1package agent
2
3import (
4 "bufio"
5 "context"
6 "encoding/json"
7 "errors"
8 "fmt"
9 "io"
10 "os"
11 "os/signal"
12 "path/filepath"
13 "strings"
14 "sync"
15 "syscall"
16 "time"
17
18 "github.com/patriceckhart/zot/packages/agent/extensions"
19 "github.com/patriceckhart/zot/packages/agent/extproto"
20 "github.com/patriceckhart/zot/packages/agent/modes"
21 "github.com/patriceckhart/zot/packages/agent/skills"
22 "github.com/patriceckhart/zot/packages/agent/swarm"
23 "github.com/patriceckhart/zot/packages/agent/tools"
24 "github.com/patriceckhart/zot/packages/core"
25 "github.com/patriceckhart/zot/packages/provider"
26 "github.com/patriceckhart/zot/packages/provider/auth"
27 "github.com/patriceckhart/zot/packages/tui"
28)
29
30// interactiveExtHooks is a tiny adapter that lets the extension
31// manager call back into the Interactive instance built later in
32// runInteractive. The forward-declared *modes.Interactive is filled
33// in immediately after manager construction.
34type interactiveExtHooks struct {
35 ivPtr **modes.Interactive
36}
37
38func (h *interactiveExtHooks) iv() *modes.Interactive {
39 if h == nil || h.ivPtr == nil {
40 return nil
41 }
42 return *h.ivPtr
43}
44
45func (h *interactiveExtHooks) Notify(extName, level, message string) {
46 if iv := h.iv(); iv != nil {
47 iv.Notify(extName, level, message)
48 }
49}
50func (h *interactiveExtHooks) Submit(text string) {
51 if iv := h.iv(); iv != nil {
52 iv.SubmitOrQueue(text, nil)
53 }
54}
55func (h *interactiveExtHooks) SubmitSlash(text string) {
56 if iv := h.iv(); iv != nil {
57 iv.SubmitSlash(text)
58 }
59}
60func (h *interactiveExtHooks) Insert(text string) {
61 if iv := h.iv(); iv != nil {
62 iv.Insert(text)
63 }
64}
65func (h *interactiveExtHooks) Display(extName, text string) {
66 if iv := h.iv(); iv != nil {
67 iv.Display(extName, text)
68 }
69}
70func (h *interactiveExtHooks) ClearNotes(extName string) {
71 if iv := h.iv(); iv != nil {
72 iv.ClearNotes(extName)
73 }
74}
75func (h *interactiveExtHooks) OpenPanel(extName string, spec extproto.PanelSpec) {
76 if iv := h.iv(); iv != nil {
77 iv.OpenPanel(extName, spec)
78 }
79}
80func (h *interactiveExtHooks) UpdatePanel(extName, panelID, title string, lines []string, footer string) {
81 if iv := h.iv(); iv != nil {
82 iv.UpdatePanel(extName, panelID, title, lines, footer)
83 }
84}
85func (h *interactiveExtHooks) ClosePanel(extName, panelID string) {
86 if iv := h.iv(); iv != nil {
87 iv.ClosePanel(extName, panelID)
88 }
89}
90
91// extToolAdapter bridges *extensions.Manager to the
92// ExtensionToolSource interface declared in build.go (kept narrow to
93// avoid a build->extensions import cycle). One adapter instance per
94// run; used at every Resolve point so re-built agents pick up the
95// same set of extension tools.
96type extToolAdapter struct {
97 mgr *extensions.Manager
98}
99
100func (a *extToolAdapter) Tools() []ExtensionToolInfo {
101 infos := a.mgr.Tools()
102 out := make([]ExtensionToolInfo, len(infos))
103 for i, t := range infos {
104 out[i] = ExtensionToolInfo{
105 Extension: t.Extension,
106 Name: t.Name,
107 Description: t.Description,
108 Schema: t.Schema,
109 Deferred: t.Deferred,
110 }
111 }
112 return out
113}
114
115func (a *extToolAdapter) NewExtensionTool(info ExtensionToolInfo) core.Tool {
116 return extensions.NewTool(a.mgr, extensions.ToolInfo{
117 Extension: info.Extension,
118 Name: info.Name,
119 Description: info.Description,
120 Schema: info.Schema,
121 Deferred: info.Deferred,
122 })
123}
124
125// liveInteractiveAgent returns the agent currently owned by the TUI. The
126// fallback is only used before the Interactive has been constructed.
127//
128// Agent rebuilds, such as cross-provider model switches, replace the TUI's
129// agent without replacing the startup pointer held by runInteractive. Session
130// operations must therefore resolve the live agent at the time of the action.
131func liveInteractiveAgent(iv *modes.Interactive, fallback *core.Agent) *core.Agent {
132 if iv != nil {
133 return iv.Agent()
134 }
135 return fallback
136}
137
138func trimMessagesForResume(msgs []provider.Message, keepTail int) []provider.Message {
139 if keepTail <= 0 || len(msgs) <= keepTail {
140 return provider.RepairOrphanedToolResults(msgs)
141 }
142 var out []provider.Message
143 start := len(msgs) - keepTail
144 // Preserve the synthetic compaction summary when present so an
145 // already-compacted session stays compacted after resume.
146 if len(msgs) > 0 && msgs[0].Meta["compaction"] == "true" && start > 1 {
147 out = append(out, msgs[0])
148 }
149 // Avoid hydrating a tail that starts with orphan tool_result rows;
150 // provider APIs require those to be paired with an earlier tool_use.
151 for start < len(msgs) && msgs[start].Role == provider.RoleTool {
152 start++
153 }
154 out = append(out, msgs[start:]...)
155 return provider.RepairOrphanedToolResults(out)
156}
157
158// fanoutAgentEvent translates a core.AgentEvent into the wire-format
159// EventFromHost and pushes it through the extension manager. Only
160// the events that have a clear extension-facing meaning are
161// forwarded; internal-only ones (text_delta, tool_progress) are
162// dropped to keep the per-extension stream sane.
163func fanoutAgentEvent(mgr *extensions.Manager, ev core.AgentEvent) {
164 if mgr == nil {
165 return
166 }
167 switch e := ev.(type) {
168 case core.EvTurnStart:
169 mgr.EmitEvent(extproto.EventFromHost{Event: "turn_start", Step: e.Step})
170 case core.EvToolCall:
171 mgr.EmitEvent(extproto.EventFromHost{
172 Event: "tool_call", ToolID: e.ID, ToolName: e.Name, ToolArgs: e.Args,
173 })
174 case core.EvAssistantMessage:
175 // Concat the visible text portions of the message; binary
176 // blocks (tool_use, etc.) are skipped because subscribers
177 // usually want a string they can grep / display.
178 var text string
179 for _, c := range e.Message.Content {
180 if tb, ok := c.(provider.TextBlock); ok {
181 text += tb.Text
182 }
183 }
184 mgr.EmitEvent(extproto.EventFromHost{Event: "assistant_message", Text: text})
185 case core.EvTurnEnd:
186 ev := extproto.EventFromHost{Event: "turn_end", Stop: string(e.Stop)}
187 if e.Err != nil {
188 ev.Error = e.Err.Error()
189 }
190 mgr.EmitEvent(ev)
191 }
192}
193
194// Run is the top-level entrypoint for the zot binary.
195func Run(rawArgs []string, version string) error {
196 // Apply network configuration before any subcommand can make an HTTP
197 // request. Standard proxy environment variables retain precedence.
198 applyConfiguredHTTPProxy()
199
200 // Subcommand router: `zot bot ...` is handled separately so the
201 // generic flag parser doesn't reject "bot" as a positional arg.
202 if handled, err := runBotCommand(rawArgs, version); handled {
203 return err
204 }
205 if handled, err := runExtCommand(rawArgs, version); handled {
206 return err
207 }
208 if handled, err := runUpdateCommand(rawArgs, version); handled {
209 return err
210 }
211 if handled, err := runZotfileCommand(rawArgs, version); handled {
212 return err
213 }
214 return runWithArgsRaw(rawArgs, version)
215}
216
217func runWithArgsRaw(rawArgs []string, version string) error {
218 // `zot rpc` is shorthand for `zot --rpc` so third-party apps can
219 // spawn the binary with a clean argv. Strip the leading 'rpc'
220 // token and let the rest flow through the normal arg parser.
221 if len(rawArgs) > 0 && rawArgs[0] == "rpc" {
222 rawArgs = append([]string{"--rpc"}, rawArgs[1:]...)
223 }
224
225 args, err := ParseArgs(rawArgs)
226 if err != nil {
227 PrintHelp(version)
228 return err
229 }
230 if args.Help {
231 PrintHelp(version)
232 return nil
233 }
234 if args.Version {
235 fmt.Println("zot", version)
236 return nil
237 }
238 // Model catalog: load any cached discovery data before we inspect
239 // the model list (list-models, print/json, interactive).
240 prepareRuntimeCatalog()
241
242 if args.ListModels {
243 printModels()
244 return nil
245 }
246
247 return runWithArgs(args, version)
248}
249
250func runWithArgs(args Args, version string) error {
251 ctx := context.Background()
252 switch args.Mode {
253 case ModePrint:
254 return runPrintMode(ctx, args, version)
255 case ModeStream:
256 return runStreamMode(ctx, args, version)
257 case ModeJSON:
258 return runJSONMode(ctx, args, version)
259 case ModeRPC:
260 return runRPCMode(ctx, args, version)
261 case ModeSwarmAgent:
262 return runSwarmAgentMode(ctx, args, version)
263 default:
264 return runInteractive(ctx, args, version)
265 }
266}
267
268// ---- print / json modes: require credentials, run single-shot ----
269
270// nonInteractiveExtHooks is the HostHooks impl used by print / json
271// modes. They have no TUI, so notify / display go to stderr and
272// submit / insert are no-ops (the extension can't steer a
273// single-shot run once it's in flight anyway).
274type nonInteractiveExtHooks struct{}
275
276func (nonInteractiveExtHooks) Notify(ext, level, message string) {
277 fmt.Fprintf(os.Stderr, "[%s] %s: %s\n", ext, level, message)
278}
279func (nonInteractiveExtHooks) Submit(string) {}
280func (nonInteractiveExtHooks) SubmitSlash(string) {}
281func (nonInteractiveExtHooks) Insert(string) {}
282func (nonInteractiveExtHooks) Display(string, string) {}
283func (nonInteractiveExtHooks) ClearNotes(string) {}
284func (nonInteractiveExtHooks) OpenPanel(string, extproto.PanelSpec) {}
285func (nonInteractiveExtHooks) UpdatePanel(string, string, string, []string, string) {}
286func (nonInteractiveExtHooks) ClosePanel(string, string) {}
287
288// setupNonInteractiveExtensions loads --ext paths and (unless
289// --no-ext) runs discovery. Returns the manager so the caller can
290// wire tools into the resolved registry, and a cleanup closure to
291// defer. Mirrors the interactive-mode setup minus the TUI hooks.
292func setupNonInteractiveExtensions(ctx context.Context, args Args, r *Resolved, version string) (*extensions.Manager, func()) {
293 extMgr := extensions.New(ZotHome(), r.CWD, version, r.Provider, r.Model, nonInteractiveExtHooks{})
294 for _, e := range extMgr.LoadExplicit(ctx, args.Exts) {
295 fmt.Fprintln(os.Stderr, "extension load:", e)
296 }
297 if !args.NoExt {
298 for _, e := range extMgr.Discover(ctx) {
299 fmt.Fprintln(os.Stderr, "extension load:", e)
300 }
301 }
302 extMgr.WaitForReady(3 * time.Second)
303 r.MergeExtensionTools(&extToolAdapter{mgr: extMgr})
304 extMgr.EmitEvent(extproto.EventFromHost{Event: "session_start"})
305 return extMgr, func() { extMgr.Stop(2 * time.Second) }
306}
307
308// wireNonInteractiveAgentExtHooks installs the same BeforeToolExecute
309// / BeforeTurn / BeforeAssistantMessage / OnEvent hooks the
310// interactive path wires up, so extensions get their normal
311// event-intercept surface in print / json / rpc flows too.
312func wireNonInteractiveAgentExtHooks(ctx context.Context, ag *core.Agent, extMgr *extensions.Manager) {
313 if ag == nil || extMgr == nil {
314 return
315 }
316 ag.BeforeToolExecute = func(call provider.ToolCallBlock) (bool, string, json.RawMessage) {
317 res := extMgr.InterceptToolCall(ctx, call.ID, call.Name, call.Arguments)
318 if res.Block {
319 return false, res.Reason, nil
320 }
321 return true, "", res.ModifiedArgs
322 }
323 ag.BeforeTurn = func(step int) (bool, string) {
324 res := extMgr.InterceptTurnStart(ctx, step)
325 return !res.Block, res.Reason
326 }
327 ag.BeforeAssistantMessage = func(text string) (bool, string, string) {
328 res := extMgr.InterceptAssistantMessage(ctx, text)
329 if res.Block {
330 return false, res.Reason, ""
331 }
332 return true, "", res.ReplaceText
333 }
334 ag.OnEvent = func(ev core.AgentEvent) { fanoutAgentEvent(extMgr, ev) }
335}
336
337type printStats struct {
338 Provider string `json:"provider"`
339 Model string `json:"model"`
340 PromptTokens int `json:"prompt_tokens"`
341 ReasoningTokens *int `json:"reasoning_tokens"`
342 GeneratedOutputTokens int `json:"generated_output_tokens"`
343 ElapsedMS int64 `json:"elapsed_ms"`
344}
345
346func writePrintStats(path, providerID, model string, usage provider.Usage, elapsed time.Duration) error {
347 var reasoning *int
348 generated := usage.OutputTokens
349 if usage.ReasoningTokensKnown {
350 count := usage.ReasoningTokens
351 reasoning = &count
352 generated -= count
353 if generated < 0 {
354 generated = 0
355 }
356 }
357 stats := printStats{
358 Provider: providerID,
359 Model: model,
360 PromptTokens: usage.InputTokens + usage.CacheReadTokens + usage.CacheWriteTokens,
361 ReasoningTokens: reasoning,
362 GeneratedOutputTokens: generated,
363 ElapsedMS: elapsed.Milliseconds(),
364 }
365 data, err := json.MarshalIndent(stats, "", " ")
366 if err != nil {
367 return fmt.Errorf("encode stats: %w", err)
368 }
369 data = append(data, '\n')
370 if err := os.WriteFile(path, data, 0o644); err != nil {
371 return fmt.Errorf("write stats %q: %w", path, err)
372 }
373 return nil
374}
375
376func runPrintMode(ctx context.Context, args Args, version string) error {
377 if args.NoYolo {
378 fmt.Fprintln(os.Stderr, "warning: --no-yolo has no effect in print mode (no interactive prompt available); tools will run without confirmation")
379 }
380 r, err := Resolve(args, true)
381 if err != nil {
382 return err
383 }
384 extMgr, stopExt := setupNonInteractiveExtensions(ctx, args, &r, version)
385 defer stopExt()
386
387 ag := r.NewAgent()
388 wireNonInteractiveAgentExtHooks(ctx, ag, extMgr)
389 sess, _ := openOrCreateSession(args, r, ag, version)
390 defer sess.Close()
391
392 prompt := args.Prompt
393 if prompt == "" {
394 piped, _ := readAllStdin()
395 prompt = strings.TrimSpace(piped)
396 }
397 if prompt == "" {
398 return fmt.Errorf("print mode requires a prompt (arg or stdin)")
399 }
400
401 start := len(ag.Messages())
402 if err := runZotfileStartupPre(ctx, args.StartupPre, r.CWD, r.Sandbox, ag, nil, os.Stderr); err != nil {
403 return err
404 }
405 reloadResourcesAfterStartupPre(ctx, args, extMgr, r.Sandbox, ag)
406 started := time.Now()
407 usage, err := modes.RunPrint(ctx, ag, prompt, nil, os.Stdout)
408 elapsed := time.Since(started)
409 WriteNewTranscript(ag, sess, start)
410 if err != nil {
411 return err
412 }
413 if args.StatsPath != "" {
414 return writePrintStats(args.StatsPath, r.Provider, r.Model, usage, elapsed)
415 }
416 return nil
417}
418
419func runStreamMode(ctx context.Context, args Args, version string) error {
420 if args.NoYolo {
421 fmt.Fprintln(os.Stderr, "warning: --no-yolo has no effect in stream mode (no interactive prompt available); tools will run without confirmation")
422 }
423 r, err := Resolve(args, true)
424 if err != nil {
425 return err
426 }
427 extMgr, stopExt := setupNonInteractiveExtensions(ctx, args, &r, version)
428 defer stopExt()
429
430 ag := r.NewAgent()
431 wireNonInteractiveAgentExtHooks(ctx, ag, extMgr)
432 sess, _ := openOrCreateSession(args, r, ag, version)
433 defer sess.Close()
434
435 prompt := args.Prompt
436 if prompt == "" {
437 piped, _ := readAllStdin()
438 prompt = strings.TrimSpace(piped)
439 }
440 if prompt == "" {
441 return fmt.Errorf("stream mode requires a prompt (arg or stdin)")
442 }
443
444 start := len(ag.Messages())
445 preSink, finishPre := newStreamTextSink(os.Stdout)
446 if err := runZotfileStartupPre(ctx, args.StartupPre, r.CWD, r.Sandbox, ag, preSink, os.Stderr); err != nil {
447 finishPre()
448 return err
449 }
450 finishPre()
451 reloadResourcesAfterStartupPre(ctx, args, extMgr, r.Sandbox, ag)
452 err = modes.RunStream(ctx, ag, prompt, nil, os.Stdout)
453 WriteNewTranscript(ag, sess, start)
454 return err
455}
456
457func newStreamTextSink(out io.Writer) (func(core.AgentEvent), func()) {
458 var streamed, wroteText, lastWasNL bool
459 writeText := func(text string) {
460 if text == "" {
461 return
462 }
463 _, _ = fmt.Fprint(out, text)
464 if syncer, ok := out.(interface{ Sync() error }); ok {
465 _ = syncer.Sync()
466 }
467 wroteText = true
468 lastWasNL = strings.HasSuffix(text, "\n")
469 }
470 sink := func(ev core.AgentEvent) {
471 switch e := ev.(type) {
472 case core.EvAssistantStart:
473 streamed = false
474 case core.EvTextDelta:
475 streamed = true
476 writeText(e.Delta)
477 case core.EvAssistantMessage:
478 if streamed {
479 return
480 }
481 var text strings.Builder
482 for _, content := range e.Message.Content {
483 if block, ok := content.(provider.TextBlock); ok {
484 if text.Len() > 0 {
485 text.WriteString("\n")
486 }
487 text.WriteString(block.Text)
488 }
489 }
490 writeText(text.String())
491 }
492 }
493 finish := func() {
494 if wroteText && !lastWasNL {
495 writeText("\n")
496 }
497 }
498 return sink, finish
499}
500
501func runJSONMode(ctx context.Context, args Args, version string) error {
502 if args.NoYolo {
503 fmt.Fprintln(os.Stderr, "warning: --no-yolo has no effect in json mode (no interactive prompt available); tools will run without confirmation")
504 }
505 r, err := Resolve(args, true)
506 if err != nil {
507 return err
508 }
509 extMgr, stopExt := setupNonInteractiveExtensions(ctx, args, &r, version)
510 defer stopExt()
511
512 ag := r.NewAgent()
513 wireNonInteractiveAgentExtHooks(ctx, ag, extMgr)
514 sess, _ := openOrCreateSession(args, r, ag, version)
515 defer sess.Close()
516
517 prompt := args.Prompt
518 if prompt == "" {
519 piped, _ := readAllStdin()
520 prompt = strings.TrimSpace(piped)
521 }
522 if prompt == "" {
523 return fmt.Errorf("json mode requires a prompt (arg or stdin)")
524 }
525
526 start := len(ag.Messages())
527 enc := json.NewEncoder(os.Stdout)
528 preSink := func(ev core.AgentEvent) {
529 _ = enc.Encode(modes.EventToJSON(ev))
530 }
531 if err := runZotfileStartupPre(ctx, args.StartupPre, r.CWD, r.Sandbox, ag, preSink, os.Stderr); err != nil {
532 _ = enc.Encode(map[string]any{"type": "error", "message": err.Error()})
533 return err
534 }
535 reloadResourcesAfterStartupPre(ctx, args, extMgr, r.Sandbox, ag)
536 err = modes.RunJSON(ctx, ag, prompt, nil, os.Stdout)
537 WriteNewTranscript(ag, sess, start)
538 return err
539}
540
541// runZotfileStartupPre runs entry.pre before the main non-interactive prompt.
542// "!command" values execute via BashTool; other values are sent as an agent turn.
543// shellOut receives live shell chunks (typically os.Stderr); sink receives
544// agent events for plain-text pre (stream mode wires a live text sink).
545func runZotfileStartupPre(ctx context.Context, pre, cwd string, sandbox *tools.Sandbox, ag *core.Agent, sink func(core.AgentEvent), shellOut io.Writer) error {
546 pre = strings.TrimSpace(pre)
547 if pre == "" {
548 return nil
549 }
550 if cmd, ok := modes.ShellEscapeCommand(pre); ok {
551 raw, err := json.Marshal(map[string]any{"command": cmd})
552 if err != nil {
553 return err
554 }
555 if shellOut != nil {
556 fmt.Fprintf(shellOut, "$ %s\n", cmd)
557 if f, ok := shellOut.(interface{ Sync() error }); ok {
558 _ = f.Sync()
559 }
560 }
561 progress := func(chunk string) {
562 if shellOut == nil || chunk == "" {
563 return
564 }
565 fmt.Fprint(shellOut, chunk)
566 if f, ok := shellOut.(interface{ Sync() error }); ok {
567 _ = f.Sync()
568 }
569 }
570 bash := &tools.BashTool{CWD: cwd, Sandbox: sandbox}
571 res, err := bash.Execute(ctx, raw, progress)
572 if err != nil {
573 return fmt.Errorf("zotfile entry.pre: %w", err)
574 }
575 if res.IsError {
576 var sb strings.Builder
577 for _, c := range res.Content {
578 if tb, ok := c.(provider.TextBlock); ok {
579 sb.WriteString(tb.Text)
580 }
581 }
582 msg := strings.TrimSpace(sb.String())
583 if msg == "" {
584 msg = "command failed"
585 }
586 return fmt.Errorf("zotfile entry.pre: %s", msg)
587 }
588 return nil
589 }
590 if ag == nil {
591 return fmt.Errorf("zotfile entry.pre requires an agent for non-shell prompts")
592 }
593 if sink == nil {
594 sink = func(core.AgentEvent) {}
595 }
596 return ag.Prompt(ctx, pre, nil, sink)
597}
598
599// refreshAgentToolsAndPrompt re-resolves tools (including rediscovered
600// skills and currently loaded extension tools) and updates the live
601// agent's registry and system prompt. Used after /reload-ext and after
602// zotfile entry.pre installs new skills or extensions.
603// mutateRegistry, if non-nil, can inject session-specific tools (e.g. swarm_spawn).
604func refreshAgentToolsAndPrompt(args Args, sharedSandbox *tools.Sandbox, extToolAdapter ExtensionToolSource, ag *core.Agent, mutateRegistry func(core.Registry) core.Registry) {
605 if ag == nil {
606 return
607 }
608 resolved, err := Resolve(args, true)
609 if err != nil {
610 return
611 }
612 if sharedSandbox != nil {
613 resolved.UseSandbox(sharedSandbox)
614 }
615 if extToolAdapter != nil {
616 resolved.MergeExtensionTools(extToolAdapter)
617 }
618 reg := resolved.ToolRegistry
619 if mutateRegistry != nil {
620 reg = mutateRegistry(reg)
621 }
622 ag.SetTools(reg)
623 ag.System = resolved.SystemPrompt
624}
625
626// reloadResourcesAfterStartupPre reloads extensions (if any) and
627// refreshes the agent tool registry + system prompt so skills/extensions
628// installed by entry.pre are visible to the following turn.
629func reloadResourcesAfterStartupPre(ctx context.Context, args Args, extMgr *extensions.Manager, sharedSandbox *tools.Sandbox, ag *core.Agent) {
630 if strings.TrimSpace(args.StartupPre) == "" || ag == nil {
631 return
632 }
633 adapter := &extToolAdapter{mgr: extMgr}
634 if extMgr != nil {
635 _ = extMgr.Reload(ctx, 2*time.Second)
636 }
637 refreshAgentToolsAndPrompt(args, sharedSandbox, adapter, ag, nil)
638}
639
640// ---- interactive mode: opens the TUI even without credentials ----
641
642func runInteractive(ctx context.Context, args Args, version string) error {
643 // Resolve WITHOUT requiring credentials.
644 r, err := Resolve(args, false)
645 if err != nil {
646 return err
647 }
648
649 authStore := AuthStoreFor()
650 mgr := auth.NewManager(authStore)
651 defer mgr.Close()
652
653 // Keep the sandbox pointer stable across agent rebuilds (login / model
654 // switch). The Interactive UI toggles the lock via this pointer, and
655 // rebuilt tool instances must share the same one so the lock sticks.
656 sharedSandbox := r.Sandbox
657
658 // Build the extension manager BEFORE the agent so we can fold
659 // extension-defined tools into the registry. Forward-declare iv so
660 // the host hooks adapter can dereference it after construction.
661 var iv *modes.Interactive
662 extHooks := &interactiveExtHooks{ivPtr: &iv}
663 extMgr := extensions.New(ZotHome(), r.CWD, version, r.Provider, r.Model, extHooks)
664 var startupExtensionErrors []string
665 // --ext paths first so they win against installed extensions of
666 // the same name (loadOne's first-write-wins semantics).
667 for _, e := range extMgr.LoadExplicit(ctx, args.Exts) {
668 startupExtensionErrors = append(startupExtensionErrors, e.Error())
669 }
670 // --no-ext skips the global + project-local discovery scan;
671 // explicit --ext paths above are still honoured so you can run
672 // "only this extension" with --no-ext --ext ./x.
673 if !args.NoExt {
674 for _, e := range extMgr.Discover(ctx) {
675 startupExtensionErrors = append(startupExtensionErrors, e.Error())
676 }
677 }
678 // Wait briefly for extensions to flush their initial register_tool
679 // frames before we build the agent's tool registry. Half a second
680 // is plenty for any extension that's actually well-behaved; ones
681 // that don't send a ready frame eat the full grace and proceed.
682 // 3s is the per-extension grace period for the ready frame.
683 // Native binaries are instant; runtimes like `npx tsx` take ~1.5s
684 // from cold cache. The wait is tight only for extensions that
685 // haven't sent ready by then; ones that signalled earlier release
686 // the wait immediately.
687 extMgr.WaitForReady(3 * time.Second)
688 defer extMgr.Stop(2 * time.Second)
689
690 extToolAdapter := &extToolAdapter{mgr: extMgr}
691 r.MergeExtensionTools(extToolAdapter)
692
693 // Build the swarm supervisor BEFORE the agent so the auto-swarm
694 // tool can reference it during tool-registry construction. State
695 // lives under ZotHome/swarm so per-agent meta/events survive
696 // restarts; the user can hunt orphaned agents down with
697 // `git worktree list` if anything misbehaves.
698 //
699 // swarmMgr is also captured by loadSession / changeCWD closures
700 // further down the function, which is why we keep the variable
701 // in this outer scope rather than scoping it tighter.
702 var swarmMgr *swarm.Swarm
703 swarmMgr = swarm.New(swarm.Config{
704 Root: filepath.Join(ZotHome(), "swarm"),
705 RepoRoot: r.CWD,
706 ResolveCredential: func(ctx context.Context, providerID string) (swarm.Credential, error) {
707 if providerID == "ollama" {
708 return swarm.Credential{Value: "ollama", Method: "apikey"}, nil
709 }
710 credential, method, accountID, err := ResolveCredentialFullContext(ctx, providerID, "")
711 if err != nil {
712 // Providers backed only by a local endpoint do not need a key.
713 // Leave stdin untouched and let the child resolve that endpoint.
714 if !CredentialAvailable(providerID) {
715 return swarm.Credential{}, nil
716 }
717 return swarm.Credential{}, err
718 }
719 return swarm.Credential{Value: credential, Method: method, AccountID: accountID}, nil
720 },
721 })
722 // Pull any previously-spawned agents off disk so the dashboard
723 // shows them as detached and the user can resume / remove them.
724 _, _ = swarmMgr.Reload()
725
726 // onSpawnedSwarm is the OnSpawned callback the swarm_spawn tool
727 // fires after every successful spawn. It hands the agent off to
728 // the running Interactive so the watcher can flush a summary back
729 // into chat when all sub-agents finish. Reads `iv` lazily because
730 // the Interactive is constructed after the agent.
731 onSpawnedSwarm := func(a *swarm.Agent, task string) {
732 if iv != nil {
733 iv.TrackSwarmAgent(a, task)
734 }
735 }
736
737 // Inject the swarm_spawn auto-swarm tool only when /settings ->
738 // auto-swarm is currently enabled. Registering it unconditionally
739 // leaves the model trying to call it (and getting a polite error)
740 // even when the user has switched the feature off. The /settings
741 // toggle live-mutates the running agent's registry separately so
742 // flipping the flag mid-session takes effect on the next turn.
743 injectSwarmSpawn := func(reg core.Registry) core.Registry {
744 if reg == nil {
745 return reg
746 }
747 if !AutoSwarmEnabled() {
748 return reg
749 }
750 reg["swarm_spawn"] = &tools.SwarmSpawnTool{
751 Swarm: swarmMgr,
752 Enabled: AutoSwarmEnabled,
753 DefaultModel: func() string { return r.Model },
754 DefaultProvider: func() string { return r.Provider },
755 OnSpawned: onSpawnedSwarm,
756 }
757 return reg
758 }
759 injectSwarmSpawn(r.ToolRegistry)
760
761 // Confirmation gate: when --no-yolo is on, the agent must ask
762 // the user before every tool call. In interactive mode the TUI
763 // provides the Confirmer; in print/json/rpc modes there's no
764 // way to prompt, so the gate is constructed with a nil inner
765 // which auto-refuses every call with a helpful reason.
766 var confirmGate *core.ConfirmGate
767 if args.NoYolo {
768 confirmGate = core.NewConfirmGate(nil) // set below for interactive
769 }
770
771 // Capture current args in a closure so BuildAgent can re-resolve
772 // after a successful login (picks up the newly stored credential).
773 wireAgentExt := func(a *core.Agent) *core.Agent {
774 if a == nil {
775 return a
776 }
777 a.BeforeToolExecute = func(call provider.ToolCallBlock) (bool, string, json.RawMessage) {
778 // Guards run before confirmation because they may rewrite the
779 // arguments. The user must approve the effective call that will
780 // actually execute, not the model's original arguments.
781 r := extMgr.InterceptToolCall(ctx, call.ID, call.Name, call.Arguments)
782 if r.Block {
783 return false, r.Reason, nil
784 }
785 effectiveArgs := call.Arguments
786 if len(r.ModifiedArgs) > 0 && json.Valid(r.ModifiedArgs) {
787 effectiveArgs = r.ModifiedArgs
788 }
789 if confirmGate != nil {
790 var content strings.Builder
791 if tool, err := a.Tools.Get(call.Name); err == nil {
792 if previewer, ok := tool.(core.ToolPreviewer); ok {
793 preview, err := previewer.Preview(ctx, effectiveArgs)
794 if err != nil {
795 return false, err.Error(), nil
796 }
797 for _, block := range preview.Content {
798 if text, ok := block.(provider.TextBlock); ok {
799 if content.Len() > 0 {
800 content.WriteString("\n")
801 }
802 content.WriteString(text.Text)
803 }
804 }
805 }
806 }
807 ok, reason, _ := confirmGate.CheckToolCall(core.ToolCallConfirmation{
808 ID: call.ID,
809 Name: call.Name,
810 Summary: core.BuildPreview(effectiveArgs, 120),
811 Content: content.String(),
812 })
813 if !ok {
814 return false, reason, nil
815 }
816 }
817 return true, "", r.ModifiedArgs
818 }
819 a.BeforeTurn = func(step int) (bool, string) {
820 r := extMgr.InterceptTurnStart(ctx, step)
821 return !r.Block, r.Reason
822 }
823 a.BeforeAssistantMessage = func(text string) (bool, string, string) {
824 r := extMgr.InterceptAssistantMessage(ctx, text)
825 if r.Block {
826 return false, r.Reason, ""
827 }
828 return true, "", r.ReplaceText
829 }
830 a.OnEvent = func(ev core.AgentEvent) { fanoutAgentEvent(extMgr, ev) }
831 return a
832 }
833
834 buildAgent := func() (*core.Agent, string, string, error) {
835 resolved, err := Resolve(args, true)
836 if err != nil {
837 return nil, "", "", err
838 }
839 resolved.UseSandbox(sharedSandbox)
840 resolved.MergeExtensionTools(extToolAdapter)
841 injectSwarmSpawn(resolved.ToolRegistry)
842 return wireAgentExt(resolved.NewAgent()), resolved.Provider, resolved.Model, nil
843 }
844
845 // Rebuild agent with an explicit provider/model override.
846 buildAgentFor := func(providerOverride, modelOverride string) (*core.Agent, string, string, error) {
847 next := args
848 if providerOverride != "" {
849 next.Provider = providerOverride
850 }
851 if modelOverride != "" {
852 next.Model = modelOverride
853 }
854 resolved, err := Resolve(next, true)
855 if err != nil {
856 return nil, "", "", err
857 }
858 resolved.UseSandbox(sharedSandbox)
859 resolved.MergeExtensionTools(extToolAdapter)
860 injectSwarmSpawn(resolved.ToolRegistry)
861 return wireAgentExt(resolved.NewAgent()), resolved.Provider, resolved.Model, nil
862 }
863
864 // Rebuild agent for the rescue picker after a recoverable failure.
865 // Unlike buildAgentFor, this drops launch-time --api-key and
866 // --base-url overrides because those are typically the cause of the
867 // rescue (a bad key, a typo'd base URL, or a corporate gateway that
868 // only the originally-picked provider needed). Re-resolving without
869 // them lets the rescue retry use env vars / auth.json / provider
870 // defaults the way zot would have without the overrides.
871 buildAgentForRescue := func(providerOverride, modelOverride string) (*core.Agent, string, string, error) {
872 next := args
873 next.APIKey = ""
874 next.BaseURL = ""
875 if providerOverride != "" {
876 next.Provider = providerOverride
877 }
878 if modelOverride != "" {
879 next.Model = modelOverride
880 }
881 resolved, err := Resolve(next, true)
882 if err != nil {
883 return nil, "", "", err
884 }
885 resolved.UseSandbox(sharedSandbox)
886 resolved.MergeExtensionTools(extToolAdapter)
887 injectSwarmSpawn(resolved.ToolRegistry)
888 return wireAgentExt(resolved.NewAgent()), resolved.Provider, resolved.Model, nil
889 }
890
891 var ag *core.Agent
892 if r.HasCredential() {
893 ag = wireAgentExt(r.NewAgent())
894 }
895
896 // /reload-ext callback: after the manager has respawned every
897 // extension, re-resolve the tool registry (built-ins + freshly-
898 // registered extension tools) and swap it onto the current
899 // agent in-place. The current agent may have been replaced by a
900 // /model swap since spawn, so re-read the live `ag` on each
901 // invocation.
902 extMgr.SetOnReload(func() {
903 current := liveInteractiveAgent(iv, ag)
904 if current == nil {
905 return
906 }
907 refreshAgentToolsAndPrompt(args, sharedSandbox, extToolAdapter, current, injectSwarmSpawn)
908 })
909
910 // Fire session_start once we know the manager's running.
911 extMgr.EmitEvent(extproto.EventFromHost{Event: "session_start"})
912
913 var sess *core.Session
914 var sessBaselineMsgs int // messages already on disk when current session opened
915 // persistMu guards sess + sessBaselineMsgs against concurrent access
916 // from the agent loop's per-message persistence hook (runs on the
917 // agent goroutine) and the TUI's session swap / flush callbacks
918 // (run on the TUI goroutine). Without this, a /sessions swap that
919 // races with a finishing turn could double-write or lose messages.
920 var persistMu sync.Mutex
921 if !args.NoSess && ag != nil {
922 sess, _ = openOrCreateSession(args, r, ag, version)
923 if ag != nil {
924 sessBaselineMsgs = len(ag.Messages())
925 }
926 }
927 defer func() {
928 persistMu.Lock()
929 defer persistMu.Unlock()
930 if sess != nil {
931 sess.Close()
932 }
933 }()
934
935 // persistMessage is the per-message hook bound to the agent. It
936 // appends each new transcript message to the live session as soon
937 // as it lands, so a kill / closed terminal / OS crash costs at
938 // most the in-flight turn instead of the whole session. The
939 // baseline counter advances in lock-step so the exit-time flush
940 // doesn't double-write rows already on disk.
941 persistMessage := func(m provider.Message) {
942 persistMu.Lock()
943 defer persistMu.Unlock()
944 if sess == nil {
945 return
946 }
947 if err := sess.AppendMessage(m); err == nil {
948 sessBaselineMsgs++
949 }
950 }
951 persistUsage := func(cum provider.Usage) {
952 persistMu.Lock()
953 defer persistMu.Unlock()
954 if sess == nil {
955 return
956 }
957 _ = sess.AppendUsage(cum, cum)
958 }
959 persistCompaction := func(messages []provider.Message) {
960 persistMu.Lock()
961 defer persistMu.Unlock()
962 if sess == nil {
963 return
964 }
965 if err := sess.AppendCompaction(messages); err == nil {
966 sessBaselineMsgs = len(messages)
967 }
968 }
969 wireAgentPersist := func(a *core.Agent) *core.Agent {
970 if a == nil {
971 return a
972 }
973 a.OnMessageAppended = persistMessage
974 a.OnUsage = persistUsage
975 a.OnTranscriptCompacted = persistCompaction
976 return a
977 }
978 wireAgentPersist(ag)
979
980 // Re-wrap the build closures so any agent constructed by the TUI
981 // (login, /model swap to a different provider) also gets the
982 // persistence hooks. Without this, switching provider would
983 // silently revert to the old in-memory-only behaviour.
984 baseBuildAgent := buildAgent
985 buildAgent = func() (*core.Agent, string, string, error) {
986 a, p, m, err := baseBuildAgent()
987 return wireAgentPersist(a), p, m, err
988 }
989 baseBuildAgentFor := buildAgentFor
990 buildAgentFor = func(providerOverride, modelOverride string) (*core.Agent, string, string, error) {
991 a, p, m, err := baseBuildAgentFor(providerOverride, modelOverride)
992 return wireAgentPersist(a), p, m, err
993 }
994 baseBuildAgentForRescue := buildAgentForRescue
995 buildAgentForRescue = func(providerOverride, modelOverride string) (*core.Agent, string, string, error) {
996 a, p, m, err := baseBuildAgentForRescue(providerOverride, modelOverride)
997 return wireAgentPersist(a), p, m, err
998 }
999
1000 // loadSession replaces the current session with the one at path and
1001 // hands its messages to the agent. Used by the /sessions picker.
1002 loadSession := func(path string) error {
1003 currentAg := liveInteractiveAgent(iv, ag)
1004 if currentAg == nil {
1005 return fmt.Errorf("no agent running; log in first")
1006 }
1007 newSess, msgs, err := core.OpenSession(path)
1008 if err != nil {
1009 return err
1010 }
1011 fullMsgCount := len(msgs)
1012 msgs = trimMessagesForResume(msgs, 100)
1013 persistMu.Lock()
1014 // Flush any unsaved messages to the old session before swapping.
1015 // Per-message persistence keeps sessBaselineMsgs current, so
1016 // this is a defensive no-op in the common case; it still
1017 // matters for the rare race where a turn just finished and
1018 // hadn't fired its hook yet.
1019 if sess != nil {
1020 writeNewTranscriptLocked(currentAg, sess, sessBaselineMsgs)
1021 _ = sess.Close()
1022 }
1023 sess = newSess
1024 currentAg.SetMessages(msgs)
1025 if cum, last, uerr := core.SessionUsageDetail(path); uerr == nil {
1026 currentAg.SeedCost(cum)
1027 currentAg.SeedLastTurnUsage(last)
1028 }
1029 // The live agent only receives a compact resume window, but
1030 // the session file remains intact. Keep the persistence
1031 // baseline at the original on-disk message count so future
1032 // turns append after the full session instead of duplicating
1033 // the hydrated tail.
1034 sessBaselineMsgs = fullMsgCount
1035 persistMu.Unlock()
1036 // Re-scope the swarm dashboard to the new session so /swarm
1037 // only shows agents this session spawned. swarmMgr may be nil
1038 // here if we haven't reached the construction site yet (it
1039 // shouldn't be, since the interactive loop is what triggers
1040 // loadSession, but the nil check is cheap insurance).
1041 if swarmMgr != nil && newSess != nil {
1042 swarmMgr.SetActiveSession(newSess.ID)
1043 }
1044 return nil
1045 }
1046
1047 // changeCWD switches the running session to a new working directory.
1048 // Wired into InteractiveConfig.ChangeCWD and invoked by the hidden
1049 // /cd slash command (which itself is only fired by the workspaces
1050 // extension's panel-key Enter handler today; the user can type /cd
1051 // directly but it's not in autocomplete / help / the README).
1052 //
1053 // Steps, in order:
1054 // 1. resolve + validate the new path (~ expansion, abs/rel)
1055 // 2. close the current session, flushing pending messages
1056 // 3. mutate captured args.CWD + r.CWD so future buildAgent
1057 // calls bind to the new cwd
1058 // 4. re-root the shared sandbox («·«) so /jail follows the
1059 // session into the new cwd instead of widening or silently
1060 // dropping
1061 // 5. rebuild the agent via buildAgent() so tools, AGENTS.md
1062 // addendum, system prompt, sessions dir all bind correctly
1063 // 6. open a fresh session in the new cwd's bucket
1064 // 7. push the new state into the running Interactive
1065 // 8. re-scope the swarm dashboard to the freshly-opened session
1066 //
1067 // The /jail state is preserved verbatim: if the sandbox was locked
1068 // to the old cwd, it stays locked, just re-pointed at the new one.
1069 changeCWD := func(path string) error {
1070 if path == "" {
1071 return fmt.Errorf("empty path")
1072 }
1073 // ~ expansion.
1074 if path == "~" || strings.HasPrefix(path, "~/") {
1075 home, herr := os.UserHomeDir()
1076 if herr != nil || home == "" {
1077 return fmt.Errorf("cannot expand ~: %v", herr)
1078 }
1079 if path == "~" {
1080 path = home
1081 } else {
1082 path = filepath.Join(home, path[2:])
1083 }
1084 }
1085 if !filepath.IsAbs(path) {
1086 path = filepath.Join(args.CWD, path)
1087 }
1088 absPath, err := filepath.Abs(path)
1089 if err != nil {
1090 return err
1091 }
1092 info, err := os.Stat(absPath)
1093 if err != nil {
1094 return err
1095 }
1096 if !info.IsDir() {
1097 return fmt.Errorf("not a directory: %s", absPath)
1098 }
1099
1100 currentAg := ag
1101 if currentAg == nil {
1102 return fmt.Errorf("no agent running; log in first")
1103 }
1104
1105 // Close the current session before we drop the reference.
1106 // Per-message persistence keeps it current already; this is
1107 // a defensive flush + final fsync via Close.
1108 persistMu.Lock()
1109 if sess != nil {
1110 writeNewTranscriptLocked(currentAg, sess, sessBaselineMsgs)
1111 _ = sess.Close()
1112 sess = nil
1113 }
1114 sessBaselineMsgs = 0
1115 persistMu.Unlock()
1116
1117 // Mutate captured state so subsequent agent rebuilds and
1118 // session opens see the new cwd.
1119 wasJailed := sharedSandbox != nil && sharedSandbox.Locked()
1120 args.CWD = absPath
1121 r.CWD = absPath
1122 if args.PermissionSet != nil {
1123 expanded := args.PermissionSet.Expand(absPath, args.AgentDataDir)
1124 args.PermissionSet = &expanded
1125 }
1126 if sharedSandbox != nil {
1127 sharedSandbox.Root = absPath
1128 if wasJailed {
1129 sharedSandbox.Lock()
1130 } else {
1131 sharedSandbox.Unlock()
1132 }
1133 }
1134
1135 // Rebuild the agent so tools / AGENTS.md / system prompt
1136 // re-bind to the new cwd. buildAgent() reads from args + r.
1137 newAg, newProvider, newModel, berr := buildAgent()
1138 if berr != nil {
1139 return fmt.Errorf("rebuild agent: %v", berr)
1140 }
1141 ag = newAg
1142
1143 // Fresh session in the new cwd's bucket. We bypass
1144 // openOrCreateSession's --continue / --resume branches
1145 // because /cd's semantics are "start fresh here", matching
1146 // what relaunching `zot --cwd <path>` would do today.
1147 if !args.NoSess {
1148 newRoot := agentSessionsRoot(ZotHome(), args)
1149 core.PruneEmptySessions(newRoot, absPath)
1150 newSess, serr := core.NewSession(newRoot, absPath, newProvider, newModel, version)
1151 if serr != nil {
1152 return fmt.Errorf("open session in %s: %v", absPath, serr)
1153 }
1154 persistMu.Lock()
1155 sess = newSess
1156 sessBaselineMsgs = 0
1157 persistMu.Unlock()
1158 }
1159
1160 // Push the new state into the running Interactive.
1161 if iv != nil {
1162 startupPaths := instructionContextPaths(loadAgentsContext(absPath, ZotHome()))
1163 iv.ApplyChangedCWDWithStartupContext(newAg, newProvider, newModel, absPath, startupPaths)
1164 }
1165
1166 // Re-scope the swarm dashboard to the new session.
1167 if swarmMgr != nil && sess != nil {
1168 swarmMgr.SetActiveSession(sess.ID)
1169 }
1170 return nil
1171 }
1172
1173 term := tui.NewProcTerm()
1174
1175 // Kick off the async update check so the banner can appear when the
1176 // http response eventually arrives (usually <1s on cached DNS). Map
1177 // agent.UpdateInfo -> modes.UpdateInfo here to avoid a cyclic import.
1178 updateCh := make(chan modes.UpdateInfo, 1)
1179 go func() {
1180 defer close(updateCh)
1181 src := <-CheckForUpdateAsync(ZotHome(), version)
1182 updateCh <- modes.UpdateInfo{
1183 Current: src.Current,
1184 Latest: src.Latest,
1185 Available: src.Available,
1186 URL: src.URL,
1187 }
1188 }()
1189
1190 // Changelog: when the running version differs from the last
1191 // version whose release notes the user dismissed, fetch the
1192 // release body from GitHub and have the TUI show it once. On
1193 // first-ever launch (no prior LastChangelogShown), seed the
1194 // stored version silently — don't dump release notes at someone
1195 // who just installed.
1196 changelogCh := make(chan modes.ChangelogPayload, 1)
1197 go func() {
1198 defer close(changelogCh)
1199 cfg, _ := LoadConfig()
1200 if cfg.LastChangelogShown == "" {
1201 SeedChangelogVersion(version)
1202 return
1203 }
1204 if !ShouldShowChangelog(version, cfg) {
1205 return
1206 }
1207 info := <-FetchChangelogAsync(version)
1208 if info.Body == "" {
1209 return
1210 }
1211 // For dev builds (0.0.0), skip if the latest release was
1212 // already shown (stored by the dismiss callback).
1213 if version == "0.0.0" && info.Version == cfg.LastChangelogShown {
1214 return
1215 }
1216 changelogCh <- modes.ChangelogPayload{
1217 Version: info.Version,
1218 Body: info.Body,
1219 URL: info.URL,
1220 }
1221 }()
1222
1223 initialCfg, _ := LoadConfig()
1224 quickModelShortcuts := make([]modes.QuickModelShortcut, len(initialCfg.QuickModelShortcuts))
1225 for idx, s := range initialCfg.QuickModelShortcuts {
1226 quickModelShortcuts[idx] = modes.QuickModelShortcut{Provider: s.Provider, Model: s.Model}
1227 }
1228 theme, _, themeErr := tui.DetectThemeWithCustom(ZotHome(), initialCfg.Theme, 80*time.Millisecond)
1229 if themeErr != nil {
1230 fmt.Fprintln(os.Stderr, "theme load:", themeErr)
1231 if initialCfg.Theme != "" && !tui.ThemeExists(ZotHome(), initialCfg.Theme) {
1232 initialCfg.Theme = ""
1233 _ = SaveConfig(initialCfg)
1234 }
1235 }
1236
1237 // swarmMgr was constructed and reloaded earlier (before the agent
1238 // build, so the auto-swarm tool could capture it). Here we just
1239 // scope the dashboard to the active host session so /swarm only
1240 // shows agents this session spawned (and any pre-upgrade unscoped
1241 // agents — see SnapshotAll docs). Updated again whenever the
1242 // user swaps sessions via loadSession below.
1243 if sess != nil {
1244 swarmMgr.SetActiveSession(sess.ID)
1245 }
1246 // Best-effort shutdown on interactive exit: stop all running
1247 // agents so they don't outlive their parent zot.
1248 defer swarmMgr.StopAll()
1249
1250 var startupSkills []*skills.Skill
1251 if r.SkillTool != nil {
1252 startupSkills = r.SkillTool.Skills()
1253 }
1254
1255 iv = modes.NewInteractive(modes.InteractiveConfig{
1256 Terminal: term,
1257 Theme: theme,
1258 InlineImagesEnabled: initialCfg.InlineImagesEnabled,
1259 AutoSwarmEnabled: initialCfg.AutoSwarmEnabled,
1260 AutoCompactThreshold: initialCfg.AutoCompactThreshold,
1261 JailByDefault: initialCfg.JailByDefault,
1262 QuickModelShortcuts: quickModelShortcuts,
1263 RecursiveFileSuggest: initialCfg.RecursiveFileSuggest,
1264 RespectGitignore: initialCfg.RespectGitignore,
1265 CompactMode: initialCfg.CompactMode,
1266 TUIInputStyle: initialCfg.TUIInputStyle,
1267 TUIStatusPosition: initialCfg.TUIStatusPosition,
1268 TUIWorkingPosition: initialCfg.TUIWorkingPosition,
1269 ThemeName: initialCfg.Theme,
1270 FlatTools: initialCfg.FlatToolRender(),
1271 CompactUser: initialCfg.CompactUserInput(),
1272 ExtensionThemes: extMgr.ThemeOptions,
1273 AutoSwarmSystemAddendum: AutoSwarmSystemAddendum,
1274 SettingsStore: configSettingsStore{},
1275 Model: r.Model,
1276 Provider: r.Provider,
1277 AuthMethod: r.AuthMethod,
1278 BaseURL: r.BaseURL,
1279 Reasoning: r.Reasoning,
1280 SystemPrompt: r.SystemPrompt,
1281 Tools: r.ToolRegistry,
1282 MaxSteps: r.MaxSteps,
1283 CWD: r.CWD,
1284 StartupAgentName: args.AgentName,
1285 StartupContextPaths: instructionContextPaths(r.ContextFiles),
1286 StartupExtensionNames: startupExtensionNames(extMgr.All()),
1287 StartupExtensionErrors: startupExtensionErrors,
1288 StartupSkillNames: startupSkillNames(startupSkills),
1289 ShowInstructionsAtStartup: initialCfg.ShowInstructionsAtStartup,
1290 ZotHome: ZotHome(),
1291 SessionsRoot: agentSessionsRoot(ZotHome(), args),
1292 Version: version,
1293 UpdateInfoChan: updateCh,
1294 Sandbox: sharedSandbox,
1295 Agent: ag,
1296 InitialInput: args.Prompt,
1297 StartupPre: args.StartupPre,
1298 OnStartupPreDone: func() {
1299 // entry.pre often installs skills/extensions; rediscover them
1300 // before the user starts a model turn.
1301 current := liveInteractiveAgent(iv, ag)
1302 if extMgr != nil {
1303 _ = extMgr.Reload(context.Background(), 2*time.Second)
1304 }
1305 if current != nil {
1306 refreshAgentToolsAndPrompt(args, sharedSandbox, extToolAdapter, current, injectSwarmSpawn)
1307 }
1308 },
1309 AuthManager: mgr,
1310 LlamaCPPConfig: ResolveLlamaCPPConfig,
1311 RefreshLlamaCPPModels: RefreshLlamaCPPModels,
1312 BuildAgent: buildAgent,
1313 SetKimiCLIFallbackDisabled: SetKimiCLIFallbackDisabled,
1314 BuildAgentFor: buildAgentFor,
1315 BuildAgentForRescue: buildAgentForRescue,
1316 LoggedInProviders: func() []string {
1317 var out []string
1318 seen := map[string]bool{}
1319 for _, p := range knownProviders {
1320 if CredentialAvailable(p) && !seen[p] {
1321 out = append(out, p)
1322 seen[p] = true
1323 }
1324 }
1325 // Include custom providers that have credentials stored.
1326 for p := range provider.CustomProviders() {
1327 if CredentialAvailable(p) && !seen[p] {
1328 out = append(out, p)
1329 seen[p] = true
1330 }
1331 }
1332 // Ollama models are always available (no auth needed).
1333 if !seen["ollama"] {
1334 out = append(out, "ollama")
1335 }
1336 return out
1337 },
1338 LoadSession: loadSession,
1339 ChangeCWD: changeCWD,
1340 CurrentSessionPath: func() string {
1341 if sess == nil {
1342 return ""
1343 }
1344 return sess.Path
1345 },
1346 FlushSession: func() {
1347 // Append any not-yet-persisted agent messages to the
1348 // current session file, then advance the baseline so
1349 // the final WriteNewTranscript at exit doesn't write
1350 // duplicates. Per-message persistence keeps the on-
1351 // disk file current already, so this is mostly a
1352 // defensive flush — still needed for /session export
1353 // to guarantee the exported bytes include the very
1354 // last in-flight turn.
1355 currentAg := iv.Agent()
1356 if currentAg == nil {
1357 return
1358 }
1359 persistMu.Lock()
1360 defer persistMu.Unlock()
1361 if sess == nil {
1362 return
1363 }
1364 writeNewTranscriptLocked(currentAg, sess, sessBaselineMsgs)
1365 sessBaselineMsgs = len(currentAg.Messages())
1366 },
1367 Extensions: extMgr,
1368 Swarm: swarmMgr,
1369 ChangelogChan: changelogCh,
1370 OnChangelogDismiss: func() {
1371 // For dev builds (0.0.0) store the actual release version
1372 // so the same changelog doesn't show again next launch.
1373 // For real builds, store the binary version.
1374 v := version
1375 if v == "0.0.0" {
1376 if iv != nil && iv.ChangelogVersion() != "" {
1377 v = iv.ChangelogVersion()
1378 }
1379 }
1380 _ = MarkChangelogShown(v)
1381 },
1382 SkillSnapshot: func() []*skills.Skill {
1383 if args.NoSkill {
1384 // --no-skill: nothing for the picker to show.
1385 return nil
1386 }
1387 // Re-discover so the picker reflects edits made during
1388 // the session. Cheap; SKILL.md files are small. Filter
1389 // out built-in skills — they're hidden from user-facing
1390 // surfaces because they're implementation detail; the
1391 // model still sees them through the system-prompt
1392 // manifest and the skill tool.
1393 userHome, _ := os.UserHomeDir()
1394 list, _ := skills.Discover(ZotHome(), r.CWD, userHome, args.WithSkills)
1395 return skills.VisibleSkills(list)
1396 },
1397 NoYolo: args.NoYolo,
1398 ConfirmGate: confirmGate,
1399 PersistModel: func(providerName, model string) {
1400 // Update config.json so next launch uses the same pick.
1401 cfg, _ := LoadConfig()
1402 cfg.Provider = providerName
1403 cfg.Model = model
1404 _ = SaveConfig(cfg)
1405 // Update the active session's meta so resume picks this up.
1406 if sess != nil {
1407 _ = sess.UpdateModel(providerName, model)
1408 }
1409 },
1410 })
1411
1412 // Bind the interactive TUI as the Confirmer. We deferred this
1413 // until now because the gate is constructed before the TUI
1414 // (the BeforeToolExecute closure captures it). SetConfirmer
1415 // is mutex-guarded on the gate so this is safe.
1416 if confirmGate != nil {
1417 confirmGate.SetConfirmer(&confirmationEventConfirmer{
1418 inner: iv,
1419 emit: extMgr.EmitEvent,
1420 })
1421 }
1422
1423 // Signal-driven flush: a SIGTERM / SIGHUP to the zot process
1424 // (closed terminal window, system shutdown, kill) used to lose
1425 // the entire in-memory transcript because the deferred post-Run
1426 // flush below never ran. Per-message persistence above covers
1427 // most of it; this handler writes any in-flight remainder and
1428 // then exits the process so we don't double-paint over a
1429 // broken terminal that the TUI's restore deferreds can no
1430 // longer fix from a signal context.
1431 //
1432 // SIGINT is intentionally NOT handled here — the TUI consumes
1433 // Ctrl+C as a regular key event for cancel/clear semantics, and
1434 // installing a SIGINT notifier here would swallow it.
1435 sigCh := make(chan os.Signal, 1)
1436 signal.Notify(sigCh, syscall.SIGTERM, syscall.SIGHUP)
1437 defer signal.Stop(sigCh)
1438 go func() {
1439 _, ok := <-sigCh
1440 if !ok {
1441 return
1442 }
1443 if finalAg := iv.Agent(); finalAg != nil {
1444 persistMu.Lock()
1445 if sess != nil {
1446 writeNewTranscriptLocked(finalAg, sess, sessBaselineMsgs)
1447 sessBaselineMsgs = len(finalAg.Messages())
1448 _ = sess.Close()
1449 sess = nil
1450 }
1451 persistMu.Unlock()
1452 }
1453 // Exit cleanly. Re-raising the signal would skip os.Exit's
1454 // at-exit hooks; explicit exit is fine because we've already
1455 // flushed the only at-risk state (the session file).
1456 os.Exit(0)
1457 }()
1458
1459 runErr := iv.Run(ctx)
1460
1461 // Flush final transcript to session (only if we had / ended up with an agent).
1462 if finalAg := iv.Agent(); finalAg != nil {
1463 persistMu.Lock()
1464 if sess != nil {
1465 writeNewTranscriptLocked(finalAg, sess, sessBaselineMsgs)
1466 sessBaselineMsgs = len(finalAg.Messages())
1467 }
1468 persistMu.Unlock()
1469 }
1470 return runErr
1471}
1472
1473func agentSessionsRoot(root string, args Args) string {
1474 if args.AgentName == "" {
1475 return root
1476 }
1477 return filepath.Join(root, "sessions", "agents", safeAgentName(args.AgentName))
1478}
1479
1480// openOrCreateSession returns a session for the run. sess may be nil
1481// with a nil error if session persistence is disabled.
1482func openOrCreateSession(args Args, r Resolved, ag *core.Agent, version string) (*core.Session, error) {
1483 if args.NoSess {
1484 return nil, nil
1485 }
1486 // Sweep meta-only files left over from older zot versions (and from
1487 // any session that crashed before its first AppendMessage). Cheap;
1488 // reads the first few bytes of each file in the cwd's session dir.
1489 sessionsRoot := agentSessionsRoot(ZotHome(), args)
1490 core.PruneEmptySessions(sessionsRoot, args.CWD)
1491 var (
1492 s *core.Session
1493 msgs []provider.Message
1494 err error
1495 )
1496 switch {
1497 case args.Session != "":
1498 s, msgs, err = core.OpenSession(args.Session)
1499 // The swarm-agent child passes a fixed --session path that
1500 // may not exist yet on first Spawn. Treat ENOENT as "create
1501 // a fresh session AT THIS PATH" so the conversation actually
1502 // gets persisted; without this fallback the swarm child runs
1503 // with sess==nil and every Resume re-starts with no memory
1504 // of the prior turns. Other openers (--continue / --resume /
1505 // the picker) never see ENOENT here because they only choose
1506 // paths that already exist on disk.
1507 if err != nil && errors.Is(err, os.ErrNotExist) {
1508 s, err = core.NewSessionAtPath(args.Session, args.CWD, r.Provider, r.Model, version)
1509 msgs = nil
1510 }
1511 case args.Continue:
1512 latest := core.LatestSession(sessionsRoot, args.CWD)
1513 if latest != "" {
1514 s, msgs, err = core.OpenSession(latest)
1515 }
1516 case args.Resume:
1517 picked, perr := pickSession(sessionsRoot, args.CWD)
1518 if perr != nil {
1519 return nil, perr
1520 }
1521 if picked != "" {
1522 s, msgs, err = core.OpenSession(picked)
1523 }
1524 }
1525 if err != nil {
1526 return nil, err
1527 }
1528 if s != nil {
1529 ag.SetMessages(msgs)
1530 if cum, last, uerr := core.SessionUsageDetail(s.Path); uerr == nil {
1531 ag.SeedCost(cum)
1532 ag.SeedLastTurnUsage(last)
1533 }
1534 return s, nil
1535 }
1536 return core.NewSession(sessionsRoot, args.CWD, r.Provider, r.Model, version)
1537}
1538
1539func pickSession(root, cwd string) (string, error) {
1540 files := core.ListSessions(root, cwd)
1541 if len(files) == 0 {
1542 fmt.Fprintln(os.Stderr, "no sessions for", cwd)
1543 return "", nil
1544 }
1545 for i, f := range files {
1546 fmt.Fprintf(os.Stderr, " %2d) %s\n", i+1, f)
1547 }
1548 fmt.Fprint(os.Stderr, "pick #: ")
1549 rd := bufio.NewReader(os.Stdin)
1550 line, _ := rd.ReadString('\n')
1551 line = strings.TrimSpace(line)
1552 var n int
1553 if _, err := fmt.Sscanf(line, "%d", &n); err != nil || n < 1 || n > len(files) {
1554 return "", fmt.Errorf("invalid selection")
1555 }
1556 return files[n-1], nil
1557}
1558
1559// WriteNewTranscript appends only messages after index `from` from the
1560// agent's transcript to the session. Used by callers that don't hold
1561// the persistMu (non-interactive print/json modes which run a single
1562// turn under their own goroutine).
1563func WriteNewTranscript(ag *core.Agent, sess *core.Session, from int) {
1564 writeNewTranscriptLocked(ag, sess, from)
1565}
1566
1567// writeNewTranscriptLocked is the same as WriteNewTranscript. The
1568// suffix marks that interactive callers must hold persistMu when
1569// invoking it so concurrent appends from the agent loop don't race
1570// with this catch-up flush.
1571func writeNewTranscriptLocked(ag *core.Agent, sess *core.Session, from int) {
1572 if sess == nil || ag == nil {
1573 return
1574 }
1575 msgs := ag.Messages()
1576 for i := from; i < len(msgs); i++ {
1577 _ = sess.AppendMessage(msgs[i])
1578 }
1579 cum := ag.Cost()
1580 _ = sess.AppendUsage(cum, cum)
1581}
1582
1583func readAllStdin() (string, error) {
1584 fi, err := os.Stdin.Stat()
1585 if err != nil {
1586 return "", err
1587 }
1588 if (fi.Mode() & os.ModeCharDevice) != 0 {
1589 return "", nil
1590 }
1591 b, err := io.ReadAll(os.Stdin)
1592 return string(b), err
1593}
1594
1595func printModels() {
1596 models := provider.Active()
1597
1598 // Compute column widths from actual data so wide providers (e.g.
1599 // xiaomi-token-plan-sgp) and long bedrock model ids don't force the
1600 // `name` column off-screen. Floors mirror the historical layout so
1601 // short catalogs look the same as before.
1602 provW, idW, srcW := len("provider"), len("model id"), len("source")
1603 for _, m := range models {
1604 if w := len(m.Provider); w > provW {
1605 provW = w
1606 }
1607 if w := len(m.ID); w > idW {
1608 idW = w
1609 }
1610 source := m.Source
1611 if source == "" {
1612 source = "catalog"
1613 }
1614 if m.Speculative {
1615 source = "speculative"
1616 }
1617 if w := len(source); w > srcW {
1618 srcW = w
1619 }
1620 }
1621
1622 header := fmt.Sprintf("%-*s %-*s %8s %8s %s %-*s %s",
1623 provW, "provider",
1624 idW, "model id",
1625 "context", "max-out", "reasoning",
1626 srcW, "source",
1627 "name")
1628 fmt.Println(header)
1629
1630 for _, m := range models {
1631 reason := " "
1632 if m.Reasoning {
1633 reason = "✓"
1634 }
1635 source := m.Source
1636 if source == "" {
1637 source = "catalog"
1638 }
1639 if m.Speculative {
1640 source = "speculative"
1641 }
1642 fmt.Printf("%-*s %-*s %8d %8d %s %-*s %s\n",
1643 provW, m.Provider,
1644 idW, m.ID,
1645 m.ContextWindow, m.MaxOutput,
1646 reason,
1647 srcW, source,
1648 m.DisplayName)
1649 }
1650}