1package provider
2
3import (
4 "bytes"
5 "context"
6 "encoding/base64"
7 "encoding/json"
8 "fmt"
9 "io"
10 "net/http"
11 "strings"
12 "time"
13)
14
15const openaiDefaultBaseURL = "https://api.openai.com"
16
17type openaiClient struct {
18 apiKey string
19 baseURL string
20 name string
21 oauth bool // when true, apiKey actually holds an OAuth access token
22 headers map[string]string
23 http *http.Client
24}
25
26// NewOpenAI creates an OpenAI client using an API key. baseURL may be empty.
27func NewOpenAI(apiKey, baseURL string) Client {
28 if baseURL == "" {
29 baseURL = openaiDefaultBaseURL
30 }
31 return &openaiClient{
32 apiKey: apiKey,
33 baseURL: strings.TrimRight(baseURL, "/"),
34 name: "openai",
35 http: &http.Client{Timeout: 0},
36 }
37}
38
39// NewOpenAIOAuth creates an OpenAI client using a subscription OAuth access token.
40// The token is sent as an HTTP Bearer credential on the standard chat/completions endpoint.
41func NewOpenAIOAuth(accessToken, baseURL string) Client {
42 if baseURL == "" {
43 baseURL = openaiDefaultBaseURL
44 }
45 return &openaiClient{
46 apiKey: accessToken,
47 baseURL: strings.TrimRight(baseURL, "/"),
48 name: "openai",
49 oauth: true,
50 http: &http.Client{Timeout: 0},
51 }
52}
53
54// NewKimi creates a Kimi/Moonshot client. Kimi's chat API is OpenAI-compatible.
55func NewKimi(apiKey, baseURL string) Client {
56 return NewKimiWithHeaders(apiKey, baseURL, nil)
57}
58
59// NewDeepSeek creates a DeepSeek client. DeepSeek's chat API is
60// OpenAI-compatible at https://api.deepseek.com/v1.
61func NewDeepSeek(apiKey, baseURL string) Client {
62 if baseURL == "" {
63 baseURL = "https://api.deepseek.com/v1"
64 }
65 return &openaiClient{
66 apiKey: apiKey,
67 baseURL: strings.TrimRight(baseURL, "/"),
68 name: "deepseek",
69 http: &http.Client{Timeout: 0},
70 }
71}
72
73// NewKimiWithHeaders creates a Kimi/Moonshot client with extra headers.
74// Subscription tokens from Kimi Code need the official CLI's X-Msh-* headers.
75func NewKimiWithHeaders(apiKey, baseURL string, headers map[string]string) Client {
76 if baseURL == "" {
77 baseURL = "https://api.kimi.com/coding/v1"
78 }
79 return &openaiClient{
80 apiKey: apiKey,
81 baseURL: strings.TrimRight(baseURL, "/"),
82 name: "kimi",
83 headers: headers,
84 http: &http.Client{Timeout: 0},
85 }
86}
87
88func (c *openaiClient) Name() string {
89 if c.name != "" {
90 return c.name
91 }
92 return "openai"
93}
94
95// ---- wire types ----
96
97type oaiContentText struct {
98 Type string `json:"type"` // "text"
99 Text string `json:"text"`
100}
101
102type oaiContentImage struct {
103 Type string `json:"type"` // "image_url"
104 ImageURL struct {
105 URL string `json:"url"`
106 } `json:"image_url"`
107}
108
109type oaiToolCallFn struct {
110 Name string `json:"name"`
111 Arguments string `json:"arguments"` // JSON string
112}
113
114type oaiToolCall struct {
115 ID string `json:"id"`
116 Type string `json:"type"` // "function"
117 Function oaiToolCallFn `json:"function"`
118}
119
120type oaiMessage struct {
121 Role string `json:"role"`
122 Content interface{} `json:"content,omitempty"` // string or []block
123 Name string `json:"name,omitempty"`
124 ToolCalls []oaiToolCall `json:"tool_calls,omitempty"`
125 ToolCallID string `json:"tool_call_id,omitempty"`
126 // ReasoningContent carries the model's chain-of-thought summary
127 // alongside an assistant tool-call message. Required by Kimi's
128 // chat completions endpoint when thinking is enabled and the
129 // assistant message contains a tool call; OpenAI ignores it.
130 ReasoningContent string `json:"reasoning_content,omitempty"`
131}
132
133type oaiTool struct {
134 Type string `json:"type"` // "function"
135 Function struct {
136 Name string `json:"name"`
137 Description string `json:"description,omitempty"`
138 Parameters json.RawMessage `json:"parameters"`
139 } `json:"function"`
140}
141
142type oaiStreamOptions struct {
143 IncludeUsage bool `json:"include_usage"`
144}
145
146type oaiRequest struct {
147 Model string `json:"model"`
148 Messages []oaiMessage `json:"messages"`
149 Tools []oaiTool `json:"tools,omitempty"`
150 ToolChoice string `json:"tool_choice,omitempty"`
151 Stream bool `json:"stream"`
152 StreamOptions *oaiStreamOptions `json:"stream_options,omitempty"`
153 Temperature *float32 `json:"temperature,omitempty"`
154 MaxTokens *int `json:"max_tokens,omitempty"`
155 MaxCompletionTok *int `json:"max_completion_tokens,omitempty"`
156 ReasoningEffort string `json:"reasoning_effort,omitempty"`
157}
158
159// ---- request building ----
160
161func (c *openaiClient) buildRequest(req Request) (*oaiRequest, error) {
162 // Look up the model by id across all providers (not just openai)
163 // because the OpenAI client is also used for ollama and other
164 // OpenAI-compatible backends.
165 m, err := FindModel("", req.Model)
166 if err != nil {
167 // Unknown model: use sensible defaults so local/custom
168 // models still work without a catalog entry.
169 m = Model{
170 ID: req.Model,
171 ContextWindow: 32768,
172 MaxOutput: 8192,
173 }
174 }
175 out := &oaiRequest{
176 Model: req.Model,
177 Stream: true,
178 StreamOptions: &oaiStreamOptions{IncludeUsage: true},
179 Temperature: req.Temperature,
180 }
181
182 maxTok := req.MaxTokens
183 if maxTok <= 0 {
184 maxTok = m.MaxOutput
185 }
186 if m.Reasoning {
187 out.MaxCompletionTok = &maxTok
188 if req.Reasoning != "" {
189 out.ReasoningEffort = strings.ToLower(req.Reasoning)
190 }
191 } else {
192 out.MaxTokens = &maxTok
193 }
194
195 if req.System != "" {
196 out.Messages = append(out.Messages, oaiMessage{Role: "system", Content: req.System})
197 }
198
199 // DeepSeek's chat-completions API rejects the multimodal content
200 // schema (parts arrays containing image_url). Force every user/tool
201 // message to a plain string and silently drop image blocks for
202 // this provider so historical sessions with screenshots still replay.
203 textOnly := c.name == "deepseek"
204
205 req.Messages = RepairOrphanedToolResults(req.Messages)
206 for _, msg := range req.Messages {
207 switch msg.Role {
208 case RoleUser:
209 content := buildOAIUserContent(msg.Content, textOnly)
210 out.Messages = append(out.Messages, oaiMessage{Role: "user", Content: content})
211 case RoleAssistant:
212 am := oaiMessage{Role: "assistant"}
213 var text strings.Builder
214 var reasoning strings.Builder
215 for _, b := range msg.Content {
216 switch v := b.(type) {
217 case TextBlock:
218 if strings.TrimSpace(v.Text) == "" {
219 continue
220 }
221 if text.Len() > 0 {
222 text.WriteString("\n")
223 }
224 text.WriteString(v.Text)
225 case ToolCallBlock:
226 args := v.Arguments
227 if len(args) == 0 || !json.Valid(args) {
228 args = json.RawMessage("{}")
229 }
230 am.ToolCalls = append(am.ToolCalls, oaiToolCall{
231 ID: v.ID,
232 Type: "function",
233 Function: oaiToolCallFn{
234 Name: v.Name,
235 Arguments: string(args),
236 },
237 })
238 case ReasoningBlock:
239 if v.Summary != "" {
240 if reasoning.Len() > 0 {
241 reasoning.WriteString("\n")
242 }
243 reasoning.WriteString(v.Summary)
244 }
245 }
246 }
247 if text.Len() > 0 {
248 am.Content = text.String()
249 }
250 if reasoning.Len() > 0 && len(am.ToolCalls) > 0 {
251 am.ReasoningContent = reasoning.String()
252 }
253 // Kimi rejects assistant messages with neither visible text nor
254 // tool calls ("assistant must not be empty"). This can happen when
255 // a previous stream produced only reasoning_content, which zot keeps
256 // internally for provider replay but cannot send back as standalone
257 // assistant content on OpenAI-compatible chat-completions APIs.
258 if am.Content == nil && len(am.ToolCalls) == 0 {
259 continue
260 }
261 out.Messages = append(out.Messages, am)
262 case RoleTool:
263 // Each ToolResultBlock becomes its own tool message. Preserve
264 // image blocks for vision-capable OpenAI models instead of
265 // flattening the tool output to plain text.
266 for _, b := range msg.Content {
267 if tr, ok := b.(ToolResultBlock); ok {
268 content := buildOAIToolContent(tr.Content, tr.IsError, textOnly)
269 out.Messages = append(out.Messages, oaiMessage{
270 Role: "tool",
271 ToolCallID: tr.CallID,
272 Content: content,
273 })
274 }
275 }
276 }
277 }
278
279 for _, t := range req.Tools {
280 var tool oaiTool
281 tool.Type = "function"
282 tool.Function.Name = t.Name
283 tool.Function.Description = t.Description
284 tool.Function.Parameters = t.Schema
285 out.Tools = append(out.Tools, tool)
286 }
287 if len(out.Tools) > 0 {
288 out.ToolChoice = "auto"
289 }
290
291 return out, nil
292}
293
294func buildOAIUserContent(blocks []Content, textOnly bool) interface{} {
295 hasImage := false
296 for _, b := range blocks {
297 if _, ok := b.(ImageBlock); ok {
298 hasImage = true
299 break
300 }
301 }
302 if textOnly || !hasImage {
303 var sb strings.Builder
304 for _, b := range blocks {
305 if tb, ok := b.(TextBlock); ok {
306 if sb.Len() > 0 {
307 sb.WriteString("\n")
308 }
309 sb.WriteString(tb.Text)
310 }
311 }
312 return sb.String()
313 }
314 return buildOAIContentBlocks(blocks, false)
315}
316
317func buildOAIToolContent(blocks []Content, isError, textOnly bool) interface{} {
318 hasImage := false
319 for _, b := range blocks {
320 if _, ok := b.(ImageBlock); ok {
321 hasImage = true
322 break
323 }
324 }
325 if textOnly || !hasImage {
326 var sb strings.Builder
327 for _, b := range blocks {
328 if tb, ok := b.(TextBlock); ok {
329 if sb.Len() > 0 {
330 sb.WriteString("\n")
331 }
332 sb.WriteString(tb.Text)
333 }
334 }
335 if isError && sb.Len() > 0 {
336 sb.WriteString(" [error]")
337 }
338 return sb.String()
339 }
340 return buildOAIContentBlocks(blocks, isError)
341}
342
343func buildOAIContentBlocks(blocks []Content, isError bool) []interface{} {
344 var arr []interface{}
345 for _, b := range blocks {
346 switch v := b.(type) {
347 case TextBlock:
348 arr = append(arr, oaiContentText{Type: "text", Text: v.Text})
349 case ImageBlock:
350 var img oaiContentImage
351 img.Type = "image_url"
352 img.ImageURL.URL = "data:" + v.MimeType + ";base64," + base64.StdEncoding.EncodeToString(v.Data)
353 arr = append(arr, img)
354 }
355 }
356 if isError {
357 arr = append(arr, oaiContentText{Type: "text", Text: "[error]"})
358 }
359 return arr
360}
361
362// ---- streaming ----
363
364func (c *openaiClient) Stream(ctx context.Context, req Request) (<-chan Event, error) {
365 apiPath := "/v1/chat/completions"
366 if strings.HasSuffix(c.baseURL, "/v1") {
367 apiPath = "/chat/completions"
368 }
369 wire, err := c.buildRequest(req)
370 if err != nil {
371 return nil, err
372 }
373 body, err := json.Marshal(wire)
374 if err != nil {
375 return nil, err
376 }
377 newReq := func() (*http.Request, error) {
378 httpReq, err := http.NewRequestWithContext(ctx, "POST", c.baseURL+apiPath, bytes.NewReader(body))
379 if err != nil {
380 return nil, err
381 }
382 httpReq.Header.Set("content-type", "application/json")
383 httpReq.Header.Set("accept", "text/event-stream")
384 httpReq.Header.Set("authorization", "Bearer "+c.apiKey)
385 for k, v := range c.headers {
386 httpReq.Header.Set(k, v)
387 }
388 return httpReq, nil
389 }
390
391 resp, err := doStreamWithRetry(ctx, c.http, newReq)
392 if err != nil {
393 return nil, fmt.Errorf("%s: %w", c.Name(), err)
394 }
395 if resp.StatusCode != http.StatusOK {
396 b, _ := io.ReadAll(resp.Body)
397 resp.Body.Close()
398 return nil, fmt.Errorf("%s: http %d: %s", c.Name(), resp.StatusCode, strings.TrimSpace(string(b)))
399 }
400
401 out := make(chan Event, 16)
402 go c.runStream(ctx, resp, req, out)
403 return out, nil
404}
405
406func (c *openaiClient) runStream(ctx context.Context, resp *http.Response, req Request, out chan<- Event) {
407 defer close(out)
408 defer resp.Body.Close()
409
410 model, _ := FindModel("", req.Model)
411 out <- EventStart{Model: req.Model, Provider: c.Name()}
412
413 raw := make(chan sseEvent, 16)
414 go readSSE(resp.Body, raw)
415
416 // Interleaved block tracking: text and tool_calls preserve their
417 // emission order so the assistant message renders in the same order
418 // the model produced it. The builder fires one kind of block at a
419 // time — incoming text deltas after a tool_call split into a fresh
420 // text block; subsequent tool_calls each get their own slot.
421 type blockEntry struct {
422 kind string // "text" | "tool_use"
423 textBuf strings.Builder
424 toolID string
425 toolName string
426 toolArgs strings.Builder
427 announced bool
428 }
429 var (
430 blocks []*blockEntry
431 currentText *blockEntry // most-recent text block, nil if none
432 toolByIdx = map[int]*blockEntry{} // openai tool_call index -> block
433 reasoningBuf strings.Builder
434 usage Usage
435 stop StopReason = StopEnd
436 finalErr error
437 )
438
439 appendText := func(delta string) {
440 if currentText == nil {
441 currentText = &blockEntry{kind: "text"}
442 blocks = append(blocks, currentText)
443 }
444 currentText.textBuf.WriteString(delta)
445 }
446
447 getOrCreateTool := func(idx int) *blockEntry {
448 if t, ok := toolByIdx[idx]; ok {
449 return t
450 }
451 t := &blockEntry{kind: "tool_use"}
452 toolByIdx[idx] = t
453 blocks = append(blocks, t)
454 // A new tool block breaks the current text block. Subsequent text
455 // deltas will start a fresh text block after this tool.
456 currentText = nil
457 return t
458 }
459
460 assembleMsg := func() Message {
461 content := []Content{}
462 for _, b := range blocks {
463 switch b.kind {
464 case "text":
465 if b.textBuf.Len() > 0 {
466 content = append(content, TextBlock{Text: b.textBuf.String()})
467 }
468 case "tool_use":
469 args := b.toolArgs.String()
470 if args == "" || !json.Valid([]byte(args)) {
471 args = "{}"
472 }
473 content = append(content, ToolCallBlock{
474 ID: b.toolID, Name: b.toolName, Arguments: json.RawMessage(args),
475 })
476 }
477 }
478 if reasoningBuf.Len() > 0 {
479 content = append(content, ReasoningBlock{Summary: reasoningBuf.String()})
480 }
481 return Message{Role: RoleAssistant, Content: content, Time: time.Now()}
482 }
483
484 sendDone := func() {
485 usage.CostUSD = ComputeCost(model, usage)
486 out <- EventUsage{Usage: usage}
487 out <- EventDone{Stop: stop, Err: finalErr, Message: assembleMsg()}
488 }
489
490 for {
491 select {
492 case <-ctx.Done():
493 stop = StopAborted
494 finalErr = ctx.Err()
495 sendDone()
496 return
497 case ev, ok := <-raw:
498 if !ok {
499 sendDone()
500 return
501 }
502 if ev.Data == "[DONE]" {
503 sendDone()
504 return
505 }
506 var chunk struct {
507 Choices []struct {
508 Index int `json:"index"`
509 Delta struct {
510 Content string `json:"content"`
511 ReasoningContent string `json:"reasoning_content"`
512 ToolCalls []struct {
513 Index int `json:"index"`
514 ID string `json:"id"`
515 Type string `json:"type"`
516 Function struct {
517 Name string `json:"name"`
518 Arguments string `json:"arguments"`
519 } `json:"function"`
520 } `json:"tool_calls"`
521 } `json:"delta"`
522 FinishReason string `json:"finish_reason"`
523 } `json:"choices"`
524 Usage *struct {
525 PromptTokens int `json:"prompt_tokens"`
526 CompletionTokens int `json:"completion_tokens"`
527 PromptTokensDetails struct {
528 CachedTokens int `json:"cached_tokens"`
529 } `json:"prompt_tokens_details"`
530 } `json:"usage"`
531 Error *struct {
532 Message string `json:"message"`
533 Type string `json:"type"`
534 } `json:"error"`
535 }
536 if err := json.Unmarshal([]byte(ev.Data), &chunk); err != nil {
537 continue
538 }
539 if chunk.Error != nil {
540 stop = StopError
541 finalErr = fmt.Errorf("openai: %s", chunk.Error.Message)
542 sendDone()
543 return
544 }
545 if chunk.Usage != nil {
546 usage.InputTokens = chunk.Usage.PromptTokens - chunk.Usage.PromptTokensDetails.CachedTokens
547 if usage.InputTokens < 0 {
548 usage.InputTokens = chunk.Usage.PromptTokens
549 }
550 usage.OutputTokens = chunk.Usage.CompletionTokens
551 usage.CacheReadTokens = chunk.Usage.PromptTokensDetails.CachedTokens
552 }
553 for _, ch := range chunk.Choices {
554 if ch.Delta.ReasoningContent != "" {
555 reasoningBuf.WriteString(ch.Delta.ReasoningContent)
556 }
557 if ch.Delta.Content != "" {
558 appendText(ch.Delta.Content)
559 out <- EventTextDelta{Delta: ch.Delta.Content}
560 }
561 for _, tc := range ch.Delta.ToolCalls {
562 t := getOrCreateTool(tc.Index)
563 if tc.ID != "" {
564 t.toolID = tc.ID
565 }
566 if tc.Function.Name != "" {
567 t.toolName = tc.Function.Name
568 }
569 if !t.announced && t.toolID != "" && t.toolName != "" {
570 t.announced = true
571 out <- EventToolStart{ID: t.toolID, Name: t.toolName}
572 }
573 if tc.Function.Arguments != "" {
574 t.toolArgs.WriteString(tc.Function.Arguments)
575 if t.announced {
576 out <- EventToolArgs{ID: t.toolID, Delta: tc.Function.Arguments}
577 }
578 }
579 }
580 switch ch.FinishReason {
581 case "stop":
582 stop = StopEnd
583 case "length":
584 stop = StopLength
585 case "tool_calls", "function_call":
586 stop = StopToolUse
587 for _, b := range blocks {
588 if b.kind == "tool_use" && b.announced {
589 out <- EventToolEnd{ID: b.toolID}
590 }
591 }
592 }
593 }
594 }
595 }
596}