osborn 0.9.228 → 0.9.229

package/dist/pipeline-fastbrain.js

@@ -147,43 +147,56 @@ async function executeTool(name, args, sessionId, workingDir, agentControl) {
 // SEARCH HELPERS
 // ============================================================
 async function executeSearch(phrases, sessionId, workingDir) {
-    const { ripgrepSearch } = await import('./jsonl-search.js');
-    const { getIndexPath, readFullContent } = await import('./summary-index.js');
-    const indexPath = getIndexPath(sessionId, workingDir);
-    if (indexPath) {
-        const sections = [];
-        const matchedRefs = [];
-        const seenLines = new Set();
-        for (const phrase of phrases.slice(0, 6)) {
-            const results = ripgrepSearch(indexPath, phrase, { maxResults: 8, fromEnd: true, contextLines: 0 });
-            const newResults = results.filter((r) => {
-                const key = `${r.lineNumber}`;
-                if (seenLines.has(key))
-                    return false;
-                seenLines.add(key);
-                return true;
-            });
-            if (newResults.length > 0) {
-                sections.push(`["${phrase}" — ${newResults.length} matches]`);
-                for (const r of newResults) {
-                    const parts = r.content.split('|');
-                    if (parts.length >= 6) {
-                        matchedRefs.push({ lineNum: parseInt(parts[0], 10), byteOffset: parseInt(parts[1], 10), source: parts[3] });
-                        sections.push(r.content);
-                    }
-                }
+    // DB-FIRST: query the recall store (full untruncated text + hybrid keyword+vector).
+    // Vector/hybrid is PREFERRED; recall() degrades to keyword INSIDE when the vec table is
+    // empty or no embedder is available. This replaces the old ripgrep-over-search-index.txt
+    // two-step (grep summary → read JSONL by offset) with one call that already returns full text.
+    const { storeExists, openStore, recall } = await import('./session-store.js');
+    const dbPath = storeExists(sessionId, workingDir);
+    if (dbPath) {
+        let embed = undefined;
+        try {
+            if (process.env.OSBORN_EMBED !== '0') {
+                const { getEmbedder } = await import('./embedder.js');
+                // Tight timeout so a slow/cold embed never blows the fast-brain deadline → keyword.
+                embed = (await getEmbedder(2500)) ?? undefined;
             }
         }
-        if (matchedRefs.length > 0) {
-            try {
-                const fullTexts = readFullContent(matchedRefs, sessionId, workingDir, undefined, 2000);
-                if (fullTexts.length > 0)
-                    sections.push('', `[FULL CONTENT — ${fullTexts.length} entries]`, ...fullTexts);
+        catch { }
+        // Best-effort: a recall/decompress throw must degrade to the ripgrep fallback below,
+        // never collapse the whole fast-brain turn (mirrors buildRecallInjection's swallow).
+        let dbResult = null;
+        const db = openStore(dbPath, { readonly: true });
+        try {
+            const query = phrases.slice(0, 6).join(' ');
+            const hits = await recall(db, query, { mode: embed ? 'hybrid' : 'keyword', topK: 12, embed });
+            if (!hits.length) {
+                dbResult = `No matches for: ${phrases.join(', ')}`;
+            }
+            else {
+                const sections = [];
+                for (const h of hits) {
+                    const src = `${h.source} L${h.lineNum} · ${h.msgType}${h.toolName ? `:${h.toolName}` : ''} · ${h.matchedBy}`;
+                    let body = h.text.replace(/\n{3,}/g, '\n\n').trim();
+                    if (body.length > 2000)
+                        body = body.slice(0, 2000) + ' …';
+                    sections.push(`[${src}]${h.ts ? ' ' + h.ts : ''}\n${body}`);
+                }
+                dbResult = sections.join('\n\n');
             }
-            catch { }
         }
-        return sections.length === 0 ? `No matches for: ${phrases.join(', ')}` : sections.join('\n');
+        catch (err) {
+            console.warn('[fast-brain] recall DB search failed, falling back to raw JSONL:', err?.message || err);
+            // dbResult stays null → fall through to the ripgrep fallback below
+        }
+        finally {
+            db.close();
+        }
+        if (dbResult !== null)
+            return dbResult;
     }
+    // FALLBACK: store not built yet (brand-new session before the first sweep) → raw-JSONL ripgrep.
+    const { ripgrepSearch } = await import('./jsonl-search.js');
     const { getSessionPaths } = await import('./session-access.js');
     const paths = getSessionPaths(sessionId, workingDir);
     if (!paths.exists)
@@ -199,31 +212,33 @@ async function executeSearch(phrases, sessionId, workingDir) {
     return sections.length > 0 ? sections.join('\n') : `No matches for: ${phrases.join(', ')}`;
 }
 async function getRecentEntries(sessionId, workingDir, _, count) {
-    const { readFileSync } = await import('fs');
-    const { getIndexPath, readFullContent } = await import('./summary-index.js');
-    const indexPath = getIndexPath(sessionId, workingDir);
-    if (!indexPath)
+    // Newest content rows straight from the recall DB (append-only, so highest id = latest).
+    const { storeExists, openStore, recentRows } = await import('./session-store.js');
+    const dbPath = storeExists(sessionId, workingDir);
+    if (!dbPath)
         return 'Index not built yet.';
-    const content = readFileSync(indexPath, 'utf-8');
-    const recentLines = content.split('\n').filter(Boolean).slice(-count);
-    const refs = [];
-    const summaries = [`[RECENT — last ${recentLines.length} entries]`];
-    for (const line of recentLines) {
-        summaries.push(line);
-        const parts = line.split('|');
-        if (parts.length >= 6) {
-            refs.push({ lineNum: parseInt(parts[0], 10), byteOffset: parseInt(parts[1], 10), source: parts[3] });
+    const db = openStore(dbPath, { readonly: true });
+    try {
+        const rows = recentRows(db, count); // newest first
+        if (!rows.length)
+            return 'No entries yet.';
+        const out = [`[RECENT — last ${rows.length} entries, newest first]`];
+        for (const h of rows) {
+            const src = `${h.source} L${h.lineNum} · ${h.msgType}${h.toolName ? `:${h.toolName}` : ''}`;
+            let body = h.text.replace(/\n{3,}/g, '\n\n').trim();
+            if (body.length > 1500)
+                body = body.slice(0, 1500) + ' …';
+            out.push(`[${src}]${h.ts ? ' ' + h.ts : ''}\n${body}`);
         }
+        return out.join('\n\n');
     }
-    if (refs.length > 0) {
-        try {
-            const fullTexts = readFullContent(refs, sessionId, workingDir, undefined, 1500);
-            if (fullTexts.length > 0)
-                summaries.push('', `[FULL CONTENT — ${fullTexts.length} entries]`, ...fullTexts);
-        }
-        catch { }
+    catch (err) {
+        console.warn('[fast-brain] recent-entries read failed:', err?.message || err);
+        return 'Could not read recent entries.';
+    }
+    finally {
+        db.close();
     }
-    return summaries.join('\n');
 }
 // ============================================================
 // SYSTEM PROMPT
@@ -249,9 +264,11 @@ function buildSystemPrompt(chatHistory, researchContext) {
         `or use vague references ("that thing", "the error"). Interpret intent, not just words.`,
         ``,
         `== RESULTS FORMAT ==`,
-        `Each line: lineNum|byteOffset|timestamp|source|msgType|summary`,
-        `  source: "main" = conversation, "agent-XXXX" = sub-agent research`,
-        `Full content sections have complete untruncated text.`,
+        `Each hit is a header line then its FULL untruncated text:`,
+        `  [<source> L<lineNum> · <msgType>[:<tool>] · <matchedBy>] <timestamp>`,
+        `  <full text of that message/tool result>`,
+        `  source: "main" = conversation, "agent-XXXX" = sub-agent research.`,
+        `  matchedBy: keyword | vector | both (how recall found it — vector/hybrid is preferred).`,
         ``,
         `== HOW TO SEARCH ==`,
         `Think about what words people ACTUALLY USED when this topic came up.`,

package/dist/session-store.d.ts

@@ -88,3 +88,9 @@ export interface RecallOpts {
  * Falls back to keyword automatically when no embeddings/embedder are available.
  */
 export declare function recall(db: Database.Database, query: string, opts?: RecallOpts): Promise<RecallHit[]>;
+/**
+ * Return the most recent N content rows (highest id = latest, append-only order),
+ * newest first, as RecallHit-shaped objects. Powers fast-brain's "recent entries"
+ * path. Encapsulates decompress so callers never touch the blob/compression.
+ */
+export declare function recentRows(db: Database.Database, count: number): RecallHit[];

package/dist/session-store.js

@@ -571,3 +571,27 @@ export async function recall(db, query, opts) {
     }
     return hits;
 }
+/**
+ * Return the most recent N content rows (highest id = latest, append-only order),
+ * newest first, as RecallHit-shaped objects. Powers fast-brain's "recent entries"
+ * path. Encapsulates decompress so callers never touch the blob/compression.
+ */
+export function recentRows(db, count) {
+    const rows = db.prepare(`SELECT id, source, line_num, byte_offset, ts, msg_type, model, git_branch, cwd, tool_name, blob
+     FROM content ORDER BY id DESC LIMIT ?`).all(Math.max(1, count));
+    return rows.map((row) => ({
+        id: row.id,
+        source: row.source,
+        lineNum: row.line_num,
+        byteOffset: row.byte_offset,
+        ts: row.ts,
+        msgType: row.msg_type,
+        model: row.model,
+        gitBranch: row.git_branch,
+        cwd: row.cwd,
+        toolName: row.tool_name,
+        text: decompress(row.blob),
+        score: 0,
+        matchedBy: 'keyword',
+    }));
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "osborn",
-  "version": "0.9.228",
+  "version": "0.9.229",
   "description": "Voice AI coding assistant - local agent that connects to Osborn frontend",
   "type": "module",
   "bin": {