]> git.giacomolli.uk - onechat/blob - daemon.go
991edfaa2a444fd6802f90d4ae859bd75f611f20
[onechat] / daemon.go
1 package main
2
3 import (
4 "bufio"
5 "context"
6 "encoding/json"
7 "errors"
8 "fmt"
9 "io"
10 "log"
11 "net"
12 "os"
13 "path/filepath"
14 "sync"
15 "time"
16 )
17
18 const (
19 // daemonAgentDirName is the directory of the root agent in the data
20 // directory.
21 daemonAgentDirName = "agent"
22
23 // daemonClientEvents is the number of events that can wait for a slow
24 // client before the daemon drops it.
25 daemonClientEvents = 1024
26 )
27
28 // daemon runs "onechat daemon": it owns the root agent and its files and serves
29 // the clients on the Unix socket in the data directory until ctx ends. When
30 // telegram.json exists, it also runs the Telegram bot. It also keeps a linked
31 // WhatsApp device connected. It holds the lock of the token file, so it is the
32 // only process that refreshes the tokens and opens the WhatsApp session.
33 func daemon(ctx context.Context, out io.Writer) (err error) {
34 dir, err := dataDir()
35 if err != nil {
36 return err
37 }
38
39 // Check the socket path before the daemon writes a file, so that a data
40 // directory with a long path stays empty.
41 socket := filepath.Join(dir, socketFileName)
42 err = checkSocketPath(socket)
43 if err != nil {
44 return err
45 }
46
47 config, err := loadConfig(dir)
48 if err != nil {
49 return err
50 }
51
52 botSettings, hasBot, err := loadTelegramConfig(dir)
53 if err != nil {
54 return err
55 }
56
57 tokens, err := lockTokens(dir)
58 if err != nil {
59 return err
60 }
61
62 defer func() {
63 err = errors.Join(err, tokens.Close())
64 }()
65
66 whatsapp, err := openWhatsApp(ctx, dir, time.Now, log.New(out, "whatsapp: ", log.LstdFlags), whatsappNewClient)
67 if err != nil {
68 return err
69 }
70
71 defer func() {
72 err = errors.Join(err, whatsapp.Close())
73 }()
74
75 address, auth, err := modelEndpoint(config, dir)
76 if err != nil {
77 return err
78 }
79
80 agent, err := openAgent(filepath.Join(dir, daemonAgentDirName))
81 if err != nil {
82 return err
83 }
84
85 client := newModelClient(address, auth, agent.CacheKey(), filepath.Join(agent.dir, "usage.jsonl"))
86 memory, err := openMemory(ctx, agent.dir, time.Now, newCompactor(client, agent.Prompt, config.Compactor))
87 if err != nil {
88 return err
89 }
90
91 defer func() {
92 err = errors.Join(err, memory.Close())
93 }()
94
95 subagents := newSubagents(filepath.Join(agent.dir, subagentDirName), time.Now, subagentNameRandom)
96 defer func() {
97 err = errors.Join(err, subagents.Close())
98 }()
99
100 listener, err := daemonListenSocket(socket)
101 if err != nil {
102 return err
103 }
104
105 _, err = fmt.Fprintf(out, "onechat: %s on %s, %s at %s effort, in %s\n", agent.config.Name, listener.Addr(), agent.config.Model, agent.config.Effort, agent.config.Workdir)
106 if err != nil {
107 return errors.Join(err, listener.Close())
108 }
109
110 var telegram *TelegramBot
111 if hasBot {
112 telegram = newTelegramBot(newTelegramAPI(telegramAPIURL, botSettings.Token), botSettings.Owner, filepath.Join(agent.dir, telegramOffsetFileName), log.New(out, "telegram: ", log.LstdFlags))
113 _, err = fmt.Fprintln(out, "onechat: the Telegram bot waits for messages of its owner")
114 if err != nil {
115 return errors.Join(err, listener.Close())
116 }
117 }
118
119 return daemonServe(ctx, listener, telegram, whatsapp, func(notify func(Event)) (*TurnLoop, error) {
120 return newTurnLoop(agent, memory, subagents, client, time.Now, notify)
121 })
122 }
123
124 // DaemonHub holds the connected clients and sends each event to all of them. It
125 // runs the WhatsApp commands of the clients on whatsapp.
126 type DaemonHub struct {
127 whatsapp *WhatsApp
128
129 mutex sync.Mutex
130 clients map[*DaemonClient]bool
131 workers sync.WaitGroup
132 }
133
134 // Accept serves each client of listener until the listener closes. A new
135 // client gets a ready event with the name of the agent, and a busy event when a
136 // turn runs.
137 func (h *DaemonHub) Accept(ctx context.Context, listener net.Listener, l *TurnLoop) {
138 for {
139 conn, err := listener.Accept()
140 if err != nil {
141 return
142 }
143
144 client := &DaemonClient{
145 conn: conn,
146 events: make(chan Event, daemonClientEvents),
147 }
148
149 h.mutex.Lock()
150 h.clients[client] = true
151 h.mutex.Unlock()
152
153 h.workers.Add(2)
154 go h.write(client)
155 go h.read(ctx, client, l)
156
157 h.Send(client, Event{
158 Kind: "ready",
159 Text: l.agent.config.Name,
160 })
161
162 if l.Busy() {
163 h.Send(client, Event{
164 Kind: "busy",
165 })
166 }
167 }
168 }
169
170 // write sends the events of client as JSON lines until its channel closes.
171 func (h *DaemonHub) write(client *DaemonClient) {
172 defer h.workers.Done()
173
174 encoder := json.NewEncoder(client.conn)
175 for event := range client.events {
176 err := encoder.Encode(event)
177 if err != nil {
178 h.drop(client)
179 }
180 }
181 }
182
183 // read runs the requests of client until it closes its connection.
184 func (h *DaemonHub) read(ctx context.Context, client *DaemonClient, l *TurnLoop) {
185 defer h.workers.Done()
186 defer h.drop(client)
187
188 scanner := bufio.NewScanner(client.conn)
189 scanner.Buffer(make([]byte, 64*1024), socketRequestBytes)
190 for ctx.Err() == nil && scanner.Scan() {
191 var request SocketRequest
192 err := json.Unmarshal(scanner.Bytes(), &request)
193 if err == nil {
194 err = daemonRunRequest(l, h.whatsapp, request, func(text string) {
195 h.Send(client, Event{
196 Kind: "notice",
197 Text: text,
198 })
199 })
200 }
201
202 if err != nil {
203 h.Send(client, Event{
204 Kind: "error",
205 Text: err.Error(),
206 })
207 }
208 }
209 }
210
211 // Send gives event to one client, or drops the client when its events do not
212 // fit.
213 func (h *DaemonHub) Send(client *DaemonClient, event Event) {
214 h.mutex.Lock()
215 defer h.mutex.Unlock()
216
217 h.send(client, event)
218 }
219
220 // drop removes client and closes its connection.
221 func (h *DaemonHub) drop(client *DaemonClient) {
222 h.mutex.Lock()
223 defer h.mutex.Unlock()
224
225 h.remove(client)
226 }
227
228 // send gives event to client. The caller holds the mutex.
229 func (h *DaemonHub) send(client *DaemonClient, event Event) {
230 if !h.clients[client] {
231 return
232 }
233
234 select {
235 case client.events <- event:
236 default:
237 h.remove(client)
238 }
239 }
240
241 // remove removes client, closes its connection, and ends its writer. The
242 // caller holds the mutex.
243 func (h *DaemonHub) remove(client *DaemonClient) {
244 delete(h.clients, client)
245 client.once.Do(func() {
246 close(client.events)
247 _ = client.conn.Close()
248 })
249 }
250
251 // Broadcast gives event to every client.
252 func (h *DaemonHub) Broadcast(event Event) {
253 h.mutex.Lock()
254 defer h.mutex.Unlock()
255
256 for client := range h.clients {
257 h.send(client, event)
258 }
259 }
260
261 // CloseAll drops every client and waits for their readers and writers.
262 func (h *DaemonHub) CloseAll() {
263 h.mutex.Lock()
264 for client := range h.clients {
265 h.remove(client)
266 }
267 h.mutex.Unlock()
268
269 h.workers.Wait()
270 }
271
272 // DaemonClient is one connected client. Its events wait in a channel for the
273 // writer, so a slow client does not stop the agent.
274 type DaemonClient struct {
275 conn net.Conn
276 events chan Event
277 once sync.Once
278 }
279
280 // daemonListenSocket listens on the Unix socket at path, with mode 0600. The
281 // caller owns the chat, so a socket file at path is left from a daemon that
282 // ended, and daemonListenSocket removes it.
283 func daemonListenSocket(path string) (net.Listener, error) {
284 err := checkSocketPath(path)
285 if err != nil {
286 return nil, err
287 }
288
289 err = os.Remove(path)
290 if err != nil && !errors.Is(err, os.ErrNotExist) {
291 return nil, fmt.Errorf("remove old socket: %w", err)
292 }
293
294 listener, err := net.Listen("unix", path)
295 if err != nil {
296 return nil, fmt.Errorf("listen on socket: %w", err)
297 }
298
299 err = os.Chmod(path, 0o600)
300 if err != nil {
301 return nil, errors.Join(fmt.Errorf("set socket mode: %w", err), listener.Close())
302 }
303
304 return listener, nil
305 }
306
307 // daemonServe runs the loop that start makes and serves the clients of listener
308 // until ctx ends or the loop fails. It closes listener. It runs whatsapp, which
309 // logs its batches and notes in the loop and tells all clients its notices, and
310 // a failure of whatsapp stops the agent. When bot is not nil, it runs the bot
311 // too, and a failure of the bot stops the agent.
312 func daemonServe(ctx context.Context, listener net.Listener, bot *TelegramBot, whatsapp *WhatsApp, start func(notify func(Event)) (*TurnLoop, error)) error {
313 ctx, cancel := context.WithCancel(ctx)
314 defer cancel()
315
316 hub := &DaemonHub{
317 whatsapp: whatsapp,
318 clients: make(map[*DaemonClient]bool),
319 }
320
321 notify := hub.Broadcast
322 if bot != nil {
323 notify = func(event Event) {
324 hub.Broadcast(event)
325 bot.Notify(event)
326 }
327 }
328
329 l, err := start(notify)
330 if err != nil {
331 return errors.Join(err, listener.Close())
332 }
333
334 var workers sync.WaitGroup
335 workers.Add(2)
336 go func() {
337 defer workers.Done()
338
339 hub.Accept(ctx, listener, l)
340 }()
341
342 var whatsappErr error
343 go func() {
344 defer workers.Done()
345
346 whatsappErr = whatsapp.Run(ctx, l, notify)
347 cancel()
348 }()
349
350 var botErr error
351 if bot != nil {
352 workers.Add(1)
353 go func() {
354 defer workers.Done()
355
356 botErr = bot.Run(ctx, l, whatsapp)
357 cancel()
358 }()
359 }
360
361 err = l.Run(ctx)
362 if errors.Is(err, context.Canceled) {
363 err = nil
364 }
365
366 cancel()
367 err = errors.Join(err, listener.Close())
368 hub.CloseAll()
369 workers.Wait()
370
371 return errors.Join(err, whatsappErr, botErr)
372 }
373
374 // daemonRunRequest runs one request of a client. The answers of a WhatsApp
375 // command go to answer.
376 func daemonRunRequest(l *TurnLoop, whatsapp *WhatsApp, request SocketRequest, answer func(text string)) error {
377 switch request.Type {
378 case "send":
379 if request.Text == "" {
380 return errors.New("empty message")
381 }
382
383 _, err := l.Submit(terminalSource, request.Text)
384
385 return err
386 case "interrupt":
387 if !l.Interrupt() {
388 return errors.New("no turn runs")
389 }
390
391 return nil
392 case "whatsapp":
393 whatsapp.Command(request.Text, answer)
394
395 return nil
396 case "stop":
397 found, err := l.StopSubagent(request.Text)
398 if err == nil && !found {
399 err = fmt.Errorf("no subagent %s runs", request.Text)
400 }
401
402 return err
403 default:
404 return fmt.Errorf("unknown request %q", request.Type)
405 }
406 }