scai 0.1.117 → 0.1.119

package/dist/daemon/runFolderCapsuleBatch.js

@@ -0,0 +1,247 @@
+import fg from 'fast-glob';
+import path from 'path';
+import { getDbForRepo } from '../db/client.js';
+import { log } from '../utils/log.js';
+import { generate } from '../lib/generate.js';
+import { cleanupModule } from '../pipeline/modules/cleanupModule.js';
+import { IGNORED_FOLDER_GLOBS } from '../fileRules/ignoredPaths.js';
+export async function runFolderCapsuleBatch(maxFolders = 20) {
+    const db = getDbForRepo();
+    log('📦 Starting folder capsule batch...');
+    // --------------------------------------------------
+    // Stop if we already have enough capsules
+    // --------------------------------------------------
+    const existing = db.prepare(`SELECT COUNT(*) as c FROM folder_capsules`).get();
+    log(`📦 Existing folder capsules: ${existing.c}`);
+    if (existing.c >= maxFolders) {
+        log(`📦 Folder capsule batch: cap reached (${existing.c})`);
+        return false;
+    }
+    // --------------------------------------------------
+    // Scan folders with fast-glob
+    // --------------------------------------------------
+    log('🔍 Scanning folders...');
+    const folderPathsRaw = await fg('**/', {
+        cwd: process.cwd(),
+        onlyDirectories: true,
+        ignore: IGNORED_FOLDER_GLOBS,
+        dot: false,
+        absolute: true,
+    });
+    const folderPaths = folderPathsRaw.map(p => path.normalize(p).replace(/\\/g, '/'));
+    log(`📂 Found ${folderPaths.length} folders`);
+    if (!folderPaths.length) {
+        log('📦 No folders found after filtering');
+        return false;
+    }
+    // --------------------------------------------------
+    // Collect uncapsuled folders from DB
+    // --------------------------------------------------
+    log('📦 Collecting uncapsuled folders from DB...');
+    const uncapsuledFoldersRaw = db.prepare(`
+        SELECT folder, COUNT(*) AS fileCount
+        FROM (
+            SELECT substr(path, 1, length(path) - length(filename) - 1) AS folder,
+                   filename, type
+            FROM files
+            WHERE processing_status != 'skipped'
+        )
+        LEFT JOIN folder_capsules fc ON fc.path = folder
+        WHERE fc.path IS NULL
+        GROUP BY folder
+    `).all();
+    const uncapsuledFolders = uncapsuledFoldersRaw
+        .map(f => ({ ...f, folder: path.normalize(path.resolve(f.folder)).replace(/\\/g, '/') }))
+        .filter(f => folderPaths.includes(f.folder));
+    log(`📦 ${uncapsuledFolders.length} uncapsuled folders match actual paths`);
+    if (!uncapsuledFolders.length) {
+        log('📦 No uncapsuled folders found');
+        return false;
+    }
+    // --------------------------------------------------
+    // Heuristic folder scoring
+    // --------------------------------------------------
+    function scoreFolder(folder, fileCount) {
+        let score = 0;
+        const depth = folder.split('/').length;
+        const p = folder.toLowerCase();
+        if (fileCount > 5)
+            score += 2;
+        if (fileCount > 15)
+            score += 2;
+        if (depth <= 6)
+            score += 2;
+        if (/src|modules|pipeline|agents|services|commands|lib/.test(p))
+            score += 3;
+        if (/css|sql|html|assets/.test(p))
+            score -= 3;
+        return score;
+    }
+    const ranked = uncapsuledFolders
+        .map(f => ({ ...f, score: scoreFolder(f.folder, f.fileCount) }))
+        .sort((a, b) => b.score - a.score)
+        .slice(0, maxFolders);
+    log(`📦 Top ${ranked.length} folders selected for processing`);
+    // --------------------------------------------------
+    // Process each top folder with try/catch
+    // --------------------------------------------------
+    for (const target of ranked) {
+        const folderPath = target.folder;
+        if (!folderPath) {
+            log(`⚠️ Skipping folder because folderPath is undefined`);
+            continue;
+        }
+        try {
+            log(`📦 Building folder capsule: ${folderPath} (score=${target.score})`);
+            // --------------------------------------------------
+            // Load files in folder from DB (inline folderPath)
+            // --------------------------------------------------
+            const folderPathSQL = folderPath.replace(/'/g, "''"); // escape single quotes
+            const files = db.prepare(`
+                SELECT path, filename, type
+                FROM files
+                WHERE path LIKE '${folderPathSQL}/%' AND processing_status != 'skipped'
+            `).all();
+            log(`📄 ${files.length} files found in folder`);
+            if (!files.length) {
+                log('⚠️ Folder empty, skipping');
+                continue;
+            }
+            // --------------------------------------------------
+            // Compute file-level scores
+            // --------------------------------------------------
+            const fileScores = files.map(f => {
+                const heuristicName = /index|main|cli|app|server|config/i.test(f.filename) ? 5 : 0;
+                const basename = path.basename(f.path).replace(/'/g, "''");
+                const incoming = db.prepare(`
+                    SELECT COUNT(*) AS c
+                    FROM files
+                    WHERE path LIKE '${folderPathSQL}/%' AND processing_status != 'skipped' AND content_text LIKE '%${basename}%'
+                `).get();
+                const incomingScore = (incoming?.c || 0) * 2;
+                const row = db.prepare(`SELECT content_text FROM files_fts WHERE path = '${f.path.replace(/'/g, "''")}'`).get();
+                const code = row?.contentText || '';
+                const outgoingCount = (code.match(/from\s+['"].+['"]/g)?.length || 0) + (code.match(/require\(['"].+['"]\)/g)?.length || 0);
+                const sizeScore = Math.min(code.length / 2000, 3);
+                const totalScore = heuristicName + incomingScore + outgoingCount + sizeScore;
+                log(`📄 File score: ${f.filename} -> ${totalScore}`);
+                return { ...f, score: totalScore };
+            });
+            const representativeFiles = fileScores.sort((a, b) => b.score - a.score).slice(0, 2);
+            log(`📌 Representative files: ${representativeFiles.map(f => f.filename).join(', ')}`);
+            const candidateFiles = representativeFiles.map(f => {
+                const row = db.prepare(`SELECT content_text FROM files_fts WHERE path = '${f.path.replace(/'/g, "''")}'`).get();
+                return { path: f.path, code: row?.contentText?.slice(0, 2000) || '' };
+            });
+            // --------------------------------------------------
+            // Ask LLM to generate folder summary
+            // --------------------------------------------------
+            const prompt = `
+You are analyzing a source code folder.
+
+Folder path:
+${folderPath}
+
+Candidate files with code snippets:
+${JSON.stringify(candidateFiles, null, 2)}
+
+Task:
+- Identify up to TWO files that best represent the purpose of this folder.
+- Explain the folder's responsibility in ONE concise sentence.
+- Return ONLY valid JSON.
+
+Expected JSON shape:
+{
+  "summary": "one sentence description",
+  "files": [
+    { "path": "absolute/file/path", "summary": "optional short note" }
+  ]
+}
+`.trim();
+            let folderSummary = '';
+            let keyFiles = [];
+            try {
+                log('🤖 Asking LLM for folder summary...');
+                const response = await generate({ content: prompt, query: '' });
+                const cleaned = await cleanupModule.run({ query: '', content: response.data });
+                const data = typeof cleaned.data === 'string' ? JSON.parse(cleaned.data) : cleaned.data;
+                if (data && typeof data === 'object') {
+                    if (typeof data.summary === 'string')
+                        folderSummary = data.summary;
+                    if (Array.isArray(data.files)) {
+                        keyFiles = data.files
+                            .filter((f) => typeof f?.path === 'string')
+                            .slice(0, 2)
+                            .map((f) => ({ path: f.path, reason: f.summary || f.reason || 'representative file' }));
+                    }
+                }
+                log(`📌 Folder summary: ${folderSummary}`);
+            }
+            catch (err) {
+                if (err instanceof Error) {
+                    log(`🔥 Failed processing folder ${folderPath}: ${err.message}\n${err.stack}`);
+                }
+                else {
+                    log(`🔥 Failed processing folder ${folderPath}:`, err);
+                }
+            }
+            // --------------------------------------------------
+            // Compute stats by file type
+            // --------------------------------------------------
+            const byType = {};
+            for (const f of files) {
+                const ext = f.type || path.extname(f.filename || '').replace('.', '') || 'unknown';
+                byType[ext] = (byType[ext] || 0) + 1;
+            }
+            // --------------------------------------------------
+            // Folder-level dependencies
+            // --------------------------------------------------
+            const importsFrom = new Set();
+            for (const f of files) {
+                const dir = path.dirname(f.path).replace(/\\/g, '/');
+                if (dir !== folderPath)
+                    importsFrom.add(dir);
+            }
+            // --------------------------------------------------
+            // Build capsule
+            // --------------------------------------------------
+            const capsule = {
+                path: folderPath,
+                depth: folderPath.split('/').length,
+                stats: { fileCount: files.length, byType },
+                roles: [],
+                concerns: [],
+                keyFiles,
+                dependencies: { importsFrom: Array.from(importsFrom), usedBy: [] },
+                summary: folderSummary,
+                confidence: Math.min(0.9, 0.4 + target.score * 0.1),
+            };
+            // --------------------------------------------------
+            // Persist capsule (inline folderPath)
+            // --------------------------------------------------
+            const now = new Date().toISOString();
+            db.prepare(`
+                INSERT INTO folder_capsules (
+                    path, depth, capsule_json, confidence, last_generated, source_file_count
+                )
+                VALUES ('${capsule.path.replace(/'/g, "''")}', ${capsule.depth}, '${JSON.stringify(capsule)}', ${capsule.confidence}, '${now}', ${files.length})
+            `).run();
+            db.prepare(`
+                UPDATE files
+                SET processing_status = 'capsuled'
+                WHERE path LIKE '${folderPathSQL}/%'
+            `).run();
+            log(`✅ Folder capsule written: ${folderPath}`);
+        }
+        catch (err) {
+            if (err instanceof Error) {
+                log(`🔥 Failed processing folder ${folderPath}: ${err.message}\n${err.stack}`);
+            }
+            else {
+                log(`🔥 Failed processing folder ${folderPath}:`, err);
+            }
+        }
+    }
+    log('📦 Folder capsule batch complete.');
+    return true;
+}

package/dist/daemon/runIndexingBatch.js

@@ -0,0 +1,147 @@
+import fs from 'fs';
+import path from 'path';
+import lockfile from 'proper-lockfile';
+import { log } from '../utils/log.js';
+import { getDbForRepo, getDbPathForRepo } from '../db/client.js';
+import { sleep } from '../utils/sleep.js';
+import { selectUnprocessedFiles, markFileAsSkippedByPath, countUnprocessedFiles, upsertFileTemplate, upsertFileFtsTemplate, markFileAsIndexed } from '../db/sqlTemplates.js';
+// --------------------------------------------------
+// DB LOCK
+// --------------------------------------------------
+async function lockDbWithRetry(retries = 3, delayMs = 100) {
+    for (let i = 0; i < retries; i++) {
+        try {
+            return await lockfile.lock(getDbPathForRepo());
+        }
+        catch (err) {
+            if (i < retries - 1) {
+                log(`⏳ DB lock busy, retrying... (${i + 1})`);
+                await sleep(delayMs);
+            }
+            else {
+                log('❌ Failed to acquire DB lock after retries:', err);
+                throw err;
+            }
+        }
+    }
+}
+// --------------------------------------------------
+// SINGLE FILE INDEXER (daemon-side)
+// --------------------------------------------------
+export function indexFile(filePath, summary, type) {
+    const stats = fs.statSync(filePath);
+    const lastModified = stats.mtime.toISOString();
+    const indexedAt = new Date().toISOString();
+    const normalizedPath = path.normalize(filePath).replace(/\\/g, '/');
+    const filename = path.basename(normalizedPath);
+    // ----------------------------------------------
+    // Extract text content (guarded)
+    // ----------------------------------------------
+    let contentText = '';
+    try {
+        if (stats.size <= 2000000) {
+            const buffer = fs.readFileSync(filePath);
+            const nonTextRatio = buffer
+                .slice(0, 1000)
+                .filter(b => b < 9 || (b > 13 && b < 32))
+                .length / Math.min(buffer.length, 1000);
+            if (nonTextRatio <= 0.3) {
+                contentText = buffer.toString('utf-8');
+            }
+            else {
+                log(`⚠️ Binary-like content skipped: ${normalizedPath}`);
+            }
+        }
+        else {
+            log(`⚠️ Large file content skipped: ${normalizedPath}`);
+        }
+    }
+    catch (err) {
+        log(`⚠️ Failed reading content: ${normalizedPath}`, err);
+    }
+    const db = getDbForRepo();
+    // ----------------------------------------------
+    // Metadata upsert
+    // ----------------------------------------------
+    try {
+        db.prepare(upsertFileTemplate).run({
+            path: normalizedPath,
+            filename,
+            summary,
+            type,
+            lastModified,
+            indexedAt,
+        });
+    }
+    catch (err) {
+        log(`⚠️ Failed metadata upsert for ${normalizedPath}:`, err);
+    }
+    // ----------------------------------------------
+    // FTS upsert
+    // ----------------------------------------------
+    db.prepare(upsertFileFtsTemplate).run({
+        path: normalizedPath,
+        filename,
+        summary,
+        contentText,
+    });
+    // ----------------------------------------------
+    // Mark as indexed
+    // ----------------------------------------------
+    db.prepare(markFileAsIndexed).run({ path: normalizedPath });
+}
+// --------------------------------------------------
+// FTS REBUILD
+// --------------------------------------------------
+function rebuildFts() {
+    const db = getDbForRepo();
+    log('🔍 Rebuilding FTS index...');
+    db.exec(`INSERT INTO files_fts(files_fts) VALUES('rebuild');`);
+}
+// --------------------------------------------------
+// INDEXING BATCH
+// --------------------------------------------------
+export async function runIndexingBatch() {
+    log('⚡ Starting indexing batch...');
+    const db = getDbForRepo();
+    const BATCH_SIZE = 25; // adjust as needed
+    const rows = db.prepare(selectUnprocessedFiles).all(BATCH_SIZE);
+    if (rows.length === 0) {
+        log('✅ No files left to index.');
+        return false;
+    }
+    const release = await lockDbWithRetry();
+    let didIndexWork = false;
+    log('Release: ', release);
+    try {
+        for (const row of rows) {
+            log(`📄 Indexing: ${row.path}`);
+            try {
+                indexFile(row.path, null, 'auto');
+                didIndexWork = true;
+            }
+            catch (err) {
+                log(`⚠️ Failed indexing ${row.path}`, err);
+                db.prepare(markFileAsSkippedByPath).run({ path: row.path });
+            }
+        }
+        try {
+            rebuildFts();
+        }
+        catch (err) {
+            log('⚠️ Failed FTS rebuild:', err);
+        }
+        const remaining = db.prepare(countUnprocessedFiles).get();
+        log(`📦 Remaining unindexed files: ${remaining.count}`);
+        return didIndexWork;
+    }
+    finally {
+        if (release) {
+            await release();
+            log('🔓 DB lock released');
+        }
+        else {
+            log('⚠️ DB lock was not acquired, nothing to release');
+        }
+    }
+}

package/dist/daemon/runKgBatch.js

@@ -0,0 +1,104 @@
+// File: src/daemon/runKgBatch.ts
+import fs from 'fs/promises';
+import fsSync from 'fs';
+import path from 'path';
+import { insertGraphTagTemplate, selectGraphTagIdTemplate, insertGraphEntityTagTemplate, } from '../db/sqlTemplates.js';
+import { getDbForRepo } from '../db/client.js';
+import { log } from '../utils/log.js';
+import { kgModule } from '../pipeline/modules/kgModule.js';
+import { KG_IGNORED_EXTENSIONS } from '../fileRules/ignoredExtensions.js';
+import * as sqlTemplates from '../db/sqlTemplates.js'; // import the template
+import { indexCodeForFile } from '../db/functionIndex.js';
+const MAX_KG_FILES_PER_BATCH = 3;
+export async function runKgBatch() {
+    const db = getDbForRepo();
+    const rows = db.prepare(`
+    SELECT id, path, summary
+    FROM files
+    WHERE functions_extracted_at IS NULL
+      AND processing_status NOT IN ('skipped', 'failed', 'kg_done')
+    LIMIT ?
+  `).all(MAX_KG_FILES_PER_BATCH);
+    if (rows.length === 0) {
+        log('🧠 KG batch: no pending files');
+        return;
+    }
+    for (const row of rows) {
+        log(`\n🔗 KG: Processing ${row.path}`);
+        if (!fsSync.existsSync(row.path)) {
+            log('⚠️ KG skipped: file missing');
+            db.prepare(sqlTemplates.markFileAsKgDone).run({ path: row.path });
+            continue;
+        }
+        // --- CODE EXTRACTION SECTION ---
+        log('\n');
+        log('📄 Code Extraction');
+        log('--------------------------------------------------------------------');
+        try {
+            const success = await indexCodeForFile(row.path, row.id);
+            if (success) {
+                log('✅ Indexed code');
+            }
+            else {
+                log('⚡ No code elements extracted or unsupported file type');
+            }
+        }
+        catch (err) {
+            log(`❌ Code extraction failed for ${row.path}:`, err);
+        }
+        // --- KNOWLEDGE GRAPH SECTION ---
+        log('\n');
+        log('🔗 Knowledge Graph');
+        log('--------------------------------------------------------------------');
+        const ext = path.extname(row.path).toLowerCase();
+        if (KG_IGNORED_EXTENSIONS.includes(ext)) {
+            log(`⚠️ KG skipped (ignored extension): ${ext}`);
+            db.prepare(sqlTemplates.markFileAsKgDone).run({ path: row.path });
+            continue;
+        }
+        try {
+            const content = await fs.readFile(row.path, 'utf-8');
+            const kgInput = {
+                fileId: row.id,
+                filepath: row.path,
+                summary: row.summary || undefined,
+            };
+            const kgResult = await kgModule.run(kgInput, content);
+            log(`✅ KG built: ${kgResult.entities.length} entities, ${kgResult.edges.length} edges`);
+            if (kgResult.entities.length > 0) {
+                const insertTag = db.prepare(insertGraphTagTemplate);
+                const getTagId = db.prepare(selectGraphTagIdTemplate);
+                const insertEntityTag = db.prepare(insertGraphEntityTagTemplate);
+                for (const entity of kgResult.entities) {
+                    if (!entity.type || !Array.isArray(entity.tags))
+                        continue;
+                    for (const tag of entity.tags) {
+                        if (!tag)
+                            continue;
+                        try {
+                            insertTag.run({ name: tag });
+                            const tagRow = getTagId.get({ name: tag });
+                            if (!tagRow)
+                                continue;
+                            insertEntityTag.run({
+                                entity_type: entity.type,
+                                entity_unique_id: `${entity.name}@${row.path}`,
+                                tag_id: tagRow.id,
+                            });
+                        }
+                        catch {
+                            // ignore per-entity failures
+                        }
+                    }
+                }
+            }
+            // ✅ Mark KG as done for this file
+            db.prepare(sqlTemplates.markFileAsKgDone).run({ path: row.path });
+        }
+        catch (err) {
+            log(`❌ KG failed for ${row.path}:`, err);
+            // Still mark as done to avoid infinite retries
+            db.prepare(sqlTemplates.markFileAsKgDone).run({ path: row.path });
+        }
+    }
+}