18 // rebuild merges added messages with the log of agent, in date order, and
19 // builds a new memory beside it. The caller holds the token lock. The model
20 // uses the original agent prompt and cache key. A failed or stopped build
21 // removes the new directory and leaves the original directory unchanged.
22 // Success copies the other agent files, adds this run's usage, and swaps the
23 // directories, with the original kept as a timestamped backup. Message IDs are
24 // new positions in the merged log, but kind, text, and date stay exact.
25 func rebuild(ctx context.Context, agent *Agent, added []DatedMessage, client *ModelClient, summarizer Summarizer, out io.Writer) (backup string, err error) {
26 err = rebuildIdle(agent.dir)
28 return "", fmt.Errorf("check rebuild work: %w", err)
31 old, err := rebuildReadLog(ctx, agent.dir)
33 return "", fmt.Errorf("read original log: %w", err)
36 entries, err := rebuildMerge(old, added)
38 return "", fmt.Errorf("merge rebuild messages: %w", err)
41 root, err := os.MkdirTemp(filepath.Dir(agent.dir), ".agent-rebuild-")
43 return "", fmt.Errorf("create rebuild directory: %w", err)
47 closeErr := os.RemoveAll(root)
49 err = errors.Join(err, fmt.Errorf("remove rebuild directory: %w", closeErr))
53 memory, err := openMemory(ctx, root, time.Now, summarizer, log.New(io.Discard, "", 0))
55 return "", fmt.Errorf("open rebuilt memory: %w", err)
58 client.LimitCalls(buildCallLimit)
63 usagePath: client.UsagePath(),
66 appended, stop := run.Run(ctx, entries)
67 sizes, err := memory.buildModelLines()
69 err = errors.Join(err, memory.Settle(ctx), memory.rebuildComplete(), memory.Flush(ctx))
72 err = errors.Join(err, memory.Close())
73 records, usageErr := buildReadUsage(client.UsagePath())
74 err = errors.Join(err, usageErr, buildWriteReport(out, appended, stop, records, sizes))
76 err = errors.Join(err, fmt.Errorf("rebuild stopped: %s", stop))
80 return "", fmt.Errorf("finish rebuilt memory: %w", err)
83 err = rebuildCopy(agent.dir, root)
85 return "", fmt.Errorf("copy other agent files: %w", err)
88 usage, err := os.ReadFile(client.UsagePath())
89 if errors.Is(err, os.ErrNotExist) {
94 return "", fmt.Errorf("read rebuild usage: %w", err)
98 err = rebuildAddUsage(filepath.Join(root, "usage.jsonl"), usage)
100 return "", fmt.Errorf("append rebuild usage: %w", err)
104 if ctx.Err() != nil {
108 backup = agent.dir + "-backup-" + time.Now().Format("20060102-150405.000000000")
109 err = os.Rename(agent.dir, backup)
111 return "", fmt.Errorf("back up agent: %w", err)
114 err = os.Rename(root, agent.dir)
116 restoreErr := os.Rename(backup, agent.dir)
117 if restoreErr != nil {
118 err = errors.Join(err, fmt.Errorf("restore original agent: %w", restoreErr))
121 return "", fmt.Errorf("install rebuilt agent: %w", err)
127 // rebuildIdle refuses an agent with unfinished subagents, including ones that a
128 // daemon restart must resume. A missing work file lists no work.
129 func rebuildIdle(root string) error {
130 data, err := os.ReadFile(filepath.Join(root, subagentWorkFileName))
131 if errors.Is(err, os.ErrNotExist) {
136 return fmt.Errorf("read running subagents: %w", err)
139 var work SubagentWorkFile
141 err = json.Unmarshal(data, &work)
143 return fmt.Errorf("decode running subagents: %w", err)
146 if len(work.Running) > 0 {
147 return errors.New("rebuild refused: subagents have unfinished work")
153 // rebuildReadLog reads and validates the day files without a write or an index
154 // repair. Even an incomplete final line is an error, not a repair.
155 func rebuildReadLog(ctx context.Context, root string) ([]DatedMessage, error) {
156 paths, err := filepath.Glob(filepath.Join(root, "main", "*.jsonl"))
158 return nil, fmt.Errorf("list message files: %w", err)
163 var entries []DatedMessage
165 for _, path := range paths {
168 day, err = journalDayNumber(strings.TrimSuffix(filepath.Base(path), ".jsonl"))
170 return nil, fmt.Errorf("parse message day: %w", err)
173 err = journalScanFile(ctx, path, day, 0, false, func(line []byte, _ Position) error {
176 err = json.Unmarshal(line, &message)
178 return fmt.Errorf("decode message: %w", err)
181 err = logValidateMessage(message)
183 return fmt.Errorf("validate message: %w", err)
186 if message.ID != len(entries) {
187 return fmt.Errorf("expected message %d, found %d", len(entries), message.ID)
190 entries = append(entries, DatedMessage{
199 return nil, fmt.Errorf("read message file: %w", err)
206 // rebuildMerge preserves the input order for equal dates, with old messages
207 // before added messages. Each date must be explicit and valid.
208 func rebuildMerge(old, added []DatedMessage) ([]DatedMessage, error) {
209 entries := append(slices.Clone(old), added...)
210 dates := make(map[string]time.Time)
211 for _, entry := range entries {
212 err := logValidateMessage(Message{
215 Size: len(entry.Text),
219 return nil, fmt.Errorf("validate dated message: %w", err)
224 date, err = time.Parse(time.RFC3339Nano, entry.Date)
226 return nil, fmt.Errorf("parse merge date: %w", err)
229 dates[entry.Date] = date
232 slices.SortStableFunc(entries, func(a, b DatedMessage) int {
233 return dates[a.Date].Compare(dates[b.Date])
239 // rebuildCopy copies the other agent directories and regular files. It excludes
240 // the memory and its lock, and refuses all other file types.
241 func rebuildCopy(source, target string) error {
242 err := filepath.WalkDir(source, func(path string, entry fs.DirEntry, err error) error {
244 return fmt.Errorf("read agent entry: %w", err)
247 relative, err := filepath.Rel(source, path)
249 return fmt.Errorf("find agent relative path: %w", err)
257 case "main", "tree", "index":
258 return filepath.SkipDir
259 case "lock", "view.json":
263 destination := filepath.Join(target, relative)
266 err = os.Mkdir(destination, 0o700)
268 return fmt.Errorf("create agent directory: %w", err)
270 case entry.Type().IsRegular():
273 data, err = os.ReadFile(path)
275 return fmt.Errorf("read agent file: %w", err)
278 err = os.WriteFile(destination, data, 0o600)
280 return fmt.Errorf("copy agent file: %w", err)
283 return fmt.Errorf("copy agent: unsupported file type for %s", relative)
289 return fmt.Errorf("copy agent files: %w", err)
295 // rebuildAddUsage keeps both the original usage and the calls of this run.
296 func rebuildAddUsage(path string, data []byte) (err error) {
297 file, err := os.OpenFile(path, os.O_CREATE|os.O_WRONLY|os.O_APPEND, 0o600)
299 return fmt.Errorf("open rebuilt usage: %w", err)
303 closeErr := file.Close()
305 err = errors.Join(err, fmt.Errorf("close rebuilt usage: %w", closeErr))
309 _, err = file.Write(data)
311 return fmt.Errorf("append rebuilt usage: %w", err)