1// Package sdk is the public Go SDK for embedding the zut agent
2// runtime in third-party programs. It is the only stable, importable
3// surface the project exposes; everything under internal/ is subject
4// to change without notice.
5//
6// Lifecycle:
7//
8// rt, err := sdk.New(sdk.Config{Provider: "anthropic", Model: "claude-sonnet-4-5"})
9// if err != nil { ... }
10// defer rt.Close()
11//
12// events, err := rt.Prompt(ctx, "fix the failing test", nil)
13// if err != nil { ... }
14// for ev := range events {
15// fmt.Printf("%s: %+v\n", ev.Type(), ev)
16// }
17//
18// Concurrency model: one prompt at a time per Runtime. Spawn one
19// Runtime per project / cwd. The Cancel call interrupts the active
20// prompt; subsequent prompts work normally.
21//
22// For a non-Go consumer, run `zut rpc` and speak the same JSON
23// schema over stdin/stdout. See docs/rpc.md.
24package sdk
25
26import (
27 "context"
28 "errors"
29 "fmt"
30 "strings"
31 "sync"
32
33 "github.com/bnema/zut/packages/agent"
34 "github.com/bnema/zut/packages/agent/subagents"
35 agenttools "github.com/bnema/zut/packages/agent/tools"
36 "github.com/bnema/zut/packages/core"
37 "github.com/bnema/zut/packages/provider"
38)
39
40// Config configures a Runtime. All fields are optional; sensible
41// defaults are read from $ZUT_HOME/config.json, env vars, and the
42// resolver chain (the same one the cli uses).
43type Config struct {
44 // Provider is a configured provider id such as "anthropic", "openai",
45 // or "opencode-go". Empty = use the user's default from config.json or env.
46 Provider string
47
48 // Model is the model id. Empty = use the provider's default.
49 Model string
50
51 // CWD is the working directory the agent operates in. Tools are
52 // confined to this directory when Lock is true. Empty = current
53 // process cwd.
54 CWD string
55
56 // SystemPrompt overrides the built-in system prompt.
57 SystemPrompt string
58
59 // AppendSystemPrompt is appended to the built-in (or overridden)
60 // system prompt. Useful for project-specific instructions.
61 AppendSystemPrompt []string
62
63 // Reasoning sets the reasoning effort for models that support it
64 // ("minimum", "low", "medium", "high", "xhigh", "max"). Empty disables it.
65 Reasoning string
66
67 // Temperature sets the sampling temperature. Nil = provider default.
68 Temperature *float32
69
70 // MaxSteps caps the agent loop iterations per Prompt call.
71 // 0 leaves the loop unlimited.
72 MaxSteps int
73
74 // APIKey overrides the credential lookup chain.
75 APIKey string
76
77 // BaseURL overrides the provider base url (for tests / proxies).
78 BaseURL string
79
80 // Tools is the list of tools to enable. Nil/empty enables the ordinary
81 // built-ins: read, write, edit, bash, python, create_worktree, grep, glob, ast, lsp,
82 // update_goal, and update_plan. The public-web capability is excluded unless this list
83 // explicitly contains "web_search"; that selector enables web_search,
84 // web_open, web_find, and web_click. Pass NoTools=true to disable everything.
85 Tools []string
86
87 // NoTools disables every tool. Useful for chat-only embeddings.
88 NoTools bool
89
90 // Lock confines tools to CWD. Same effect as the /jail command.
91 Lock bool
92}
93
94// Runtime is one zut agent session. Safe for use from one goroutine
95// at a time per Runtime; create separate Runtimes for parallel work.
96type Runtime struct {
97 mu sync.Mutex
98 agent *core.Agent
99 provider string
100 model string
101 cwd string
102 // modelCatalog is a private metadata snapshot for dynamic providers.
103 modelCatalog []provider.Model
104 modelCatalogAuthoritative bool
105
106 // activeCancel is set while a Prompt or Compact is running.
107 activeCancel context.CancelFunc
108 activeDone chan struct{}
109
110 // closed signals that Close has been called.
111 closed bool
112}
113
114// New constructs a Runtime from cfg. Returns an error if no
115// credential is available for the requested provider.
116func New(cfg Config) (*Runtime, error) {
117 return NewContext(context.Background(), cfg)
118}
119
120// NewContext constructs a Runtime from cfg and uses ctx for credential
121// resolution and initial model catalog discovery. Canceling ctx stops setup
122// and any refresh started for this runtime.
123func NewContext(ctx context.Context, cfg Config) (*Runtime, error) {
124 if ctx == nil {
125 ctx = context.Background()
126 }
127 webSearchPolicy := subagents.WebSearchDeny
128 for _, name := range cfg.Tools {
129 if name == "web_search" {
130 webSearchPolicy = subagents.WebSearchAllow
131 break
132 }
133 }
134 args := agent.Args{
135 Mode: agent.ModeJSON, // headless
136 Provider: cfg.Provider,
137 Model: cfg.Model,
138 CWD: cfg.CWD,
139 APIKey: cfg.APIKey,
140 BaseURL: cfg.BaseURL,
141 SystemPrompt: cfg.SystemPrompt,
142 AppendSystemPrompt: cfg.AppendSystemPrompt,
143 Reasoning: cfg.Reasoning,
144 Temperature: cfg.Temperature,
145 MaxSteps: cfg.MaxSteps,
146 Tools: cfg.Tools,
147 NoTools: cfg.NoTools,
148 WebSearchPolicy: webSearchPolicy,
149 NoSess: true, // SDK callers manage persistence themselves
150 }
151 r, err := agent.ResolveSDK(ctx, args)
152 if err != nil {
153 return nil, err
154 }
155 if cfg.Lock && r.Sandbox != nil {
156 r.Sandbox.Lock()
157 }
158 ag := r.NewAgent()
159 return &Runtime{
160 agent: ag,
161 provider: r.Provider,
162 model: r.Model,
163 cwd: r.CWD,
164 modelCatalog: r.ModelCatalogSnapshot(),
165 modelCatalogAuthoritative: r.ModelCatalogIsAuthoritative(),
166 }, nil
167}
168
169// Provider returns the active provider id.
170func (r *Runtime) Provider() string { r.mu.Lock(); defer r.mu.Unlock(); return r.provider }
171
172// Model returns the active model id.
173func (r *Runtime) Model() string { r.mu.Lock(); defer r.mu.Unlock(); return r.model }
174
175// CWD returns the working directory the agent operates in.
176func (r *Runtime) CWD() string { r.mu.Lock(); defer r.mu.Unlock(); return r.cwd }
177
178// Messages returns a copy of the current transcript.
179func (r *Runtime) Messages() []Message {
180 if r.agent == nil {
181 return nil
182 }
183 src := r.agent.Messages()
184 out := make([]Message, len(src))
185 for i, m := range src {
186 out[i] = Message{
187 Role: string(m.Role),
188 Content: convertContent(m.Content),
189 Time: m.Time.Format("2006-01-02T15:04:05Z"),
190 }
191 }
192 return out
193}
194
195// SetMessages replaces the transcript. Use to seed history from a
196// session file or to clear with nil.
197func (r *Runtime) SetMessages(msgs []Message) {
198 if r.agent == nil {
199 return
200 }
201 out := make([]provider.Message, 0, len(msgs))
202 for _, m := range msgs {
203 out = append(out, provider.Message{
204 Role: provider.Role(m.Role),
205 Content: rebuildContent(m.Content),
206 })
207 }
208 r.agent.SetMessages(out)
209}
210
211// Cost returns the cumulative usage and dollar cost for this Runtime.
212func (r *Runtime) Cost() Usage {
213 if r.agent == nil {
214 return Usage{}
215 }
216 u := r.agent.Cost()
217 return Usage{
218 Input: u.InputTokens,
219 Output: u.OutputTokens,
220 Reasoning: sdkReasoningTokens(u),
221 CacheRead: u.CacheReadTokens,
222 CacheWrite: u.CacheWriteTokens,
223 CacheMeasuredPrompt: u.CacheMeasuredPromptTokens,
224 CacheMeasuredRead: u.CacheMeasuredReadTokens,
225 CostUSD: u.CostUSD,
226 }
227}
228
229// Prompt sends a user message and runs the agent loop. Returns a
230// channel that emits one Event per agent action, closed when the
231// turn finishes (cleanly or with error). Only one Prompt may be
232// active at a time per Runtime; concurrent calls return ErrBusy.
233func (r *Runtime) Prompt(ctx context.Context, text string, images []Image) (<-chan Event, error) {
234 r.mu.Lock()
235 if r.closed {
236 r.mu.Unlock()
237 return nil, ErrClosed
238 }
239 if r.activeCancel != nil {
240 r.mu.Unlock()
241 return nil, ErrBusy
242 }
243 if r.agent == nil {
244 r.mu.Unlock()
245 return nil, fmt.Errorf("sdk: no agent (login first via /login or set credentials)")
246 }
247 subCtx, cancel := context.WithCancel(ctx)
248 done := make(chan struct{})
249 r.activeCancel = cancel
250 r.activeDone = done
251 r.mu.Unlock()
252
253 imgBlocks := make([]provider.ImageBlock, 0, len(images))
254 for _, img := range images {
255 imgBlocks = append(imgBlocks, provider.ImageBlock{MimeType: img.MimeType, Data: img.Data})
256 }
257
258 out := make(chan Event, 16)
259 go func() {
260 defer close(out)
261 defer func() {
262 r.mu.Lock()
263 r.activeCancel = nil
264 r.activeDone = nil
265 close(done)
266 r.mu.Unlock()
267 }()
268 err := r.agent.Prompt(subCtx, text, imgBlocks, func(ev core.AgentEvent) {
269 out <- toEvent(ev)
270 })
271 if err != nil && !errors.Is(err, context.Canceled) {
272 out <- Event{Type: "error", Error: err.Error()}
273 }
274 }()
275 return out, nil
276}
277
278// Cancel interrupts the active Prompt, if any. Safe to call when no
279// prompt is in flight (no-op).
280func (r *Runtime) Cancel() {
281 r.mu.Lock()
282 cancel := r.activeCancel
283 r.mu.Unlock()
284 if cancel != nil {
285 cancel()
286 }
287}
288
289// Compact summarises the current transcript into a single synthetic
290// user message, freeing up context. Blocks until done.
291func (r *Runtime) Compact(ctx context.Context, customInstructions string) (CompactResult, error) {
292 if r.agent == nil {
293 return CompactResult{}, fmt.Errorf("sdk: no agent")
294 }
295 r.mu.Lock()
296 if r.activeCancel != nil {
297 r.mu.Unlock()
298 return CompactResult{}, ErrBusy
299 }
300 subCtx, cancel := context.WithCancel(ctx)
301 done := make(chan struct{})
302 r.activeCancel = cancel
303 r.activeDone = done
304 r.mu.Unlock()
305 defer func() {
306 r.mu.Lock()
307 r.activeCancel = nil
308 r.activeDone = nil
309 close(done)
310 r.mu.Unlock()
311 }()
312
313 summary, err := r.agent.Compact(subCtx, 4, nil)
314 if err != nil {
315 return CompactResult{}, err
316 }
317 return CompactResult{Summary: summary, Messages: r.Messages()}, nil
318}
319
320// SetModel switches the active model. Same provider only; for cross-
321// provider switches, create a new Runtime.
322func (r *Runtime) SetModel(model string) error {
323 r.mu.Lock()
324 defer r.mu.Unlock()
325 if r.agent == nil {
326 return fmt.Errorf("sdk: no agent")
327 }
328 if r.activeCancel != nil {
329 return ErrBusy
330 }
331 model = strings.TrimSpace(model)
332 if model == "" {
333 return fmt.Errorf("sdk: model must not be empty")
334 }
335
336 var metadata provider.Model
337 found := false
338 if r.provider == provider.ProviderOpenCodeGo {
339 for _, candidate := range r.modelCatalog {
340 if candidate.ID == model {
341 metadata = candidate
342 found = true
343 break
344 }
345 }
346 if !found && r.modelCatalogAuthoritative {
347 return fmt.Errorf("unknown model %q (provider=%q)", model, r.provider)
348 }
349 if !found {
350 metadata = provider.DynamicOpenCodeGoModel(r.provider, model, "")
351 }
352 } else {
353 var err error
354 metadata, err = provider.FindModel(r.provider, model)
355 if err != nil {
356 return err
357 }
358 found = true
359 }
360 if found || r.provider == provider.ProviderOpenCodeGo {
361 if setter, ok := r.agent.Client.(provider.ModelMetadataSetter); ok {
362 setter.SetModelMetadata(metadata)
363 }
364 if metadata.ContextWindow > 0 {
365 r.agent.ContextWindow = metadata.ContextWindow
366 }
367 if metadata.MaxOutput > 0 {
368 r.agent.MaxTokens = metadata.MaxOutput
369 }
370 }
371 r.agent.Model = model
372 r.model = model
373 return nil
374}
375
376// SetReasoning changes the reasoning level used by subsequent prompts.
377func (r *Runtime) SetReasoning(level string) error {
378 r.mu.Lock()
379 defer r.mu.Unlock()
380 if r.agent == nil {
381 return fmt.Errorf("sdk: no agent")
382 }
383 reasoning := provider.NormalizeReasoning(level)
384 switch reasoning {
385 case "", "minimum", "low", "medium", "high", "xhigh", "max":
386 r.agent.Reasoning = reasoning
387 return nil
388 default:
389 return fmt.Errorf("sdk: reasoning must be off|minimum|low|medium|high|xhigh|max")
390 }
391}
392
393// State returns a snapshot of the runtime's current state. Useful for
394// driving UIs.
395func (r *Runtime) State() State {
396 r.mu.Lock()
397 defer r.mu.Unlock()
398 return State{
399 Provider: r.provider,
400 Model: r.model,
401 Reasoning: r.agent.Reasoning,
402 CWD: r.cwd,
403 Busy: r.activeCancel != nil,
404 MessageCount: len(r.Messages()),
405 }
406}
407
408// Close releases resources held by the Runtime. Cancels any active
409// prompt. Subsequent calls return ErrClosed.
410func (r *Runtime) Close() error {
411 r.mu.Lock()
412 if r.closed {
413 r.mu.Unlock()
414 return nil
415 }
416 r.closed = true
417 cancel := r.activeCancel
418 done := r.activeDone
419 r.mu.Unlock()
420 if cancel != nil {
421 cancel()
422 }
423 if done != nil {
424 <-done
425 }
426 r.mu.Lock()
427 ag := r.agent
428 r.mu.Unlock()
429 if ag != nil {
430 return agenttools.CloseLSPManagers(ag.ToolsSnapshot())
431 }
432 return nil
433}
434
435// ListModels returns every model known to the runtime for the
436// current provider (catalog + live discovery if cached).
437func (r *Runtime) ListModels() []ModelInfo {
438 r.mu.Lock()
439 providerName := r.provider
440 models := append([]provider.Model(nil), r.modelCatalog...)
441 r.mu.Unlock()
442 if providerName != provider.ProviderOpenCodeGo {
443 models = provider.ModelsForProvider(providerName)
444 }
445 out := make([]ModelInfo, 0, len(models))
446 for _, m := range models {
447 out = append(out, ModelInfo{
448 ID: m.ID,
449 Provider: m.Provider,
450 ContextWindow: m.ContextWindow,
451 MaxOutput: m.MaxOutput,
452 Reasoning: m.Reasoning,
453 })
454 }
455 return out
456}
457
458// ---- public errors ----
459
460// ErrBusy is returned when a Prompt or Compact is started while
461// another is in flight on the same Runtime.
462var ErrBusy = errors.New("sdk: runtime is busy")
463
464// ErrClosed is returned by methods on a Runtime after Close.
465var ErrClosed = errors.New("sdk: runtime is closed")