@jlongo78/agent-spaces 0.9.6 → 0.9.8

package/.next/standalone/src/lib/codex/parser.ts

@@ -1,287 +1,287 @@
-import fs from 'fs';
-import path from 'path';
-
-/**
- * Metadata extracted from a Codex CLI session rollout file.
- */
-export interface CodexSessionMeta {
-  id: string;
-  cwd: string;
-  timestamp: string;
-  firstPrompt: string;
-  messageCount: number;
-  modified: string;
-}
-
-/**
- * Read a bounded prefix of a file, stopping once we've seen enough newline
- * boundaries to parse the session metadata and the first few events.
- */
-function readPrefix(filePath: string, maxBytes: number, minNewlines: number): string {
-  let fd: number | null = null;
-  try {
-    fd = fs.openSync(filePath, 'r');
-    const chunks: Buffer[] = [];
-    let offset = 0;
-    let newlineCount = 0;
-
-    while (offset < maxBytes && newlineCount < minNewlines) {
-      const toRead = Math.min(8192, maxBytes - offset);
-      const buf = Buffer.alloc(toRead);
-      const bytesRead = fs.readSync(fd, buf, 0, toRead, offset);
-      if (bytesRead <= 0) break;
-
-      const chunk = buf.subarray(0, bytesRead);
-      chunks.push(chunk);
-      offset += bytesRead;
-
-      for (let i = 0; i < chunk.length; i++) {
-        if (chunk[i] === 0x0a) newlineCount++;
-      }
-    }
-
-    return Buffer.concat(chunks).toString('utf-8');
-  } finally {
-    if (fd !== null) fs.closeSync(fd);
-  }
-}
-
-/**
- * Extract the session UUID from a rollout filename.
- * Expected format: rollout-<timestamp>-<session-uuid>.jsonl
- * The session UUID is the last hyphen-separated segment before .jsonl,
- * which itself is a standard UUID (8-4-4-4-12 hex).
- */
-function sessionIdFromFilename(filename: string): string {
-  // Strip extension
-  const base = filename.replace(/\.jsonl$/, '');
-  // Match a UUID at the end: 8-4-4-4-12 hex pattern
-  const uuidMatch = base.match(
-    /([0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12})$/i
-  );
-  if (uuidMatch) return uuidMatch[1];
-  // Fallback: everything after "rollout-"
-  return base.replace(/^rollout-/, '');
-}
-
-/**
- * Try to extract a user message string from a parsed JSONL event.
- *
- * Codex CLI uses several shapes depending on version:
- *   - { payload: { type: "user_message", text: "..." } }
- *   - { payload: { type: "UserMessage", text: "..." } }
- *   - { type: "user_message", text: "..." }
- *   - { type: "UserMessage", text: "..." }
- *   - { payload: { type: "user_message", content: "..." } }
- *   - { payload: { type: "user_message", message: { content: "..." } } }
- */
-function extractUserMessage(obj: Record<string, unknown>): string | null {
-  const candidates = [obj.payload as Record<string, unknown> | undefined, obj];
-  for (const c of candidates) {
-    if (!c || typeof c !== 'object') continue;
-    const t = c.type;
-    if (t === 'user_message' || t === 'UserMessage') {
-      // Direct text field
-      if (typeof c.text === 'string') return c.text;
-      // Content field
-      if (typeof c.content === 'string') return c.content;
-      // Nested message.content
-      const msg = c.message as Record<string, unknown> | undefined;
-      if (msg && typeof msg.content === 'string') return msg.content;
-    }
-  }
-  return null;
-}
-
-/**
- * Check if a parsed JSONL line looks like a conversation event
- * (something we should count toward messageCount).
- */
-function isConversationEvent(obj: Record<string, unknown>): boolean {
-  const payload = obj.payload as Record<string, unknown> | undefined;
-  const t = (payload?.type ?? obj.type) as string | undefined;
-  if (!t) return false;
-  // Count user messages, assistant messages, and tool outputs
-  const eventTypes = new Set([
-    'user_message', 'UserMessage',
-    'assistant_message', 'AssistantMessage',
-    'message', 'response',
-    'tool_call', 'ToolCall',
-    'tool_result', 'ToolResult',
-  ]);
-  return eventTypes.has(t);
-}
-
-/**
- * Parse a single rollout JSONL file into session metadata.
- * Reads a bounded prefix large enough to include Codex's oversized
- * `session_meta` line plus a few subsequent conversation events.
- */
-function parseRolloutFile(filePath: string, filename: string): CodexSessionMeta | null {
-  try {
-    const head = readPrefix(filePath, 256 * 1024, 6);
-    const stat = fs.statSync(filePath);
-    const lines = head.split('\n').filter(l => l.trim().length > 0);
-
-    if (lines.length === 0) return null;
-
-    let id = '';
-    let cwd = '';
-    let timestamp = '';
-    let firstPrompt = '';
-    let messageCount = 0;
-
-    for (let i = 0; i < lines.length; i++) {
-      let parsed: Record<string, unknown>;
-      try {
-        parsed = JSON.parse(lines[i]);
-      } catch {
-        continue;
-      }
-
-      // First valid JSON line: extract session metadata
-      if (i === 0 || (!id && !cwd && !timestamp)) {
-        const payload = parsed.payload as Record<string, unknown> | undefined;
-        // Session ID: try item.id, id, session_id
-        const item = parsed.item as Record<string, unknown> | undefined;
-        id = (
-          (item?.id as string) ||
-          (payload?.id as string) ||
-          (parsed.id as string) ||
-          (parsed.session_id as string) ||
-          ''
-        );
-        // CWD: try item.cwd, cwd
-        cwd = (
-          (item?.cwd as string) ||
-          (payload?.cwd as string) ||
-          (parsed.cwd as string) ||
-          ''
-        );
-        // Timestamp: try item.timestamp, timestamp, created_at, created
-        timestamp = (
-          (item?.timestamp as string) ||
-          (payload?.timestamp as string) ||
-          (parsed.timestamp as string) ||
-          (item?.created_at as string) ||
-          (payload?.created_at as string) ||
-          (parsed.created_at as string) ||
-          (parsed.created as string) ||
-          ''
-        );
-      }
-
-      // Count conversation events
-      if (isConversationEvent(parsed)) {
-        messageCount++;
-      }
-
-      // Extract first user message
-      if (!firstPrompt) {
-        const msg = extractUserMessage(parsed);
-        if (msg) {
-          firstPrompt = msg.slice(0, 500);
-        }
-      }
-    }
-
-    // Fallback: extract session ID from filename
-    if (!id) {
-      id = sessionIdFromFilename(filename);
-    }
-
-    // Derive timestamp from filename if not found in content
-    // Filename format: rollout-<timestamp>-<uuid>.jsonl
-    if (!timestamp) {
-      const tsMatch = filename.match(/^rollout-(\d+)/);
-      if (tsMatch) {
-        const epochMs = parseInt(tsMatch[1], 10);
-        // Codex uses millisecond timestamps typically
-        const date = new Date(epochMs > 1e12 ? epochMs : epochMs * 1000);
-        if (!isNaN(date.getTime())) {
-          timestamp = date.toISOString();
-        }
-      }
-    }
-
-    // Last resort: use file mtime for timestamp
-    if (!timestamp) {
-      timestamp = stat.mtime.toISOString();
-    }
-
-    return {
-      id,
-      cwd,
-      timestamp,
-      firstPrompt,
-      messageCount,
-      modified: stat.mtime.toISOString(),
-    };
-  } catch {
-    return null;
-  }
-}
-
-/**
- * Recursively collect all `rollout-*.jsonl` files under the YYYY/MM/DD
- * directory tree inside the Codex sessions directory.
- */
-function collectRolloutFiles(dir: string): { filePath: string; filename: string }[] {
-  const results: { filePath: string; filename: string }[] = [];
-
-  try {
-    const entries = fs.readdirSync(dir, { withFileTypes: true });
-    for (const entry of entries) {
-      const fullPath = path.join(dir, entry.name);
-      if (entry.isDirectory()) {
-        // Recurse into YYYY, MM, DD subdirectories
-        results.push(...collectRolloutFiles(fullPath));
-      } else if (
-        entry.isFile() &&
-        entry.name.startsWith('rollout-') &&
-        entry.name.endsWith('.jsonl')
-      ) {
-        results.push({ filePath: fullPath, filename: entry.name });
-      }
-    }
-  } catch {
-    // Directory unreadable or doesn't exist
-  }
-
-  return results;
-}
-
-/**
- * Scan the Codex sessions directory (~/.codex/sessions/) for all rollout
- * files and return metadata for each session.
- *
- * The directory structure is: YYYY/MM/DD/rollout-<timestamp>-<session-uuid>.jsonl
- *
- * Each file is read only partially (first 8KB) for efficiency.
- * Bad or unreadable files are silently skipped.
- */
-export function scanCodexSessions(sessionsDir: string): CodexSessionMeta[] {
-  if (!sessionsDir || !fs.existsSync(sessionsDir)) {
-    return [];
-  }
-
-  const files = collectRolloutFiles(sessionsDir);
-  const sessions: CodexSessionMeta[] = [];
-
-  for (const { filePath, filename } of files) {
-    const meta = parseRolloutFile(filePath, filename);
-    if (meta) {
-      sessions.push(meta);
-    }
-  }
-
-  // Sort by timestamp descending (most recent first)
-  sessions.sort((a, b) => {
-    if (a.timestamp && b.timestamp) {
-      return b.timestamp.localeCompare(a.timestamp);
-    }
-    return b.modified.localeCompare(a.modified);
-  });
-
-  return sessions;
-}
+import fs from 'fs';
+import path from 'path';
+
+/**
+ * Metadata extracted from a Codex CLI session rollout file.
+ */
+export interface CodexSessionMeta {
+  id: string;
+  cwd: string;
+  timestamp: string;
+  firstPrompt: string;
+  messageCount: number;
+  modified: string;
+}
+
+/**
+ * Read a bounded prefix of a file, stopping once we've seen enough newline
+ * boundaries to parse the session metadata and the first few events.
+ */
+function readPrefix(filePath: string, maxBytes: number, minNewlines: number): string {
+  let fd: number | null = null;
+  try {
+    fd = fs.openSync(filePath, 'r');
+    const chunks: Buffer[] = [];
+    let offset = 0;
+    let newlineCount = 0;
+
+    while (offset < maxBytes && newlineCount < minNewlines) {
+      const toRead = Math.min(8192, maxBytes - offset);
+      const buf = Buffer.alloc(toRead);
+      const bytesRead = fs.readSync(fd, buf, 0, toRead, offset);
+      if (bytesRead <= 0) break;
+
+      const chunk = buf.subarray(0, bytesRead);
+      chunks.push(chunk);
+      offset += bytesRead;
+
+      for (let i = 0; i < chunk.length; i++) {
+        if (chunk[i] === 0x0a) newlineCount++;
+      }
+    }
+
+    return Buffer.concat(chunks).toString('utf-8');
+  } finally {
+    if (fd !== null) fs.closeSync(fd);
+  }
+}
+
+/**
+ * Extract the session UUID from a rollout filename.
+ * Expected format: rollout-<timestamp>-<session-uuid>.jsonl
+ * The session UUID is the last hyphen-separated segment before .jsonl,
+ * which itself is a standard UUID (8-4-4-4-12 hex).
+ */
+function sessionIdFromFilename(filename: string): string {
+  // Strip extension
+  const base = filename.replace(/\.jsonl$/, '');
+  // Match a UUID at the end: 8-4-4-4-12 hex pattern
+  const uuidMatch = base.match(
+    /([0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12})$/i
+  );
+  if (uuidMatch) return uuidMatch[1];
+  // Fallback: everything after "rollout-"
+  return base.replace(/^rollout-/, '');
+}
+
+/**
+ * Try to extract a user message string from a parsed JSONL event.
+ *
+ * Codex CLI uses several shapes depending on version:
+ *   - { payload: { type: "user_message", text: "..." } }
+ *   - { payload: { type: "UserMessage", text: "..." } }
+ *   - { type: "user_message", text: "..." }
+ *   - { type: "UserMessage", text: "..." }
+ *   - { payload: { type: "user_message", content: "..." } }
+ *   - { payload: { type: "user_message", message: { content: "..." } } }
+ */
+function extractUserMessage(obj: Record<string, unknown>): string | null {
+  const candidates = [obj.payload as Record<string, unknown> | undefined, obj];
+  for (const c of candidates) {
+    if (!c || typeof c !== 'object') continue;
+    const t = c.type;
+    if (t === 'user_message' || t === 'UserMessage') {
+      // Direct text field
+      if (typeof c.text === 'string') return c.text;
+      // Content field
+      if (typeof c.content === 'string') return c.content;
+      // Nested message.content
+      const msg = c.message as Record<string, unknown> | undefined;
+      if (msg && typeof msg.content === 'string') return msg.content;
+    }
+  }
+  return null;
+}
+
+/**
+ * Check if a parsed JSONL line looks like a conversation event
+ * (something we should count toward messageCount).
+ */
+function isConversationEvent(obj: Record<string, unknown>): boolean {
+  const payload = obj.payload as Record<string, unknown> | undefined;
+  const t = (payload?.type ?? obj.type) as string | undefined;
+  if (!t) return false;
+  // Count user messages, assistant messages, and tool outputs
+  const eventTypes = new Set([
+    'user_message', 'UserMessage',
+    'assistant_message', 'AssistantMessage',
+    'message', 'response',
+    'tool_call', 'ToolCall',
+    'tool_result', 'ToolResult',
+  ]);
+  return eventTypes.has(t);
+}
+
+/**
+ * Parse a single rollout JSONL file into session metadata.
+ * Reads a bounded prefix large enough to include Codex's oversized
+ * `session_meta` line plus a few subsequent conversation events.
+ */
+function parseRolloutFile(filePath: string, filename: string): CodexSessionMeta | null {
+  try {
+    const head = readPrefix(filePath, 256 * 1024, 6);
+    const stat = fs.statSync(filePath);
+    const lines = head.split('\n').filter(l => l.trim().length > 0);
+
+    if (lines.length === 0) return null;
+
+    let id = '';
+    let cwd = '';
+    let timestamp = '';
+    let firstPrompt = '';
+    let messageCount = 0;
+
+    for (let i = 0; i < lines.length; i++) {
+      let parsed: Record<string, unknown>;
+      try {
+        parsed = JSON.parse(lines[i]);
+      } catch {
+        continue;
+      }
+
+      // First valid JSON line: extract session metadata
+      if (i === 0 || (!id && !cwd && !timestamp)) {
+        const payload = parsed.payload as Record<string, unknown> | undefined;
+        // Session ID: try item.id, id, session_id
+        const item = parsed.item as Record<string, unknown> | undefined;
+        id = (
+          (item?.id as string) ||
+          (payload?.id as string) ||
+          (parsed.id as string) ||
+          (parsed.session_id as string) ||
+          ''
+        );
+        // CWD: try item.cwd, cwd
+        cwd = (
+          (item?.cwd as string) ||
+          (payload?.cwd as string) ||
+          (parsed.cwd as string) ||
+          ''
+        );
+        // Timestamp: try item.timestamp, timestamp, created_at, created
+        timestamp = (
+          (item?.timestamp as string) ||
+          (payload?.timestamp as string) ||
+          (parsed.timestamp as string) ||
+          (item?.created_at as string) ||
+          (payload?.created_at as string) ||
+          (parsed.created_at as string) ||
+          (parsed.created as string) ||
+          ''
+        );
+      }
+
+      // Count conversation events
+      if (isConversationEvent(parsed)) {
+        messageCount++;
+      }
+
+      // Extract first user message
+      if (!firstPrompt) {
+        const msg = extractUserMessage(parsed);
+        if (msg) {
+          firstPrompt = msg.slice(0, 500);
+        }
+      }
+    }
+
+    // Fallback: extract session ID from filename
+    if (!id) {
+      id = sessionIdFromFilename(filename);
+    }
+
+    // Derive timestamp from filename if not found in content
+    // Filename format: rollout-<timestamp>-<uuid>.jsonl
+    if (!timestamp) {
+      const tsMatch = filename.match(/^rollout-(\d+)/);
+      if (tsMatch) {
+        const epochMs = parseInt(tsMatch[1], 10);
+        // Codex uses millisecond timestamps typically
+        const date = new Date(epochMs > 1e12 ? epochMs : epochMs * 1000);
+        if (!isNaN(date.getTime())) {
+          timestamp = date.toISOString();
+        }
+      }
+    }
+
+    // Last resort: use file mtime for timestamp
+    if (!timestamp) {
+      timestamp = stat.mtime.toISOString();
+    }
+
+    return {
+      id,
+      cwd,
+      timestamp,
+      firstPrompt,
+      messageCount,
+      modified: stat.mtime.toISOString(),
+    };
+  } catch {
+    return null;
+  }
+}
+
+/**
+ * Recursively collect all `rollout-*.jsonl` files under the YYYY/MM/DD
+ * directory tree inside the Codex sessions directory.
+ */
+function collectRolloutFiles(dir: string): { filePath: string; filename: string }[] {
+  const results: { filePath: string; filename: string }[] = [];
+
+  try {
+    const entries = fs.readdirSync(dir, { withFileTypes: true });
+    for (const entry of entries) {
+      const fullPath = path.join(dir, entry.name);
+      if (entry.isDirectory()) {
+        // Recurse into YYYY, MM, DD subdirectories
+        results.push(...collectRolloutFiles(fullPath));
+      } else if (
+        entry.isFile() &&
+        entry.name.startsWith('rollout-') &&
+        entry.name.endsWith('.jsonl')
+      ) {
+        results.push({ filePath: fullPath, filename: entry.name });
+      }
+    }
+  } catch {
+    // Directory unreadable or doesn't exist
+  }
+
+  return results;
+}
+
+/**
+ * Scan the Codex sessions directory (~/.codex/sessions/) for all rollout
+ * files and return metadata for each session.
+ *
+ * The directory structure is: YYYY/MM/DD/rollout-<timestamp>-<session-uuid>.jsonl
+ *
+ * Each file is read only partially (first 8KB) for efficiency.
+ * Bad or unreadable files are silently skipped.
+ */
+export function scanCodexSessions(sessionsDir: string): CodexSessionMeta[] {
+  if (!sessionsDir || !fs.existsSync(sessionsDir)) {
+    return [];
+  }
+
+  const files = collectRolloutFiles(sessionsDir);
+  const sessions: CodexSessionMeta[] = [];
+
+  for (const { filePath, filename } of files) {
+    const meta = parseRolloutFile(filePath, filename);
+    if (meta) {
+      sessions.push(meta);
+    }
+  }
+
+  // Sort by timestamp descending (most recent first)
+  sessions.sort((a, b) => {
+    if (a.timestamp && b.timestamp) {
+      return b.timestamp.localeCompare(a.timestamp);
+    }
+    return b.modified.localeCompare(a.modified);
+  });
+
+  return sessions;
+}