18 // rebuildCallLimit is the most model calls of a rebuild. A rebuild summarizes
19 // the whole merged log, which can need more calls than a probe sample, and the
20 // quota stop of the build run still guards the account.
21 const rebuildCallLimit = 1500
23 // rebuild merges added messages with the log of agent, in date order, and
24 // builds a new memory beside it. The caller holds the token lock. The model
25 // uses the original agent prompt and cache key. A failed or stopped build
26 // removes the new directory and leaves the original directory unchanged.
27 // Success copies the other agent files, adds this run's usage, and swaps the
28 // directories, with the original kept as a timestamped backup. Message IDs are
29 // new positions in the merged log, but kind, text, and date stay exact.
30 func rebuild(ctx context.Context, agent *Agent, added []DatedMessage, client *ModelClient, summarizer Summarizer, out io.Writer) (backup string, err error) {
31 err = rebuildIdle(agent.dir)
33 return "", fmt.Errorf("check rebuild work: %w", err)
36 old, err := rebuildReadLog(ctx, agent.dir)
38 return "", fmt.Errorf("read original log: %w", err)
41 entries, err := rebuildMerge(old, added)
43 return "", fmt.Errorf("merge rebuild messages: %w", err)
46 root, err := os.MkdirTemp(filepath.Dir(agent.dir), ".agent-rebuild-")
48 return "", fmt.Errorf("create rebuild directory: %w", err)
52 closeErr := os.RemoveAll(root)
54 err = errors.Join(err, fmt.Errorf("remove rebuild directory: %w", closeErr))
58 memory, err := openMemory(ctx, root, time.Now, summarizer, log.New(io.Discard, "", 0))
60 return "", fmt.Errorf("open rebuilt memory: %w", err)
63 client.LimitCalls(rebuildCallLimit)
68 usagePath: client.UsagePath(),
71 appended, stop := run.Run(ctx, entries)
72 sizes, err := memory.buildModelLines()
74 err = errors.Join(err, memory.Settle(ctx), memory.rebuildComplete(), memory.Flush(ctx))
77 err = errors.Join(err, memory.Close())
78 records, usageErr := buildReadUsage(client.UsagePath())
79 err = errors.Join(err, usageErr, buildWriteReport(out, appended, stop, records, sizes))
81 err = errors.Join(err, fmt.Errorf("rebuild stopped: %s", stop))
85 return "", fmt.Errorf("finish rebuilt memory: %w", err)
88 err = rebuildCopy(agent.dir, root)
90 return "", fmt.Errorf("copy other agent files: %w", err)
93 usage, err := os.ReadFile(client.UsagePath())
94 if errors.Is(err, os.ErrNotExist) {
99 return "", fmt.Errorf("read rebuild usage: %w", err)
103 err = rebuildAddUsage(filepath.Join(root, "usage.jsonl"), usage)
105 return "", fmt.Errorf("append rebuild usage: %w", err)
109 if ctx.Err() != nil {
113 backup = agent.dir + "-backup-" + time.Now().Format("20060102-150405.000000000")
114 err = os.Rename(agent.dir, backup)
116 return "", fmt.Errorf("back up agent: %w", err)
119 err = os.Rename(root, agent.dir)
121 restoreErr := os.Rename(backup, agent.dir)
122 if restoreErr != nil {
123 err = errors.Join(err, fmt.Errorf("restore original agent: %w", restoreErr))
126 return "", fmt.Errorf("install rebuilt agent: %w", err)
132 // rebuildIdle refuses an agent with unfinished subagents, including ones that a
133 // daemon restart must resume. A missing work file lists no work.
134 func rebuildIdle(root string) error {
135 data, err := os.ReadFile(filepath.Join(root, subagentWorkFileName))
136 if errors.Is(err, os.ErrNotExist) {
141 return fmt.Errorf("read running subagents: %w", err)
144 var work SubagentWorkFile
146 err = json.Unmarshal(data, &work)
148 return fmt.Errorf("decode running subagents: %w", err)
151 if len(work.Running) > 0 {
152 return errors.New("rebuild refused: subagents have unfinished work")
158 // rebuildReadLog reads and validates the day files without a write or an index
159 // repair. Even an incomplete final line is an error, not a repair.
160 func rebuildReadLog(ctx context.Context, root string) ([]DatedMessage, error) {
161 paths, err := filepath.Glob(filepath.Join(root, "main", "*.jsonl"))
163 return nil, fmt.Errorf("list message files: %w", err)
168 var entries []DatedMessage
170 for _, path := range paths {
173 day, err = journalDayNumber(strings.TrimSuffix(filepath.Base(path), ".jsonl"))
175 return nil, fmt.Errorf("parse message day: %w", err)
178 err = journalScanFile(ctx, path, day, 0, false, func(line []byte, _ Position) error {
181 err = json.Unmarshal(line, &message)
183 return fmt.Errorf("decode message: %w", err)
186 err = logValidateMessage(message)
188 return fmt.Errorf("validate message: %w", err)
191 if message.ID != len(entries) {
192 return fmt.Errorf("expected message %d, found %d", len(entries), message.ID)
195 entries = append(entries, DatedMessage{
204 return nil, fmt.Errorf("read message file: %w", err)
211 // rebuildMerge preserves the input order for equal dates, with old messages
212 // before added messages. Each date must be explicit and valid.
213 func rebuildMerge(old, added []DatedMessage) ([]DatedMessage, error) {
214 entries := append(slices.Clone(old), added...)
215 dates := make(map[string]time.Time)
216 for _, entry := range entries {
217 err := logValidateMessage(Message{
220 Size: len(entry.Text),
224 return nil, fmt.Errorf("validate dated message: %w", err)
229 date, err = time.Parse(time.RFC3339Nano, entry.Date)
231 return nil, fmt.Errorf("parse merge date: %w", err)
234 dates[entry.Date] = date
237 slices.SortStableFunc(entries, func(a, b DatedMessage) int {
238 return dates[a.Date].Compare(dates[b.Date])
244 // rebuildCopy copies the other agent directories and regular files. It excludes
245 // the memory and its lock, and refuses all other file types.
246 func rebuildCopy(source, target string) error {
247 err := filepath.WalkDir(source, func(path string, entry fs.DirEntry, err error) error {
249 return fmt.Errorf("read agent entry: %w", err)
252 relative, err := filepath.Rel(source, path)
254 return fmt.Errorf("find agent relative path: %w", err)
262 case "main", "tree", "index":
263 return filepath.SkipDir
264 case "lock", "view.json":
268 destination := filepath.Join(target, relative)
271 err = os.Mkdir(destination, 0o700)
273 return fmt.Errorf("create agent directory: %w", err)
275 case entry.Type().IsRegular():
278 data, err = os.ReadFile(path)
280 return fmt.Errorf("read agent file: %w", err)
283 err = os.WriteFile(destination, data, 0o600)
285 return fmt.Errorf("copy agent file: %w", err)
288 return fmt.Errorf("copy agent: unsupported file type for %s", relative)
294 return fmt.Errorf("copy agent files: %w", err)
300 // rebuildAddUsage keeps both the original usage and the calls of this run.
301 func rebuildAddUsage(path string, data []byte) (err error) {
302 file, err := os.OpenFile(path, os.O_CREATE|os.O_WRONLY|os.O_APPEND, 0o600)
304 return fmt.Errorf("open rebuilt usage: %w", err)
308 closeErr := file.Close()
310 err = errors.Join(err, fmt.Errorf("close rebuilt usage: %w", closeErr))
314 _, err = file.Write(data)
316 return fmt.Errorf("append rebuilt usage: %w", err)