@comfanion/usethis_search 4.4.0 → 4.5.1

package/tools/codeindex.ts

@@ -9,7 +9,7 @@ import { tool } from "@opencode-ai/plugin"
 import path from "path"
 import fs from "fs/promises"
 
-import { CodebaseIndexer } from "../vectorizer/index.ts"
+import { CodebaseIndexer, getIndexer, releaseIndexer, destroyIndexer } from "../vectorizer/index.ts"
 
 const INDEX_EXTENSIONS: Record<string, string[]> = {
   code: [".js", ".ts", ".jsx", ".tsx", ".go", ".py", ".rs", ".java", ".kt", ".swift", ".c", ".cpp", ".h", ".cs", ".rb", ".php"],
@@ -98,14 +98,17 @@ Available indexes:
         output += `### Active Indexes\n\n`
         for (const idx of indexes) {
           try {
-            const indexer = await new CodebaseIndexer(projectRoot, idx).init()
-            const stats = await indexer.getStats()
-            await indexer.unloadModel()
-            const desc = INDEX_DESCRIPTIONS[idx] || "Custom index"
-            const features = stats.features
-              ? ` | chunking: ${stats.features.chunking}, hybrid: ${stats.features.hybrid ? "on" : "off"}`
-              : ""
-            output += `- **${idx}** - ${desc} (files: ${stats.fileCount}, chunks: ${stats.chunkCount}${features})\n`
+            const indexer = await getIndexer(projectRoot, idx)
+            try {
+              const stats = await indexer.getStats()
+              const desc = INDEX_DESCRIPTIONS[idx] || "Custom index"
+              const features = stats.features
+                ? ` | chunking: ${stats.features.chunking}, hybrid: ${stats.features.hybrid ? "on" : "off"}`
+                : ""
+              output += `- **${idx}** - ${desc} (files: ${stats.fileCount}, chunks: ${stats.chunkCount}${features})\n`
+            } finally {
+              releaseIndexer(projectRoot, idx)
+            }
           } catch {
             output += `- ${idx}\n`
           }
@@ -119,41 +122,44 @@ Available indexes:
     if (args.action === "status") {
       const hashesFile = path.join(vectorsDir, indexName, "hashes.json")
       try {
-        const indexer = await new CodebaseIndexer(projectRoot, indexName).init()
-        const stats = await indexer.getStats()
-        await indexer.unloadModel()
-
-        const sampleFiles = Object.keys(JSON.parse(await fs.readFile(hashesFile, "utf8"))).slice(0, 5)
-        const desc = INDEX_DESCRIPTIONS[indexName] || "Custom index"
-
-        let output = `## Index Status: "${indexName}"\n\n`
-        output += `**Description:** ${desc}\n`
-        output += `**Files indexed:** ${stats.fileCount}\n`
-        output += `**Total chunks:** ${stats.chunkCount}\n`
-        output += `**Model:** ${stats.model}\n`
-
-        if (stats.features) {
-          output += `\n**Features:**\n`
-          output += `- Chunking strategy: ${stats.features.chunking}\n`
-          output += `- Hybrid search: ${stats.features.hybrid ? "enabled" : "disabled"}\n`
-          output += `- Metrics: ${stats.features.metrics ? "enabled" : "disabled"}\n`
-          output += `- Query cache: ${stats.features.cache ? "enabled" : "disabled"}\n`
-        }
-
-        // Show metrics summary if available
+        const indexer = await getIndexer(projectRoot, indexName)
         try {
-          const metrics = await indexer.getMetrics()
-          if (metrics.total_queries > 0) {
-            output += `\n**Search Metrics:**\n`
-            output += `- Total queries: ${metrics.total_queries}\n`
-            output += `- Avg results/query: ${metrics.avg_results_per_query.toFixed(1)}\n`
-            output += `- Zero results rate: ${(metrics.zero_results_rate * 100).toFixed(1)}%\n`
-            output += `- Avg relevance: ${metrics.avg_relevance.toFixed(3)}\n`
+          const stats = await indexer.getStats()
+
+          const sampleFiles = Object.keys(JSON.parse(await fs.readFile(hashesFile, "utf8"))).slice(0, 5)
+          const desc = INDEX_DESCRIPTIONS[indexName] || "Custom index"
+
+          let output = `## Index Status: "${indexName}"\n\n`
+          output += `**Description:** ${desc}\n`
+          output += `**Files indexed:** ${stats.fileCount}\n`
+          output += `**Total chunks:** ${stats.chunkCount}\n`
+          output += `**Model:** ${stats.model}\n`
+
+          if (stats.features) {
+            output += `\n**Features:**\n`
+            output += `- Chunking strategy: ${stats.features.chunking}\n`
+            output += `- Hybrid search: ${stats.features.hybrid ? "enabled" : "disabled"}\n`
+            output += `- Metrics: ${stats.features.metrics ? "enabled" : "disabled"}\n`
+            output += `- Query cache: ${stats.features.cache ? "enabled" : "disabled"}\n`
           }
-        } catch {}
 
-        output += `\n**Sample indexed files:**\n${sampleFiles.map((f) => `- ${f}`).join("\n")}${stats.fileCount > 5 ? `\n- ... and ${stats.fileCount - 5} more` : ""}`
-        return output
+          // Show metrics summary if available
+          try {
+            const metrics = await indexer.getMetrics()
+            if (metrics.total_queries > 0) {
+              output += `\n**Search Metrics:**\n`
+              output += `- Total queries: ${metrics.total_queries}\n`
+              output += `- Avg results/query: ${metrics.avg_results_per_query.toFixed(1)}\n`
+              output += `- Zero results rate: ${(metrics.zero_results_rate * 100).toFixed(1)}%\n`
+              output += `- Avg relevance: ${metrics.avg_relevance.toFixed(3)}\n`
+            }
+          } catch {}
+
+          output += `\n**Sample indexed files:**\n${sampleFiles.map((f) => `- ${f}`).join("\n")}${stats.fileCount > 5 ? `\n- ... and ${stats.fileCount - 5} more` : ""}`
+          return output
+        } finally {
+          releaseIndexer(projectRoot, indexName)
+        }
       } catch {
         return `## Index Status: "${indexName}"\n\nIndex "${indexName}" not created yet. Create it with: codeindex({ action: "reindex", index: "${indexName}" })`
       }
@@ -161,6 +167,8 @@ Available indexes:
 
     if (args.action === "reindex") {
       try {
+        // Destroy any pooled instance to get fresh state for reindex
+        await destroyIndexer(projectRoot, indexName)
         const indexer = await new CodebaseIndexer(projectRoot, indexName).init()
 
         const baseDir = args.dir ? path.resolve(projectRoot, args.dir) : projectRoot
@@ -195,8 +203,8 @@ Available indexes:
           } catch {}
         }
 
-        await indexer.unloadModel()
         const stats = await indexer.getStats()
+        await indexer.unloadModel()
 
         let output = `## Re-indexing Complete\n\n`
         output += `**Index:** ${indexName}\n`
@@ -258,41 +266,43 @@ Available indexes:
           return `## Gold Dataset Test\n\nNo test queries found in gold dataset.`
         }
 
-        const indexer = await new CodebaseIndexer(projectRoot, indexName).init()
-        let passed = 0
-        let failed = 0
-        let output = `## Gold Dataset Test Results\n\n`
-
-        for (const t of tests) {
-          const results = await indexer.search(t.query, 10, false)
-          const foundFiles = results.map((r: any) => r.file)
-          const foundExpected = t.expected_files.filter(f => foundFiles.includes(f))
-          const topScore = results.length > 0 && results[0]._distance != null
-            ? 1 - results[0]._distance
-            : 0
-
-          const pass = foundExpected.length >= Math.ceil(t.expected_files.length * 0.5) && topScore >= t.min_relevance
-
-          if (pass) {
-            passed++
-            output += `**PASS** Query: "${t.query}"\n`
-          } else {
-            failed++
-            output += `**FAIL** Query: "${t.query}"\n`
-          }
+        const indexer = await getIndexer(projectRoot, indexName)
+        try {
+          let passed = 0
+          let failed = 0
+          let output = `## Gold Dataset Test Results\n\n`
+
+          for (const t of tests) {
+            const results = await indexer.search(t.query, 10, false)
+            const foundFiles = results.map((r: any) => r.file)
+            const foundExpected = t.expected_files.filter(f => foundFiles.includes(f))
+            const topScore = results.length > 0 && results[0]._distance != null
+              ? 1 - results[0]._distance
+              : 0
+
+            const pass = foundExpected.length >= Math.ceil(t.expected_files.length * 0.5) && topScore >= t.min_relevance
+
+            if (pass) {
+              passed++
+              output += `**PASS** Query: "${t.query}"\n`
+            } else {
+              failed++
+              output += `**FAIL** Query: "${t.query}"\n`
+            }
 
-          output += `  Found: ${foundFiles.slice(0, 3).map((f: string) => `${f} (${(1 - (results.find((r: any) => r.file === f)?._distance ?? 1)).toFixed(2)})`).join(", ")}\n`
-          if (foundExpected.length < t.expected_files.length) {
-            const missing = t.expected_files.filter(f => !foundFiles.includes(f))
-            output += `  Missing: ${missing.join(", ")}\n`
+            output += `  Found: ${foundFiles.slice(0, 3).map((f: string) => `${f} (${(1 - (results.find((r: any) => r.file === f)?._distance ?? 1)).toFixed(2)})`).join(", ")}\n`
+            if (foundExpected.length < t.expected_files.length) {
+              const missing = t.expected_files.filter(f => !foundFiles.includes(f))
+              output += `  Missing: ${missing.join(", ")}\n`
+            }
+            output += `\n`
           }
-          output += `\n`
-        }
 
-        await indexer.unloadModel()
-
-        output += `---\n**Summary:** ${passed}/${tests.length} tests passed (${Math.round(passed / tests.length * 100)}%)\n`
-        return output
+          output += `---\n**Summary:** ${passed}/${tests.length} tests passed (${Math.round(passed / tests.length * 100)}%)\n`
+          return output
+        } finally {
+          releaseIndexer(projectRoot, indexName)
+        }
       } catch (error: any) {
         return `Gold dataset test failed: ${error.message || String(error)}`
       }
@@ -301,139 +311,137 @@ Available indexes:
     // NFR-031: Graph validation
     if (args.action === "validate-graph") {
       try {
-        const indexer = await new CodebaseIndexer(projectRoot, indexName).init()
-
-        // Access internal graphDB and db
-        const graphDB = (indexer as any).graphDB
-        const db = (indexer as any).db
-
-        if (!graphDB) {
-          await indexer.unloadModel()
-          return `## Graph Validation: "${indexName}"\n\nNo graph database found. Run reindex first.`
-        }
-
-        // 1. Get all triples from graph
-        let allTriples: any[] = []
+        const indexer = await getIndexer(projectRoot, indexName)
         try {
-          allTriples = await graphDB.getAllTriples()
-        } catch (e: any) {
-          await indexer.unloadModel()
-          return `## Graph Validation: "${indexName}"\n\n**Error:** Failed to read graph database: ${e.message || String(e)}\n\nThe graph database may be corrupted. Run: codeindex({ action: "reindex", index: "${indexName}" })`
-        }
+          // Access internal graphDB and db
+          const graphDB = (indexer as any).graphDB
+          const db = (indexer as any).db
 
-        // 2. Get all chunk IDs from vector DB
-        const knownChunkIds = new Set<string>()
-        const tables = await db.tableNames()
-        if (tables.includes("chunks")) {
-          const table = await db.openTable("chunks")
+          if (!graphDB) {
+            return `## Graph Validation: "${indexName}"\n\nNo graph database found. Run reindex first.`
+          }
+
+          // 1. Get all triples from graph
+          let allTriples: any[] = []
           try {
-            const rows = await table.filter("").limit(100000).execute()
-            for (const row of rows) {
-              if (row.chunk_id) knownChunkIds.add(row.chunk_id)
-            }
+            allTriples = await graphDB.getAllTriples()
           } catch (e: any) {
-            await indexer.unloadModel()
-            return `## Graph Validation: "${indexName}"\n\n**Error:** Failed to read vector database: ${e.message || String(e)}\n\nThe vector database may be corrupted. Run: codeindex({ action: "reindex", index: "${indexName}" })`
+            return `## Graph Validation: "${indexName}"\n\n**Error:** Failed to read graph database: ${e.message || String(e)}\n\nThe graph database may be corrupted. Run: codeindex({ action: "reindex", index: "${indexName}" })`
           }
-        }
 
-        // 3. Validate: find orphaned triples (subject or object points to non-existent chunk)
-        const orphanedSubjects: Array<{ triple: string; missingId: string }> = []
-        const orphanedObjects: Array<{ triple: string; missingId: string }> = []
-        const predicateCounts: Record<string, number> = {}
-        const sourceCounts: Record<string, number> = {}
-        const fileCounts: Record<string, number> = {}
-
-        for (const t of allTriples) {
-          // Count predicates/sources
-          predicateCounts[t.predicate] = (predicateCounts[t.predicate] || 0) + 1
-          sourceCounts[t.source] = (sourceCounts[t.source] || 0) + 1
-          fileCounts[t.file] = (fileCounts[t.file] || 0) + 1
-
-          // Check subject (skip meta: prefixed subjects)
-          if (!t.subject.startsWith("meta:") && t.subject.startsWith("chunk_") && !knownChunkIds.has(t.subject)) {
-            orphanedSubjects.push({
-              triple: `${t.subject} --[${t.predicate}]--> ${t.object}`,
-              missingId: t.subject,
-            })
+          // 2. Get all chunk IDs from vector DB
+          const knownChunkIds = new Set<string>()
+          const tables = await db.tableNames()
+          if (tables.includes("chunks")) {
+            const table = await db.openTable("chunks")
+            try {
+              const rows = await table.filter("").limit(100000).execute()
+              for (const row of rows) {
+                if (row.chunk_id) knownChunkIds.add(row.chunk_id)
+              }
+            } catch (e: any) {
+              return `## Graph Validation: "${indexName}"\n\n**Error:** Failed to read vector database: ${e.message || String(e)}\n\nThe vector database may be corrupted. Run: codeindex({ action: "reindex", index: "${indexName}" })`
+            }
           }
 
-          // Check object (skip non-chunk objects like file paths, hashes)
-          if (t.object.startsWith("chunk_") && !knownChunkIds.has(t.object)) {
-            orphanedObjects.push({
-              triple: `${t.subject} --[${t.predicate}]--> ${t.object}`,
-              missingId: t.object,
-            })
-          }
-        }
+          // 3. Validate: find orphaned triples (subject or object points to non-existent chunk)
+          const orphanedSubjects: Array<{ triple: string; missingId: string }> = []
+          const orphanedObjects: Array<{ triple: string; missingId: string }> = []
+          const predicateCounts: Record<string, number> = {}
+          const sourceCounts: Record<string, number> = {}
+          const fileCounts: Record<string, number> = {}
+
+          for (const t of allTriples) {
+            // Count predicates/sources
+            predicateCounts[t.predicate] = (predicateCounts[t.predicate] || 0) + 1
+            sourceCounts[t.source] = (sourceCounts[t.source] || 0) + 1
+            fileCounts[t.file] = (fileCounts[t.file] || 0) + 1
+
+            // Check subject (skip meta: prefixed subjects)
+            if (!t.subject.startsWith("meta:") && t.subject.startsWith("chunk_") && !knownChunkIds.has(t.subject)) {
+              orphanedSubjects.push({
+                triple: `${t.subject} --[${t.predicate}]--> ${t.object}`,
+                missingId: t.subject,
+              })
+            }
 
-        // 4. Get file metadata stats
-        let fileMeta: Array<{ filePath: string; hash: string; timestamp: number }> = []
-        try {
-          fileMeta = await graphDB.getAllFileMeta()
-        } catch (e: any) {
-          // Non-fatal - continue validation without metadata
-          console.warn(`Warning: Failed to get file metadata: ${e.message || String(e)}`)
-        }
+            // Check object (skip non-chunk objects like file paths, hashes)
+            if (t.object.startsWith("chunk_") && !knownChunkIds.has(t.object)) {
+              orphanedObjects.push({
+                triple: `${t.subject} --[${t.predicate}]--> ${t.object}`,
+                missingId: t.object,
+              })
+            }
+          }
 
-        await indexer.unloadModel()
+          // 4. Get file metadata stats
+          let fileMeta: Array<{ filePath: string; hash: string; timestamp: number }> = []
+          try {
+            fileMeta = await graphDB.getAllFileMeta()
+          } catch (e: any) {
+            // Non-fatal - continue validation without metadata
+            console.warn(`Warning: Failed to get file metadata: ${e.message || String(e)}`)
+          }
 
-        // 5. Build report
-        const totalOrphaned = orphanedSubjects.length + orphanedObjects.length
-        const isHealthy = totalOrphaned === 0
+          // 5. Build report
+          const totalOrphaned = orphanedSubjects.length + orphanedObjects.length
+          const isHealthy = totalOrphaned === 0
 
-        let output = `## Graph Validation: "${indexName}"\n\n`
-        output += `**Status:** ${isHealthy ? "HEALTHY" : "ISSUES FOUND"}\n\n`
+          let output = `## Graph Validation: "${indexName}"\n\n`
+          output += `**Status:** ${isHealthy ? "HEALTHY" : "ISSUES FOUND"}\n\n`
 
-        output += `### Statistics\n`
-        output += `- **Total triples:** ${allTriples.length}\n`
-        output += `- **Known chunk IDs:** ${knownChunkIds.size}\n`
-        output += `- **Files with graph metadata:** ${fileMeta.length}\n`
-        output += `- **Unique files in graph:** ${Object.keys(fileCounts).length}\n\n`
+          output += `### Statistics\n`
+          output += `- **Total triples:** ${allTriples.length}\n`
+          output += `- **Known chunk IDs:** ${knownChunkIds.size}\n`
+          output += `- **Files with graph metadata:** ${fileMeta.length}\n`
+          output += `- **Unique files in graph:** ${Object.keys(fileCounts).length}\n\n`
 
-        output += `### Edge Types\n`
-        for (const [pred, count] of Object.entries(predicateCounts).sort((a, b) => b[1] - a[1])) {
-          output += `- **${pred}:** ${count}\n`
-        }
-        output += `\n`
+          output += `### Edge Types\n`
+          for (const [pred, count] of Object.entries(predicateCounts).sort((a, b) => b[1] - a[1])) {
+            output += `- **${pred}:** ${count}\n`
+          }
+          output += `\n`
 
-        output += `### Edge Sources\n`
-        for (const [source, count] of Object.entries(sourceCounts).sort((a, b) => b[1] - a[1])) {
-          output += `- **${source}:** ${count}\n`
-        }
-        output += `\n`
+          output += `### Edge Sources\n`
+          for (const [source, count] of Object.entries(sourceCounts).sort((a, b) => b[1] - a[1])) {
+            output += `- **${source}:** ${count}\n`
+          }
+          output += `\n`
 
-        if (totalOrphaned > 0) {
-          output += `### Orphaned References (${totalOrphaned})\n\n`
+          if (totalOrphaned > 0) {
+            output += `### Orphaned References (${totalOrphaned})\n\n`
 
-          if (orphanedSubjects.length > 0) {
-            output += `**Broken subjects** (${orphanedSubjects.length}):\n`
-            for (const o of orphanedSubjects.slice(0, 10)) {
-              output += `- \`${o.missingId}\` in: ${o.triple}\n`
-            }
-            if (orphanedSubjects.length > 10) {
-              output += `- ... and ${orphanedSubjects.length - 10} more\n`
+            if (orphanedSubjects.length > 0) {
+              output += `**Broken subjects** (${orphanedSubjects.length}):\n`
+              for (const o of orphanedSubjects.slice(0, 10)) {
+                output += `- \`${o.missingId}\` in: ${o.triple}\n`
+              }
+              if (orphanedSubjects.length > 10) {
+                output += `- ... and ${orphanedSubjects.length - 10} more\n`
+              }
+              output += `\n`
             }
-            output += `\n`
-          }
 
-          if (orphanedObjects.length > 0) {
-            output += `**Broken objects** (${orphanedObjects.length}):\n`
-            for (const o of orphanedObjects.slice(0, 10)) {
-              output += `- \`${o.missingId}\` in: ${o.triple}\n`
-            }
-            if (orphanedObjects.length > 10) {
-              output += `- ... and ${orphanedObjects.length - 10} more\n`
+            if (orphanedObjects.length > 0) {
+              output += `**Broken objects** (${orphanedObjects.length}):\n`
+              for (const o of orphanedObjects.slice(0, 10)) {
+                output += `- \`${o.missingId}\` in: ${o.triple}\n`
+              }
+              if (orphanedObjects.length > 10) {
+                output += `- ... and ${orphanedObjects.length - 10} more\n`
+              }
+              output += `\n`
             }
-            output += `\n`
+
+            output += `**Recommendation:** Run \`codeindex({ action: "reindex", index: "${indexName}" })\` to rebuild the graph.\n`
+          } else {
+            output += `### Integrity\nAll chunk references are valid. No orphaned triples found.\n`
           }
 
-          output += `**Recommendation:** Run \`codeindex({ action: "reindex", index: "${indexName}" })\` to rebuild the graph.\n`
-        } else {
-          output += `### Integrity\nAll chunk references are valid. No orphaned triples found.\n`
+          return output
+        } finally {
+          releaseIndexer(projectRoot, indexName)
         }
-
-        return output
       } catch (error: any) {
         return `Graph validation failed: ${error.message || String(error)}`
       }