scai 0.1.40 → 0.1.42

package/dist/commands/AskCmd.js

@@ -1,9 +1,16 @@
-import readline from 'readline';
+import fs from 'fs';
 import path from 'path';
-import { searchFiles, queryFiles } from "../db/fileIndex.js";
-import { sanitizeQueryForFts } from "../utils/sanitizeQuery.js";
-import { generate } from "../lib/generate.js";
+import readline from 'readline';
+import { searchFiles, queryFiles } from '../db/fileIndex.js';
+import { sanitizeQueryForFts } from '../utils/sanitizeQuery.js';
+import { generate } from '../lib/generate.js';
+import { buildContextualPrompt } from '../utils/buildContextualPrompt.js';
+import { generateFileTree } from '../utils/fileTree.js';
+import { log } from '../utils/log.js';
+import { PROMPT_LOG_PATH, SCAI_HOME, INDEX_DIR } from '../constants.js';
+const MAX_RELATED_FILES = 5;
 export async function runAskCommand(query) {
+    // 🧠 Prompt the user if no query is passed
     if (!query) {
         query = await promptOnce('🧠 Ask your question:\n> ');
     }
@@ -12,21 +19,32 @@ export async function runAskCommand(query) {
         console.error('❌ No question provided.\n👉 Usage: scai ask "your question"');
         return;
     }
+    console.log(`📁 Using index root: ${INDEX_DIR}`);
     console.log(`🔍 Searching for: "${query}"\n`);
+    // 🧠 Step 1: Semantic + fallback search
     const start = Date.now();
-    const semanticResults = await searchFiles(query, 5);
+    const semanticResults = await searchFiles(query, MAX_RELATED_FILES);
     const duration = Date.now() - start;
     console.log(`⏱️ searchFiles took ${duration}ms and returned ${semanticResults.length} result(s)`);
-    // Also run fallback keyword search
+    // 🔍 Log raw semantic results
+    console.log('🔍 Raw semantic search results:');
+    semanticResults.forEach((file, i) => {
+        console.log(`  ${i + 1}. 📄 Path: ${file.path} | Score: ${file.score?.toFixed(3) ?? 'n/a'}`);
+    });
     const safeQuery = sanitizeQueryForFts(query);
     const fallbackResults = queryFiles(safeQuery, 10);
-    // Merge semantic and fallback results
+    // 🔍 Log raw keyword fallback results
+    console.log('\n🔍 Raw fallback keyword (FTS) search results:');
+    fallbackResults.forEach((file, i) => {
+        console.log(`  ${i + 1}. 📄 Path: ${file.path}`);
+    });
+    // 🧠 Step 2: Merge results
     const seen = new Set();
     const combinedResults = [];
     for (const file of semanticResults) {
         const resolved = path.resolve(file.path);
         seen.add(resolved);
-        combinedResults.push(file); // Already scored
+        combinedResults.push(file);
     }
     for (const file of fallbackResults) {
         const resolved = path.resolve(file.path);
@@ -35,7 +53,7 @@ export async function runAskCommand(query) {
             combinedResults.push({
                 path: file.path,
                 summary: file.summary,
-                score: 0.0, // fallback score
+                score: 0.0,
             });
         }
     }
@@ -46,26 +64,43 @@ export async function runAskCommand(query) {
         });
     }
     else {
-        console.log('⚠️ No similar files found. Asking the model for context only...');
+        console.log('⚠️ No similar files found. Asking the model using question only...');
     }
-    // Aggregate summaries
-    let allSummaries = '';
-    for (const file of combinedResults) {
-        if (!file?.summary) {
-            console.warn(`⚠️ No summary available for file: ${file?.path}`);
-            continue;
-        }
-        console.log(`📝 Using stored summary for: ${file.path}`);
-        allSummaries += `\n${file.summary}`;
+    // 🧠 Step 3: Build metadata for prompt
+    const relatedFiles = combinedResults.slice(0, MAX_RELATED_FILES).map(file => ({
+        path: file.path,
+        summary: file.summary || '(No summary available)',
+    }));
+    let fileTree = '';
+    try {
+        fileTree = generateFileTree(INDEX_DIR, 2); // Limit depth
+    }
+    catch (e) {
+        console.warn('⚠️ Failed to generate file tree:', e);
+    }
+    const prompt = buildContextualPrompt({
+        baseInstruction: query,
+        code: '', // No specific code selected
+        relatedFiles,
+        projectFileTree: fileTree || undefined,
+    });
+    // 🧠 Step 4: Log prompt to file
+    try {
+        if (!fs.existsSync(SCAI_HOME))
+            fs.mkdirSync(SCAI_HOME, { recursive: true });
+        fs.writeFileSync(PROMPT_LOG_PATH, prompt, 'utf-8');
+        log(`📝 Prompt saved to ${PROMPT_LOG_PATH}`);
+    }
+    catch (err) {
+        log('❌ Failed to write prompt log:', err);
     }
-    const input = {
-        content: allSummaries ? `${query}\n\n${allSummaries}` : query,
-        filepath: '',
-    };
+    // 🧠 Step 5: Call the model
     try {
-        console.log(allSummaries.trim()
-            ? '🧠 Summaries found, sending them to the model for synthesis...'
-            : '⚠️ No summaries found. Asking the model for context only...');
+        console.log('🤖 Asking the model...');
+        const input = {
+            content: prompt,
+            filepath: '',
+        };
         const modelResponse = await generate(input, 'llama3');
         console.log(`\n📝 Model response:\n${modelResponse.content}`);
     }
@@ -77,7 +112,7 @@ function promptOnce(promptText) {
     return new Promise(resolve => {
         const rl = readline.createInterface({
             input: process.stdin,
-            output: process.stdout
+            output: process.stdout,
         });
         rl.question(promptText, answer => {
             rl.close();

package/dist/constants.js

@@ -26,6 +26,11 @@ export const CONFIG_PATH = path.join(SCAI_HOME, 'config.json');
  * ~/.scai/daemon.log
  */
 export const LOG_PATH = path.join(SCAI_HOME, 'daemon.log');
+/**
+ * Path to the last prompt sent to the model:
+ * ~/.scai/prompt.log
+ */
+export const PROMPT_LOG_PATH = path.join(SCAI_HOME, 'prompt.log');
 /**
  * Get the active index directory.
  *

package/dist/daemon/daemonBatch.js

@@ -21,9 +21,9 @@ async function lockDb() {
 export async function runDaemonBatch() {
     log('🟡 Starting daemon batch...');
     const rows = db.prepare(`
-    SELECT path, type, processing_status FROM files
-    WHERE (summary IS NULL OR summary = '') 
-      OR processing_status = 'unprocessed'
+    SELECT id, path, type, summary, indexed_at, last_modified, processing_status
+    FROM files
+    WHERE processing_status = 'unprocessed' OR summary IS NULL OR summary = ''
     ORDER BY last_modified DESC
     LIMIT ?
   `).all(MAX_FILES_PER_BATCH);
@@ -47,32 +47,42 @@ export async function runDaemonBatch() {
         }
         try {
             const content = await fs.readFile(row.path, 'utf-8');
-            const fileResult = db.prepare(`SELECT id FROM files WHERE path = @path`).get({ path: row.path });
-            const fileId = fileResult?.id;
-            if (fileId) {
-                await indexFunctionsForFile(row.path, fileId);
-                // Update processing_status to 'extracted' after function extraction attempt
-                db.prepare(`UPDATE files SET processing_status = 'extracted' WHERE id = @id`).run({ id: fileId });
+            // 🧠 Only re-generate summary if it's missing or outdated
+            const needsResummary = !row.summary ||
+                !row.indexed_at ||
+                (row.last_modified && new Date(row.last_modified) > new Date(row.indexed_at));
+            if (needsResummary) {
+                log(`📝 Generating summary for ${row.path}...`);
+                const summaryResult = await summaryModule.run({ content, filepath: row.path });
+                const summary = summaryResult?.summary?.trim() || null;
+                let embedding = null;
+                if (summary) {
+                    const vector = await generateEmbedding(summary);
+                    if (vector) {
+                        embedding = JSON.stringify(vector);
+                    }
+                }
+                db.prepare(`
+          UPDATE files
+          SET summary = @summary,
+              embedding = @embedding,
+              indexed_at = datetime('now')
+          WHERE path = @path
+        `).run({ summary, embedding, path: row.path });
+                log(`✅ Updated summary & embedding for ${row.path}`);
             }
             else {
-                log(`⚠️ Could not find fileId for ${row.path}`);
-            }
-            log(`📝 Generating summary for ${row.path}...`);
-            const summaryResult = await summaryModule.run({ content, filepath: row.path });
-            const summary = summaryResult?.summary?.trim() || null;
-            let embedding = null;
-            if (summary) {
-                const vector = await generateEmbedding(summary);
-                if (vector) {
-                    embedding = JSON.stringify(vector);
-                }
+                log(`⚡ Skipped summary (up-to-date) for ${row.path}`);
             }
+            // 🧩 Extract functions and update processing status
+            await indexFunctionsForFile(row.path, row.id);
             db.prepare(`
         UPDATE files
-        SET summary = @summary, embedding = @embedding, indexed_at = datetime('now')
-        WHERE path = @path
-      `).run({ summary, embedding, path: row.path });
-            log(`✅ Updated summary & embedding for ${row.path}\n`);
+        SET processing_status = 'extracted',
+            functions_extracted_at = datetime('now')
+        WHERE id = @id
+      `).run({ id: row.id });
+            log(`✅ Function extraction complete for ${row.path}\n`);
         }
         catch (err) {
             log(`❌ Failed: ${row.path}: ${err instanceof Error ? err.message : String(err)}\n`);

package/dist/db/schema.js

@@ -9,7 +9,8 @@ export function initSchema() {
       indexed_at TEXT,
       last_modified TEXT,
       embedding TEXT,
-      processing_status TEXT
+      processing_status TEXT,
+      functions_extracted_at TEXT
     );
 
     CREATE VIRTUAL TABLE IF NOT EXISTS files_fts

package/dist/utils/buildContextualPrompt.js

@@ -0,0 +1,20 @@
+export function buildContextualPrompt({ baseInstruction, code, summary, functions, relatedFiles, projectFileTree }) {
+    const parts = [baseInstruction];
+    if (summary) {
+        parts.push(`📄 File Summary:\n${summary}`);
+    }
+    if (functions?.length) {
+        parts.push(`🔧 Functions:\n${functions.join(', ')}`);
+    }
+    if (relatedFiles?.length) {
+        const formatted = relatedFiles
+            .map(f => `• ${f.path}: ${f.summary}`)
+            .join('\n');
+        parts.push(`📚 Related Files:\n${formatted}`);
+    }
+    if (projectFileTree) {
+        parts.push(`📁 Project File Structure:\n\`\`\`\n${projectFileTree.trim()}\n\`\`\``);
+    }
+    parts.push(`\n--- CODE START ---\n${code}\n--- CODE END ---`);
+    return parts.join('\n\n');
+}

package/dist/utils/fileTree.js

@@ -0,0 +1,30 @@
+import fs from 'fs';
+import path from 'path';
+const IGNORED_DIRS = new Set([
+    'node_modules', 'dist', '.git', '.vscode', 'coverage', 'build', 'out', 'logs', 'tmp'
+]);
+export function generateFileTree(dir, maxDepth = 3, currentDepth = 0, prefix = '') {
+    if (currentDepth > maxDepth)
+        return '';
+    let output = '';
+    const items = fs.readdirSync(dir, { withFileTypes: true })
+        .filter(item => !IGNORED_DIRS.has(item.name))
+        .sort((a, b) => {
+        // Directories first
+        if (a.isDirectory() && !b.isDirectory())
+            return -1;
+        if (!a.isDirectory() && b.isDirectory())
+            return 1;
+        return a.name.localeCompare(b.name);
+    });
+    for (const [i, item] of items.entries()) {
+        const isLast = i === items.length - 1;
+        const connector = isLast ? '└── ' : '├── ';
+        const childPrefix = prefix + (isLast ? '    ' : '│   ');
+        output += `${prefix}${connector}${item.name}\n`;
+        if (item.isDirectory()) {
+            output += generateFileTree(path.join(dir, item.name), maxDepth, currentDepth + 1, childPrefix);
+        }
+    }
+    return output;
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "scai",
-  "version": "0.1.40",
+  "version": "0.1.42",
   "type": "module",
   "bin": {
     "scai": "./dist/index.js"