
+            affectedProcesses = Array.from(entryPointMap.values())
+                .map(ep => ({
+                ...ep,
+                earliest_broken_step: ep.earliest_broken_step === Infinity ? null : ep.earliest_broken_step,
+            }))
+                .sort((a, b) => b.total_hits - a.total_hits);
+            // ── Module enrichment: use same cap as process enrichment and parameterized queries
+            const maxItems = Math.min(impacted.length, MAX_CHUNKS * CHUNK_SIZE);
+            const cappedImpacted = impacted.slice(0, maxItems);
+            const allIdsArr = cappedImpacted.map((i) => String(i.id ?? ''));
+            const d1Items = (grouped[1] || []).slice(0, maxItems);
+            const d1IdsArr = d1Items.map((i) => String(i.id ?? ''));
+            // Chunked module enrichment: run the MEMBER_OF queries in chunks
+            // to avoid large single queries or concurrent Kuzu calls that can
+            // crash (SIGSEGV) on arm64 macOS; behavior preserves existing maxItems cap and returns equivalent aggregated results.
+            const moduleHitsMap = new Map();
+            const directModuleSet = new Set();
+            // Helper to run a single module chunk and accumulate hits by name
+            const runModuleChunk = async (idsChunk) => {
+                if (!idsChunk || idsChunk.length === 0)
+                    return;
+                try {
+                    const rows = await executeParameterized(repo.id, `
+            MATCH (s)-[:CodeRelation {type: 'MEMBER_OF'}]->(c:Community)
+            WHERE s.id IN $ids
+            RETURN c.heuristicLabel AS name, COUNT(DISTINCT s.id) AS hits
+            ORDER BY hits DESC
+            LIMIT 20
+          `, { ids: idsChunk }).catch(() => []);
+                    for (const r of rows) {
+                        const name = r.name ?? r[0] ?? null;
+                        const hits = (r.hits ?? r[1]) || 0;
+                        if (!name)
+                            continue;
+                        moduleHitsMap.set(name, (moduleHitsMap.get(name) || 0) + hits);
+                    }
+                }
+                catch (e) {
+                    logQueryError('impact:module-chunk', e);
+                }
+            };
+            // Run module query chunks sequentially (safe on arm64 macOS)
+            for (let i = 0; i < allIdsArr.length; i += CHUNK_SIZE) {
+                const chunkIds = allIdsArr.slice(i, i + CHUNK_SIZE);
+                await runModuleChunk(chunkIds);
+            }
+            // Run direct module query similarly (distinct heuristic labels for depth-1 items)
+            const runDirectModuleChunk = async (idsChunk) => {
+                if (!idsChunk || idsChunk.length === 0)
+                    return;
+                try {
+                    const rows = await executeParameterized(repo.id, `
             MATCH (s)-[:CodeRelation {type: 'MEMBER_OF'}]->(c:Community)
-            WHERE s.id IN [${d1Ids}]
+            WHERE s.id IN $ids
             RETURN DISTINCT c.heuristicLabel AS name
-          `).catch(() => [])
-                : Promise.resolve([]);
-            let processRows, moduleRows, directModuleRows;
-            if (isArm64Mac) {
-                // Sequential: avoid concurrent native DB access
-                processRows = await processQuery;
-                moduleRows = await moduleQuery();
-                directModuleRows = await directModuleQuery();
+          `, { ids: idsChunk }).catch(() => []);
+                    for (const r of rows) {
+                        const name = r.name ?? r[0] ?? null;
+                        if (name)
+                            directModuleSet.add(name);
+                    }
+                }
+                catch (e) {
+                    logQueryError('impact:direct-module-chunk', e);
+                }
+            };
+            for (let i = 0; i < d1IdsArr.length; i += CHUNK_SIZE) {
+                const chunkIds = d1IdsArr.slice(i, i + CHUNK_SIZE);
+                await runDirectModuleChunk(chunkIds);
             }
-            else {
-                // Parallel: safe on non-arm64 platforms
-                processRows = await processQuery;
-                [moduleRows, directModuleRows] = await Promise.all([moduleQuery(), directModuleQuery()]);
-            }
-            affectedProcesses = processRows.map((r) => ({
-                name: r.name || r[0],
-                hits: r.hits || r[1],
-                broken_at_step: r.minStep ?? r[2],
-                step_count: r.stepCount ?? r[3],
-            }));
-            const directModuleSet = new Set(directModuleRows.map((r) => r.name || r[0]));
+            // Build final moduleRows array from aggregated hits map, sorted & limited
+            const moduleRows = Array.from(moduleHitsMap.entries())
+                .map(([name, hits]) => ({ name, hits }))
+                .sort((a, b) => b.hits - a.hits)
+                .slice(0, 20);
+            const directModuleRows = Array.from(directModuleSet).map(name => ({ name }));
+            // Build affectedModules in the same shape as original implementation
+            const directModuleNameSet = new Set(directModuleRows.map((r) => r.name || r[0]));
             affectedModules = moduleRows.map((r) => {
-                const name = r.name || r[0];
+                const name = r.name ?? r[0];
+                const hits = r.hits ?? r[1] ?? 0;
                 return {
                     name,
-                    hits: r.hits || r[1],
-                    impact: directModuleSet.has(name) ? 'direct' : 'indirect',
+                    hits,
+                    impact: directModuleNameSet.has(name) ? 'direct' : 'indirect',
                 };
             });
         }
@@ -1440,8 +1771,8 @@ export class LocalBackend {
             target: {
                 id: symId,
                 name: sym.name || sym[1],
-                type: sym.type || sym[2],
-                filePath: sym.filePath || sym[3],
+                type: symType,
+                filePath: sym.filePath || sym[2],
             },
             direction,
             impactedCount: impacted.length,
@@ -1457,6 +1788,312 @@ export class LocalBackend {
             byDepth: grouped,
         };
     }
+    /**
+     * Fetch Route nodes with their consumers in a single query.
+     * Shared by routeMap and shapeCheck to avoid N+1 query patterns.
+     */
+    async fetchRoutesWithConsumers(repoId, routeFilter, params) {
+        const rows = await executeParameterized(repoId, `
+      MATCH (n:Route)
+      WHERE n.id STARTS WITH 'Route:' ${routeFilter}
+      OPTIONAL MATCH (consumer)-[r:CodeRelation]->(n)
+      WHERE r.type = 'FETCHES'
+      RETURN n.id AS routeId, n.name AS routeName, n.filePath AS handlerFile,
+             n.responseKeys AS responseKeys, n.errorKeys AS errorKeys, n.middleware AS middleware,
+             consumer.name AS consumerName, consumer.filePath AS consumerFile,
+             r.reason AS fetchReason
+    `, params);
+        // Strip wrapping quotes from DB array elements — CSV COPY stores ['key'] which
+        // LadybugDB may return as "'key'" rather than "key"
+        const stripQuotes = (keys) => keys ? keys.map(k => k.replace(/^['"]|['"]$/g, '')) : null;
+        const routeMap = new Map();
+        for (const row of rows) {
+            const id = row.routeId ?? row[0];
+            const name = row.routeName ?? row[1];
+            const filePath = row.handlerFile ?? row[2];
+            const responseKeys = stripQuotes(row.responseKeys ?? row[3] ?? null);
+            const errorKeys = stripQuotes(row.errorKeys ?? row[4] ?? null);
+            const middleware = stripQuotes(row.middleware ?? row[5] ?? null);
+            const consumerName = row.consumerName ?? row[6];
+            const consumerFile = row.consumerFile ?? row[7];
+            const fetchReason = row.fetchReason ?? row[8] ?? null;
+            if (!routeMap.has(id)) {
+                routeMap.set(id, { id, name, filePath, responseKeys, errorKeys, middleware, consumers: [] });
+            }
+            if (consumerName && consumerFile) {
+                // Parse accessed keys from reason field: "fetch-url-match|keys:data,pagination|fetches:3"
+                let accessedKeys;
+                let fetchCount;
+                if (fetchReason) {
+                    const keysMatch = fetchReason.match(/\|keys:([^|]+)/);
+                    if (keysMatch) {
+                        accessedKeys = keysMatch[1].split(',').filter(k => k.length > 0);
+                    }
+                    const fetchesMatch = fetchReason.match(/\|fetches:(\d+)/);
+                    if (fetchesMatch) {
+                        fetchCount = parseInt(fetchesMatch[1], 10);
+                    }
+                }
+                routeMap.get(id).consumers.push({
+                    name: consumerName,
+                    filePath: consumerFile,
+                    ...(accessedKeys ? { accessedKeys } : {}),
+                    ...(fetchCount && fetchCount > 1 ? { fetchCount } : {}),
+                });
+            }
+        }
+        return [...routeMap.values()];
+    }
+    /**
+     * Batch-fetch execution flows linked to a set of Route or Tool nodes.
+     * Single query instead of N+1.
+     */
+    async fetchLinkedFlowsBatch(repoId, nodeIds) {
+        const result = new Map();
+        if (nodeIds.length === 0)
+            return result;
+        try {
+            // Use list_contains to filter at DB level instead of fetching all and filtering in memory
+            const rows = await executeParameterized(repoId, `
+        MATCH (source)-[r:CodeRelation]->(proc:Process)
+        WHERE r.type = 'ENTRY_POINT_OF'
+          AND list_contains($nodeIds, source.id)
+        RETURN source.id AS sourceId, proc.label AS name
+      `, { nodeIds });
+            for (const row of rows) {
+                const sourceId = row.sourceId ?? row[0];
+                const name = row.name ?? row[1];
+                if (!name)
+                    continue;
+                let list = result.get(sourceId);
+                if (!list) {
+                    list = [];
+                    result.set(sourceId, list);
+                }
+                list.push(name);
+            }
+        }
+        catch { /* no ENTRY_POINT_OF edges yet */ }
+        return result;
+    }
+    async routeMap(repo, params) {
+        await this.ensureInitialized(repo.id);
+        const routeFilter = params.route ? `AND n.name CONTAINS $route` : '';
+        const queryParams = params.route ? { route: params.route } : {};
+        const routes = await this.fetchRoutesWithConsumers(repo.id, routeFilter, queryParams);
+        if (routes.length === 0) {
+            return { routes: [], total: 0, message: params.route ? `No routes matching "${params.route}"` : 'No routes found in this project.' };
+        }
+        const flowMap = await this.fetchLinkedFlowsBatch(repo.id, routes.map(r => r.id));
+        return {
+            routes: routes.map(r => ({
+                route: r.name, handler: r.filePath,
+                middleware: r.middleware || [],
+                consumers: r.consumers,
+                flows: flowMap.get(r.id) || [],
+            })),
+            total: routes.length,
+        };
+    }
+    async shapeCheck(repo, params) {
+        await this.ensureInitialized(repo.id);
+        const routeFilter = params.route ? `AND n.name CONTAINS $route` : '';
+        const queryParams = params.route ? { route: params.route } : {};
+        const allRoutes = await this.fetchRoutesWithConsumers(repo.id, routeFilter, queryParams);
+        const results = allRoutes
+            .filter(r => ((r.responseKeys && r.responseKeys.length > 0) || (r.errorKeys && r.errorKeys.length > 0)) && r.consumers.length > 0)
+            .map(r => {
+            // Keys already normalized by fetchRoutesWithConsumers (quotes stripped)
+            const responseKeys = r.responseKeys ?? [];
+            const errorKeys = r.errorKeys ?? [];
+            // Combined set: consumer accessing either success or error keys is valid
+            const allKnownKeys = new Set([...responseKeys, ...errorKeys]);
+            // Check each consumer's accessed keys against the route's response shape
+            const responseKeySet = new Set(responseKeys);
+            const consumers = r.consumers.map(c => {
+                if (!c.accessedKeys || c.accessedKeys.length === 0) {
+                    return { name: c.name, filePath: c.filePath };
+                }
+                const mismatched = c.accessedKeys.filter(k => !allKnownKeys.has(k));
+                // Keys in allKnownKeys but not in responseKeys — error-path access (e.g., .error from errorKeys)
+                const errorPathKeys = c.accessedKeys.filter(k => allKnownKeys.has(k) && !responseKeySet.has(k));
+                const isMultiFetch = (c.fetchCount ?? 1) > 1;
+                return {
+                    name: c.name,
+                    filePath: c.filePath,
+                    accessedKeys: c.accessedKeys,
+                    ...(mismatched.length > 0 ? { mismatched, mismatchConfidence: isMultiFetch ? 'low' : 'high' } : {}),
+                    ...(errorPathKeys.length > 0 ? { errorPathKeys } : {}),
+                    ...(isMultiFetch ? { attributionNote: `This file fetches ${c.fetchCount} routes — accessed keys may belong to a different route.` } : {}),
+                };
+            });
+            const hasMismatches = consumers.some(c => 'mismatched' in c && c.mismatched.length > 0);
+            return {
+                route: r.name,
+                handler: r.filePath,
+                ...(responseKeys.length > 0 ? { responseKeys } : {}),
+                ...(errorKeys.length > 0 ? { errorKeys } : {}),
+                consumers,
+                ...(hasMismatches ? { status: 'MISMATCH' } : {}),
+            };
+        });
+        const mismatchCount = results.filter(r => r.status === 'MISMATCH').length;
+        return {
+            routes: results,
+            total: results.length,
+            routesWithShapes: results.length,
+            ...(mismatchCount > 0 ? { mismatches: mismatchCount } : {}),
+            message: results.length === 0
+                ? 'No routes with both response shapes and consumers found.'
+                : mismatchCount > 0
+                    ? `Found ${results.length} route(s) with response shape data. ${mismatchCount} route(s) have consumer/shape mismatches.`
+                    : `Found ${results.length} route(s) with response shape data and consumers.`,
+        };
+    }
+    async toolMap(repo, params) {
+        await this.ensureInitialized(repo.id);
+        const toolFilter = params.tool ? `AND n.name CONTAINS $tool` : '';
+        const queryParams = params.tool ? { tool: params.tool } : {};
+        const rows = await executeParameterized(repo.id, `
+      MATCH (n:Tool)
+      WHERE n.id STARTS WITH 'Tool:' ${toolFilter}
+      RETURN n.id AS id, n.name AS name, n.filePath AS filePath, n.description AS description
+    `, queryParams);
+        if (rows.length === 0) {
+            return { tools: [], total: 0, message: params.tool ? `No tools matching "${params.tool}"` : 'No tool definitions found.' };
+        }
+        const toolIds = rows.map((r) => r.id ?? r[0]);
+        const flowMap = await this.fetchLinkedFlowsBatch(repo.id, toolIds);
+        return {
+            tools: rows.map((r) => {
+                const id = r.id ?? r[0];
+                return {
+                    name: r.name ?? r[1],
+                    filePath: r.filePath ?? r[2],
+                    description: (r.description ?? r[3] ?? '').slice(0, 200),
+                    flows: flowMap.get(id) || [],
+                };
+            }),
+            total: rows.length,
+        };
+    }
+    async apiImpact(repo, params) {
+        await this.ensureInitialized(repo.id);
+        if (!params.route && !params.file) {
+            return { error: 'Either "route" or "file" parameter is required.' };
+        }
+        // If file is provided but route is not, look up the route by file path
+        let routeFilter = '';
+        const queryParams = {};
+        if (params.route) {
+            routeFilter = `AND n.name CONTAINS $route`;
+            queryParams.route = params.route;
+        }
+        else if (params.file) {
+            routeFilter = `AND n.filePath CONTAINS $file`;
+            queryParams.file = params.file;
+        }
+        const routes = await this.fetchRoutesWithConsumers(repo.id, routeFilter, queryParams);
+        if (routes.length === 0) {
+            const target = params.route || params.file;
+            return { error: `No routes found matching "${target}".` };
+        }
+        const flowMap = await this.fetchLinkedFlowsBatch(repo.id, routes.map(r => r.id));
+        // Count how many routes share the same handler file (for middleware partial detection)
+        const routeCountByHandler = new Map();
+        for (const r of routes) {
+            if (r.filePath) {
+                routeCountByHandler.set(r.filePath, (routeCountByHandler.get(r.filePath) ?? 0) + 1);
+            }
+        }
+        const results = routes.map(r => {
+            // Keys already normalized by fetchRoutesWithConsumers (quotes stripped)
+            const responseKeys = r.responseKeys ?? [];
+            const errorKeys = r.errorKeys ?? [];
+            const allKnownKeys = new Set([...responseKeys, ...errorKeys]);
+            // Build consumer list with mismatch detection
+            const consumers = r.consumers.map(c => ({
+                name: c.name,
+                file: c.filePath,
+                accesses: c.accessedKeys ?? [],
+                ...(c.fetchCount && c.fetchCount > 1 ? { attributionNote: `This file fetches ${c.fetchCount} routes — accessed keys may belong to a different route.` } : {}),
+            }));
+            // Detect mismatches: consumer accesses keys not in response shape
+            const mismatches = [];
+            if (allKnownKeys.size > 0) {
+                for (const c of r.consumers) {
+                    if (!c.accessedKeys)
+                        continue;
+                    const isMultiFetch = (c.fetchCount ?? 1) > 1;
+                    for (const key of c.accessedKeys) {
+                        if (!allKnownKeys.has(key)) {
+                            mismatches.push({
+                                consumer: c.filePath,
+                                field: key,
+                                reason: 'accessed but not in response shape',
+                                confidence: isMultiFetch ? 'low' : 'high',
+                            });
+                        }
+                    }
+                }
+            }
+            const flows = flowMap.get(r.id) || [];
+            const consumerCount = r.consumers.length;
+            // Risk level heuristic
+            let riskLevel;
+            if (consumerCount >= 10) {
+                riskLevel = 'HIGH';
+            }
+            else if (consumerCount >= 4) {
+                riskLevel = 'MEDIUM';
+            }
+            else {
+                riskLevel = 'LOW';
+            }
+            // Bump up one level if mismatches exist
+            if (mismatches.length > 0) {
+                if (riskLevel === 'LOW')
+                    riskLevel = 'MEDIUM';
+                else if (riskLevel === 'MEDIUM')
+                    riskLevel = 'HIGH';
+            }
+            const warning = consumerCount > 0
+                ? `Changing response shape will affect ${consumerCount} component${consumerCount === 1 ? '' : 's'}`
+                : undefined;
+            // Flag when middleware was detected but handler exports multiple HTTP methods
+            // (middleware chain may only reflect one export)
+            const middlewareArr = r.middleware || [];
+            const handlerRouteCount = r.filePath ? (routeCountByHandler.get(r.filePath) ?? 1) : 1;
+            const middlewarePartial = middlewareArr.length > 0 && handlerRouteCount > 1;
+            return {
+                route: r.name,
+                handler: r.filePath,
+                responseShape: {
+                    success: responseKeys,
+                    error: errorKeys,
+                },
+                middleware: middlewareArr,
+                ...(middlewarePartial ? {
+                    middlewareDetection: 'partial',
+                    middlewareNote: 'Middleware captured from first HTTP method export only — other methods in this handler may use different middleware chains.',
+                } : {}),
+                consumers,
+                ...(mismatches.length > 0 ? { mismatches } : {}),
+                executionFlows: flows,
+                impactSummary: {
+                    directConsumers: consumerCount,
+                    affectedFlows: flows.length,
+                    riskLevel,
+                    ...(warning ? { warning } : {}),
+                },
+            };
+        });
+        // If a single route was targeted, return it directly (not wrapped in array)
+        if (results.length === 1) {
+            return results[0];
+        }
+        return { routes: results, total: results.length };
+    }
     // ─── Direct Graph Queries (for resources.ts) ────────────────────
     /**
      * Query clusters (communities) directly from graph.