@plurnk/plurnk-providers 5.0.2 → 6.0.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (108) hide show
  1. package/.env.defaults +18 -84
  2. package/README.md +21 -22
  3. package/SPEC.md +110 -1190
  4. package/dist/LeadingReasoning.js +1 -1
  5. package/dist/LeadingReasoning.js.map +1 -1
  6. package/dist/Mock.d.ts +2 -2
  7. package/dist/Mock.d.ts.map +1 -1
  8. package/dist/Mock.js +2 -2
  9. package/dist/Mock.js.map +1 -1
  10. package/dist/{AiSdkProvider.d.ts → ModelProvider.d.ts} +12 -21
  11. package/dist/ModelProvider.d.ts.map +1 -0
  12. package/dist/{AiSdkProvider.js → ModelProvider.js} +140 -188
  13. package/dist/ModelProvider.js.map +1 -0
  14. package/dist/{AiSdkRequestBody.d.ts → ModelRequestBody.d.ts} +6 -10
  15. package/dist/ModelRequestBody.d.ts.map +1 -0
  16. package/dist/{AiSdkRequestBody.js → ModelRequestBody.js} +17 -63
  17. package/dist/ModelRequestBody.js.map +1 -0
  18. package/dist/Pool.js +24 -24
  19. package/dist/Pool.js.map +1 -1
  20. package/dist/ProviderRegistry.d.ts.map +1 -1
  21. package/dist/ProviderRegistry.js +15 -15
  22. package/dist/ProviderRegistry.js.map +1 -1
  23. package/dist/RequestCapture.d.ts +1 -0
  24. package/dist/RequestCapture.d.ts.map +1 -1
  25. package/dist/RequestCapture.js +9 -6
  26. package/dist/RequestCapture.js.map +1 -1
  27. package/dist/accounting.d.ts +1 -1
  28. package/dist/accounting.d.ts.map +1 -1
  29. package/dist/accounting.js +4 -4
  30. package/dist/accounting.js.map +1 -1
  31. package/dist/capacity.js +4 -4
  32. package/dist/capacity.js.map +1 -1
  33. package/dist/catalogProvider.d.ts +5 -8
  34. package/dist/catalogProvider.d.ts.map +1 -1
  35. package/dist/catalogProvider.js +218 -68
  36. package/dist/catalogProvider.js.map +1 -1
  37. package/dist/compatibleProvider.d.ts.map +1 -1
  38. package/dist/compatibleProvider.js +19 -8
  39. package/dist/compatibleProvider.js.map +1 -1
  40. package/dist/discover.d.ts +1 -1
  41. package/dist/discover.d.ts.map +1 -1
  42. package/dist/discover.js +1 -8
  43. package/dist/discover.js.map +1 -1
  44. package/dist/env.d.ts +4 -4
  45. package/dist/env.d.ts.map +1 -1
  46. package/dist/env.js +43 -80
  47. package/dist/env.js.map +1 -1
  48. package/dist/errors.d.ts +48 -0
  49. package/dist/errors.d.ts.map +1 -1
  50. package/dist/errors.js +61 -4
  51. package/dist/errors.js.map +1 -1
  52. package/dist/index.d.ts +5 -6
  53. package/dist/index.d.ts.map +1 -1
  54. package/dist/index.js +6 -7
  55. package/dist/index.js.map +1 -1
  56. package/dist/{sdkModels.d.ts → modelAdapters.d.ts} +7 -9
  57. package/dist/modelAdapters.d.ts.map +1 -0
  58. package/dist/{sdkModels.js → modelAdapters.js} +194 -87
  59. package/dist/modelAdapters.js.map +1 -0
  60. package/dist/native-tools.d.ts +1 -2
  61. package/dist/native-tools.d.ts.map +1 -1
  62. package/dist/native-tools.js +9 -11
  63. package/dist/native-tools.js.map +1 -1
  64. package/dist/ollama.d.ts.map +1 -1
  65. package/dist/ollama.js +26 -12
  66. package/dist/ollama.js.map +1 -1
  67. package/dist/openai.d.ts +3 -3
  68. package/dist/openai.d.ts.map +1 -1
  69. package/dist/openai.js +1 -1
  70. package/dist/pi-transport.d.ts +69 -0
  71. package/dist/pi-transport.d.ts.map +1 -0
  72. package/dist/pi-transport.js +697 -0
  73. package/dist/pi-transport.js.map +1 -0
  74. package/dist/providerError.js +1 -1
  75. package/dist/providerError.js.map +1 -1
  76. package/dist/retry-deadline.d.ts +3 -0
  77. package/dist/retry-deadline.d.ts.map +1 -0
  78. package/dist/retry-deadline.js +96 -0
  79. package/dist/retry-deadline.js.map +1 -0
  80. package/dist/stream-forensics.d.ts +51 -0
  81. package/dist/stream-forensics.d.ts.map +1 -0
  82. package/dist/stream-forensics.js +466 -0
  83. package/dist/stream-forensics.js.map +1 -0
  84. package/dist/types.d.ts +29 -4
  85. package/dist/types.d.ts.map +1 -1
  86. package/dist/types.js.map +1 -1
  87. package/dist/usage.js +2 -2
  88. package/dist/usage.js.map +1 -1
  89. package/docs/models.md +3 -3
  90. package/package.json +8 -21
  91. package/dist/AiSdkProvider.d.ts.map +0 -1
  92. package/dist/AiSdkProvider.js.map +0 -1
  93. package/dist/AiSdkRequestBody.d.ts.map +0 -1
  94. package/dist/AiSdkRequestBody.js.map +0 -1
  95. package/dist/RequestFields.d.ts +0 -24
  96. package/dist/RequestFields.d.ts.map +0 -1
  97. package/dist/RequestFields.js +0 -237
  98. package/dist/RequestFields.js.map +0 -1
  99. package/dist/aiSdkTransport.d.ts +0 -84
  100. package/dist/aiSdkTransport.d.ts.map +0 -1
  101. package/dist/aiSdkTransport.js +0 -651
  102. package/dist/aiSdkTransport.js.map +0 -1
  103. package/dist/model-options.d.ts +0 -3
  104. package/dist/model-options.d.ts.map +0 -1
  105. package/dist/model-options.js +0 -26
  106. package/dist/model-options.js.map +0 -1
  107. package/dist/sdkModels.d.ts.map +0 -1
  108. package/dist/sdkModels.js.map +0 -1
@@ -0,0 +1,697 @@
1
+ // {§provider-sdk-boundary} {§provider-request-authority} Vendor transport adapter over @earendil-works/pi-ai
2
+ // and OpenAI-compatible HTTP endpoints (#1305, #1313).
3
+ import { normalizeContext, } from "@earendil-works/pi-ai";
4
+ import { stream as streamAnthropic } from "@earendil-works/pi-ai/api/anthropic-messages";
5
+ import { stream as streamBedrockConverseStream } from "@earendil-works/pi-ai/api/bedrock-converse-stream";
6
+ import { stream as streamGoogleGenerativeAi } from "@earendil-works/pi-ai/api/google-generative-ai";
7
+ import { stream as streamOpenAICompletions } from "@earendil-works/pi-ai/api/openai-completions";
8
+ import { stream as streamOpenAIResponses } from "@earendil-works/pi-ai/api/openai-responses";
9
+ import { isRecord } from "@plurnk/plurnk-contracts";
10
+ import { outboundFetch, syncAmbientHttp } from "@plurnk/plurnk-meta";
11
+ import { APICallError } from "./errors.js";
12
+ import { withoutNativeTools } from "./native-tools.js";
13
+ import { nativeFixedEffort } from "./reasoning-effort.js";
14
+ import { abortFailure, attachFailureEvidence, collectChunkForensics, extractFromAssistantMessage, serializeError, toUsageFromPiMessage, } from "./stream-forensics.js";
15
+ const headersToRecord = (headers) => {
16
+ const out = {};
17
+ for (const [key, value] of headers.entries())
18
+ out[key] = value;
19
+ return out;
20
+ };
21
+ const toPiContext = (messages) => {
22
+ const systemParts = [];
23
+ const contextMessages = [];
24
+ let leadingSystem = true;
25
+ for (const message of messages) {
26
+ if (leadingSystem && message.role === "system") {
27
+ systemParts.push(typeof message.content === "string"
28
+ ? message.content
29
+ : message.content.map((part) => part.type === "text" ? part.text : "").join(""));
30
+ continue;
31
+ }
32
+ leadingSystem = false;
33
+ if (message.role === "assistant") {
34
+ const text = typeof message.content === "string"
35
+ ? message.content
36
+ : message.content.map((part) => part.type === "text" ? part.text : "").join("");
37
+ contextMessages.push({
38
+ role: "assistant",
39
+ content: [{ type: "text", text }],
40
+ api: "openai-completions",
41
+ provider: "openai",
42
+ model: "unknown",
43
+ usage: {
44
+ input: 0,
45
+ output: 0,
46
+ cacheRead: 0,
47
+ cacheWrite: 0,
48
+ totalTokens: 0,
49
+ cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0, total: 0 },
50
+ },
51
+ stopReason: "stop",
52
+ timestamp: 0,
53
+ });
54
+ continue;
55
+ }
56
+ const content = typeof message.content === "string"
57
+ ? message.content
58
+ : message.content.map((part) => part.type === "text"
59
+ ? { type: "text", text: part.text }
60
+ : { type: "image", data: Buffer.from(part.data).toString("base64"), mimeType: part.mediaType });
61
+ contextMessages.push({
62
+ role: "user",
63
+ content,
64
+ timestamp: 0,
65
+ });
66
+ }
67
+ return normalizeContext({
68
+ ...(systemParts.length > 0 ? { systemPrompt: systemParts.join("\n\n") } : {}),
69
+ messages: contextMessages,
70
+ });
71
+ };
72
+ const mapOpenAICompatibleMessages = (messages, systemCacheBreakpoint) => {
73
+ let lastLeadingSystemIndex = -1;
74
+ if (systemCacheBreakpoint) {
75
+ for (let i = 0; i < messages.length; i += 1) {
76
+ if (messages[i].role !== "system")
77
+ break;
78
+ lastLeadingSystemIndex = i;
79
+ }
80
+ }
81
+ return messages.map((message, index) => {
82
+ if (index === lastLeadingSystemIndex && typeof message.content === "string") {
83
+ return {
84
+ role: message.role,
85
+ content: [{
86
+ type: "text",
87
+ text: message.content,
88
+ cache_control: { type: "ephemeral" },
89
+ }],
90
+ };
91
+ }
92
+ if (typeof message.content === "string") {
93
+ return { role: message.role, content: message.content };
94
+ }
95
+ return {
96
+ role: message.role,
97
+ content: message.content.map((part) => {
98
+ if (part.type === "text")
99
+ return { type: "text", text: part.text };
100
+ const base64 = Buffer.from(part.data).toString("base64");
101
+ if (part.mediaType.startsWith("audio/")) {
102
+ return {
103
+ type: "input_audio",
104
+ input_audio: {
105
+ data: base64,
106
+ format: part.mediaType.slice("audio/".length),
107
+ },
108
+ };
109
+ }
110
+ return {
111
+ type: "image_url",
112
+ image_url: { url: `data:${part.mediaType};base64,${base64}` },
113
+ };
114
+ }),
115
+ };
116
+ });
117
+ };
118
+ const applyAnthropicPayloadAdjustments = (params, messages, systemCacheBreakpoint, serviceTier) => {
119
+ if (serviceTier !== undefined) {
120
+ params.service_tier = serviceTier;
121
+ }
122
+ if (systemCacheBreakpoint && Array.isArray(params.system) && params.system.length > 0) {
123
+ const last = params.system.at(-1);
124
+ if (isRecord(last)) {
125
+ last.cache_control = { type: "ephemeral" };
126
+ }
127
+ }
128
+ const wireMessages = Array.isArray(params.messages) ? params.messages : undefined;
129
+ if (wireMessages === undefined)
130
+ return;
131
+ const nonSystem = messages.filter((m) => m.role !== "system");
132
+ for (let i = 0; i < Math.min(wireMessages.length, nonSystem.length); i += 1) {
133
+ const sourceMsg = nonSystem[i];
134
+ const wireMsg = wireMessages[i];
135
+ if (!Array.isArray(sourceMsg.content) || !isRecord(wireMsg) || !Array.isArray(wireMsg.content))
136
+ continue;
137
+ for (let j = 0; j < Math.min(sourceMsg.content.length, wireMsg.content.length); j += 1) {
138
+ const sourcePart = sourceMsg.content[j];
139
+ const wireBlock = wireMsg.content[j];
140
+ if (sourcePart.type === "file"
141
+ && sourcePart.mediaType === "application/pdf"
142
+ && isRecord(wireBlock)
143
+ && wireBlock.type === "image") {
144
+ wireBlock.type = "document";
145
+ }
146
+ }
147
+ }
148
+ };
149
+ export const executePiTransportStream = async (opts) => {
150
+ const { model, source, url, providerModel, fetch: configuredFetch, requestHeaders, messages, sessionId, operationSignal, requestSampling, body, effectiveMaxOutputTokens, nativeReasoningBudget, capacity, requestBody, effort, adaptiveEffort, nativeAdaptiveThinking, reasoningRequestBody, systemCacheBreakpoint, cacheAffinity, resolvedControls, rawBody, capture, reasoningStepDown = 0, observers, } = opts;
151
+ const effectiveModel = providerModel ?? {
152
+ id: model,
153
+ api: "openai-completions",
154
+ provider: source,
155
+ baseUrl: url.replace(/\/chat\/completions\/?$/, ""),
156
+ compat: {
157
+ supportsStore: false,
158
+ supportsDeveloperRole: false,
159
+ supportsUsageInStreaming: true,
160
+ supportsFinishReason: true,
161
+ },
162
+ };
163
+ const buildPiModel = (api) => ({
164
+ id: effectiveModel.id,
165
+ name: effectiveModel.name ?? effectiveModel.id,
166
+ api,
167
+ provider: effectiveModel.provider,
168
+ baseUrl: effectiveModel.baseUrl,
169
+ reasoning: effectiveModel.reasoning ?? true,
170
+ input: [...(effectiveModel.input ?? ["text", "image"])],
171
+ cost: effectiveModel.cost ?? { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
172
+ contextWindow: effectiveModel.contextWindow ?? 128_000,
173
+ maxTokens: effectiveModel.maxTokens ?? 32_768,
174
+ ...(effectiveModel.headers === undefined ? {} : { headers: { ...effectiveModel.headers } }),
175
+ });
176
+ const context = toPiContext(messages);
177
+ const rawChunks = [];
178
+ let httpResponse;
179
+ let httpErrorBody;
180
+ let capturedTransportError;
181
+ let requestUrl = url ?? effectiveModel.baseUrl;
182
+ let requestBodyValues;
183
+ const baseFetch = capture?.fetch(configuredFetch ?? outboundFetch) ?? configuredFetch ?? outboundFetch;
184
+ const customFetch = async (input, init) => {
185
+ const effectiveInput = providerModel === undefined && url !== undefined
186
+ ? url
187
+ : input;
188
+ requestUrl = typeof effectiveInput === "string"
189
+ ? effectiveInput
190
+ : effectiveInput instanceof URL
191
+ ? effectiveInput.toString()
192
+ : effectiveInput.url;
193
+ if (typeof init?.body === "string") {
194
+ try {
195
+ requestBodyValues = JSON.parse(init.body);
196
+ }
197
+ catch {
198
+ requestBodyValues = init.body;
199
+ }
200
+ }
201
+ try {
202
+ const res = await baseFetch(effectiveInput, init);
203
+ httpResponse = {
204
+ status: res.status,
205
+ headers: headersToRecord(res.headers),
206
+ ok: res.ok,
207
+ };
208
+ if (!res.ok) {
209
+ try {
210
+ httpErrorBody = await res.clone().text();
211
+ }
212
+ catch {
213
+ // ignore clone read error
214
+ }
215
+ }
216
+ return res;
217
+ }
218
+ catch (err) {
219
+ capturedTransportError = err;
220
+ throw err;
221
+ }
222
+ };
223
+ const temperature = typeof requestSampling.temperature === "number" ? requestSampling.temperature : undefined;
224
+ const topP = typeof requestSampling.top_p === "number" ? requestSampling.top_p : undefined;
225
+ const topK = typeof requestSampling.top_k === "number" ? requestSampling.top_k : undefined;
226
+ const presencePenalty = typeof requestSampling.presence_penalty === "number" ? requestSampling.presence_penalty : undefined;
227
+ const frequencyPenalty = typeof requestSampling.frequency_penalty === "number" ? requestSampling.frequency_penalty : undefined;
228
+ const stopSequences = typeof requestSampling.stop === "string"
229
+ ? [requestSampling.stop]
230
+ : Array.isArray(requestSampling.stop) && requestSampling.stop.every((value) => typeof value === "string")
231
+ ? requestSampling.stop
232
+ : undefined;
233
+ const seed = typeof requestSampling.seed === "number" ? requestSampling.seed : undefined;
234
+ const maxTokens = providerModel === undefined
235
+ ? effectiveMaxOutputTokens
236
+ : requestBody.nativeMaxOutputTokens(capacity.outputBudget, nativeReasoningBudget);
237
+ const reasoning = providerModel === undefined || reasoningRequestBody !== undefined
238
+ ? "provider-default"
239
+ : effort.mode === "off"
240
+ ? "none"
241
+ : effort.mode === "adaptive"
242
+ ? adaptiveEffort
243
+ : nativeFixedEffort(effort.mode);
244
+ const commonOptions = {
245
+ signal: operationSignal,
246
+ headers: requestHeaders,
247
+ ...(temperature === undefined ? {} : { temperature }),
248
+ ...(maxTokens === undefined ? {} : { maxTokens }),
249
+ onProviderStreamEvent: (rawEvent) => {
250
+ rawChunks.push(rawEvent);
251
+ },
252
+ };
253
+ const onPayload = async (payload) => {
254
+ const transformed = effectiveModel.onPayload !== undefined
255
+ ? await effectiveModel.onPayload(payload, effectiveModel)
256
+ : payload;
257
+ const params = isRecord(transformed) ? { ...transformed } : {};
258
+ if (effectiveModel.api === "openai-completions") {
259
+ if (topP !== undefined)
260
+ params.top_p = topP;
261
+ if (topK !== undefined && effectiveModel.includeTopK === true)
262
+ params.top_k = topK;
263
+ if (presencePenalty !== undefined)
264
+ params.presence_penalty = presencePenalty;
265
+ if (frequencyPenalty !== undefined)
266
+ params.frequency_penalty = frequencyPenalty;
267
+ if (stopSequences !== undefined)
268
+ params.stop = stopSequences;
269
+ if (seed !== undefined)
270
+ params.seed = seed;
271
+ if (systemCacheBreakpoint
272
+ || messages.some((m) => Array.isArray(m.content) && m.content.some((p) => p.type === "file" && !p.mediaType.startsWith("image/")))) {
273
+ params.messages = mapOpenAICompatibleMessages(messages, systemCacheBreakpoint);
274
+ }
275
+ else if (messages[0]?.role === "system" && Array.isArray(params.messages)) {
276
+ const first = params.messages[0];
277
+ if (!isRecord(first) || (first.role !== "system" && first.role !== "developer")) {
278
+ params.messages.unshift({ role: "system", content: "" });
279
+ }
280
+ }
281
+ if (providerModel === undefined) {
282
+ for (const [key, value] of Object.entries(body)) {
283
+ if (value === undefined)
284
+ delete params[key];
285
+ else
286
+ params[key] = value;
287
+ }
288
+ }
289
+ else {
290
+ if (reasoningRequestBody !== undefined) {
291
+ Object.assign(params, reasoningRequestBody(effort.mode, capacity.reasoningBudget));
292
+ }
293
+ if (cacheAffinity?.target === "body") {
294
+ params[cacheAffinity.name] = sessionId;
295
+ }
296
+ if (resolvedControls.topLogprobs !== undefined) {
297
+ params.logprobs = true;
298
+ params.top_logprobs = resolvedControls.topLogprobs;
299
+ }
300
+ if (resolvedControls.serviceTier !== undefined) {
301
+ params.service_tier = resolvedControls.serviceTier;
302
+ }
303
+ }
304
+ }
305
+ else if (effectiveModel.api === "openai-responses") {
306
+ if (topP !== undefined)
307
+ params.top_p = topP;
308
+ if (cacheAffinity?.target === "body") {
309
+ params[cacheAffinity.name] = sessionId;
310
+ }
311
+ if (resolvedControls.serviceTier !== undefined) {
312
+ params.service_tier = resolvedControls.serviceTier;
313
+ }
314
+ if (resolvedControls.topLogprobs !== undefined) {
315
+ const existingInclude = Array.isArray(params.include)
316
+ ? params.include.filter((item) => typeof item === "string")
317
+ : [];
318
+ if (!existingInclude.includes("message.output_text.logprobs")) {
319
+ params.include = [...existingInclude, "message.output_text.logprobs"];
320
+ }
321
+ params.top_logprobs = resolvedControls.topLogprobs;
322
+ }
323
+ }
324
+ else if (effectiveModel.api === "anthropic-messages") {
325
+ if (topP !== undefined)
326
+ params.top_p = topP;
327
+ if (topK !== undefined)
328
+ params.top_k = topK;
329
+ if (stopSequences !== undefined)
330
+ params.stop_sequences = stopSequences;
331
+ applyAnthropicPayloadAdjustments(params, messages, systemCacheBreakpoint, resolvedControls.serviceTier);
332
+ }
333
+ else if (effectiveModel.api === "google-generative-ai") {
334
+ const config = isRecord(params.config) ? { ...params.config } : {};
335
+ if (topP !== undefined)
336
+ config.topP = topP;
337
+ if (topK !== undefined)
338
+ config.topK = topK;
339
+ if (presencePenalty !== undefined)
340
+ config.presencePenalty = presencePenalty;
341
+ if (frequencyPenalty !== undefined)
342
+ config.frequencyPenalty = frequencyPenalty;
343
+ if (stopSequences !== undefined)
344
+ config.stopSequences = stopSequences;
345
+ if (seed !== undefined)
346
+ config.seed = seed;
347
+ if (resolvedControls.serviceTier !== undefined) {
348
+ config.serviceTier = resolvedControls.serviceTier;
349
+ }
350
+ params.config = withoutNativeTools(config, "generate-content");
351
+ try {
352
+ capture?.record({
353
+ method: "POST",
354
+ origin: new URL(effectiveModel.baseUrl).origin,
355
+ body: JSON.stringify(params),
356
+ });
357
+ }
358
+ catch {
359
+ // ignore invalid URL
360
+ }
361
+ requestBodyValues = params;
362
+ return params;
363
+ }
364
+ else if (effectiveModel.api === "bedrock-converse-stream") {
365
+ const inferenceConfig = isRecord(params.inferenceConfig)
366
+ ? { ...params.inferenceConfig }
367
+ : {};
368
+ if (topP !== undefined)
369
+ inferenceConfig.topP = topP;
370
+ if (stopSequences !== undefined)
371
+ inferenceConfig.stopSequences = stopSequences;
372
+ if (Object.keys(inferenceConfig).length > 0)
373
+ params.inferenceConfig = inferenceConfig;
374
+ if (resolvedControls.serviceTier !== undefined) {
375
+ params.serviceTier = { type: resolvedControls.serviceTier };
376
+ }
377
+ }
378
+ const toolProtocol = effectiveModel.api === "bedrock-converse-stream"
379
+ ? "amazon-bedrock"
380
+ : effectiveModel.api === "anthropic-messages"
381
+ ? "messages"
382
+ : "openai";
383
+ const nextParams = withoutNativeTools(params, toolProtocol);
384
+ if (effectiveModel.api === "bedrock-converse-stream") {
385
+ try {
386
+ capture?.record({
387
+ method: "POST",
388
+ origin: new URL(effectiveModel.baseUrl).origin,
389
+ body: JSON.stringify(nextParams),
390
+ });
391
+ }
392
+ catch {
393
+ // ignore invalid URL
394
+ }
395
+ }
396
+ requestBodyValues = nextParams;
397
+ return nextParams;
398
+ };
399
+ const startStream = () => {
400
+ switch (effectiveModel.api) {
401
+ case "openai-completions": {
402
+ const reasoningEffort = providerModel === undefined || reasoning === "none" || reasoning === "provider-default"
403
+ ? undefined
404
+ : reasoning;
405
+ return streamOpenAICompletions({
406
+ ...buildPiModel("openai-completions"),
407
+ ...(effectiveModel.compat === undefined ? {} : { compat: { ...effectiveModel.compat } }),
408
+ }, context, {
409
+ ...commonOptions,
410
+ fetch: customFetch,
411
+ apiKey: effectiveModel.apiKey ?? "plurnk-unset",
412
+ ...(reasoningEffort === undefined ? {} : { reasoningEffort }),
413
+ onPayload,
414
+ });
415
+ }
416
+ case "openai-responses": {
417
+ const reasoningEffort = reasoning === "none" || reasoning === "provider-default" ? undefined : reasoning;
418
+ return streamOpenAIResponses(buildPiModel("openai-responses"), context, {
419
+ ...commonOptions,
420
+ fetch: customFetch,
421
+ apiKey: effectiveModel.apiKey ?? "plurnk-unset",
422
+ ...(reasoningEffort === undefined ? {} : { reasoningEffort }),
423
+ onPayload,
424
+ });
425
+ }
426
+ case "anthropic-messages": {
427
+ const effortOption = reasoning !== "none" && reasoning !== "provider-default"
428
+ ? (reasoning === "minimal" ? "low" : reasoning === "xhigh" ? "max" : reasoning)
429
+ : undefined;
430
+ return streamAnthropic({
431
+ ...buildPiModel("anthropic-messages"),
432
+ ...(nativeAdaptiveThinking ? { compat: { forceAdaptiveThinking: true } } : {}),
433
+ }, context, {
434
+ ...commonOptions,
435
+ fetch: customFetch,
436
+ apiKey: effectiveModel.apiKey ?? "plurnk-unset",
437
+ cacheRetention: "none",
438
+ ...(reasoning === "none"
439
+ ? { thinkingEnabled: false }
440
+ : nativeAdaptiveThinking
441
+ ? {
442
+ thinkingEnabled: true,
443
+ ...(effortOption === undefined ? {} : { effort: effortOption }),
444
+ }
445
+ : nativeReasoningBudget !== null
446
+ ? { thinkingEnabled: true, thinkingBudgetTokens: nativeReasoningBudget }
447
+ : {}),
448
+ onPayload,
449
+ });
450
+ }
451
+ case "google-generative-ai": {
452
+ syncAmbientHttp();
453
+ const googleLevels = ["HIGH", "MEDIUM", "LOW", "MINIMAL"];
454
+ const baseLevel = reasoning === "minimal"
455
+ ? "MINIMAL"
456
+ : reasoning === "low"
457
+ ? "LOW"
458
+ : reasoning === "medium"
459
+ ? "MEDIUM"
460
+ : "HIGH";
461
+ const steppedIndex = Math.min(googleLevels.length - 1, googleLevels.indexOf(baseLevel) + reasoningStepDown);
462
+ const googleLevel = googleLevels[steppedIndex];
463
+ const rawBudget = nativeReasoningBudget ?? capacity.reasoningBudget;
464
+ const reservedVisible = capacity.outputFloor !== null
465
+ && maxTokens !== undefined
466
+ && maxTokens > capacity.outputFloor
467
+ ? capacity.outputFloor
468
+ : 1;
469
+ const maxThinkingBudget = maxTokens !== undefined
470
+ ? Math.max(1, maxTokens - reservedVisible)
471
+ : null;
472
+ const boundedBudget = rawBudget === null
473
+ ? null
474
+ : maxThinkingBudget === null
475
+ ? rawBudget
476
+ : Math.min(rawBudget, maxThinkingBudget);
477
+ const steppedBudget = boundedBudget === null
478
+ ? null
479
+ : Math.max(1, Math.floor(boundedBudget / 2 ** reasoningStepDown));
480
+ const googleThinking = reasoning === "none"
481
+ ? { enabled: false }
482
+ : steppedBudget !== null
483
+ ? { enabled: true, budgetTokens: steppedBudget }
484
+ : reasoning !== "provider-default"
485
+ ? {
486
+ enabled: true,
487
+ level: googleLevel,
488
+ }
489
+ : undefined;
490
+ return streamGoogleGenerativeAi(buildPiModel("google-generative-ai"), context, {
491
+ ...commonOptions,
492
+ ...(effectiveModel.apiKey === undefined ? {} : { apiKey: effectiveModel.apiKey }),
493
+ ...(googleThinking === undefined ? {} : { thinking: googleThinking }),
494
+ onPayload,
495
+ });
496
+ }
497
+ case "bedrock-converse-stream": {
498
+ syncAmbientHttp();
499
+ const bedrockReasoning = reasoning === "none" || reasoning === "provider-default"
500
+ ? undefined
501
+ : (reasoning === "xhigh" ? "max" : reasoning);
502
+ return streamBedrockConverseStream(buildPiModel("bedrock-converse-stream"), context, {
503
+ ...commonOptions,
504
+ ...(bedrockReasoning === undefined ? {} : { reasoning: bedrockReasoning }),
505
+ ...(nativeReasoningBudget !== null && bedrockReasoning !== undefined
506
+ ? {
507
+ thinkingBudgets: {
508
+ minimal: nativeReasoningBudget,
509
+ low: nativeReasoningBudget,
510
+ medium: nativeReasoningBudget,
511
+ high: nativeReasoningBudget,
512
+ },
513
+ }
514
+ : {}),
515
+ onPayload,
516
+ });
517
+ }
518
+ }
519
+ };
520
+ let outputObserved = false;
521
+ let finalMessage;
522
+ let lastErrorMessage;
523
+ let streamError;
524
+ const stream = startStream();
525
+ for await (const event of stream) {
526
+ if (event.type === "text_delta" && event.delta.length > 0) {
527
+ outputObserved = true;
528
+ observers.observeText?.(event.delta);
529
+ }
530
+ else if (event.type === "thinking_delta" && event.delta.length > 0) {
531
+ outputObserved = true;
532
+ observers.observeReasoning?.(event.delta);
533
+ }
534
+ else if (event.type === "toolcall_start" || event.type === "toolcall_delta" || event.type === "toolcall_end") {
535
+ outputObserved = true;
536
+ }
537
+ else if (event.type === "done") {
538
+ finalMessage = event.message;
539
+ }
540
+ else if (event.type === "error") {
541
+ lastErrorMessage = event.error;
542
+ if (event.error.rawStopReason !== undefined
543
+ && !operationSignal?.aborted
544
+ && capturedTransportError === undefined
545
+ && httpResponse?.ok !== false) {
546
+ finalMessage = event.error;
547
+ }
548
+ else {
549
+ streamError = event.error.errorMessage ?? "Provider stream failed";
550
+ }
551
+ }
552
+ }
553
+ const forensics = collectChunkForensics(rawChunks);
554
+ const messageInfo = extractFromAssistantMessage(finalMessage ?? lastErrorMessage);
555
+ const responseHeaders = httpResponse?.headers;
556
+ const responseId = finalMessage?.responseId ?? lastErrorMessage?.responseId ?? forensics.id;
557
+ if (finalMessage === undefined) {
558
+ const failureRawBody = rawChunks.length > 0
559
+ ? rawChunks
560
+ : httpErrorBody !== undefined && httpErrorBody.length > 0
561
+ ? httpErrorBody
562
+ : undefined;
563
+ if (operationSignal?.aborted) {
564
+ const abortError = abortFailure(operationSignal);
565
+ const usage = toUsageFromPiMessage(lastErrorMessage, forensics.rawUsage);
566
+ attachFailureEvidence(abortError, {
567
+ status: httpResponse?.status,
568
+ usage,
569
+ chargeEvidence: {
570
+ ...(forensics.charge === undefined ? {} : { charge: forensics.charge }),
571
+ ...(forensics.rawUsage === undefined ? {} : { usage: forensics.rawUsage }),
572
+ response: {
573
+ ...(responseId === undefined ? {} : { id: responseId }),
574
+ ...(responseHeaders === undefined ? {} : { headers: responseHeaders }),
575
+ },
576
+ },
577
+ content: messageInfo.content,
578
+ reasoning: messageInfo.reasoning,
579
+ error: serializeError(abortError),
580
+ ...(failureRawBody === undefined ? {} : { rawBody: failureRawBody }),
581
+ outputObserved: outputObserved || messageInfo.content.length > 0 || messageInfo.reasoning.length > 0,
582
+ });
583
+ throw abortError;
584
+ }
585
+ const statusCode = httpResponse !== undefined && !httpResponse.ok ? httpResponse.status : undefined;
586
+ if (capturedTransportError !== undefined && statusCode === undefined) {
587
+ const usage = toUsageFromPiMessage(lastErrorMessage, forensics.rawUsage);
588
+ attachFailureEvidence(capturedTransportError, {
589
+ status: httpResponse?.status,
590
+ usage,
591
+ chargeEvidence: {
592
+ ...(forensics.charge === undefined ? {} : { charge: forensics.charge }),
593
+ ...(forensics.rawUsage === undefined ? {} : { usage: forensics.rawUsage }),
594
+ response: {
595
+ ...(responseId === undefined ? {} : { id: responseId }),
596
+ ...(responseHeaders === undefined ? {} : { headers: responseHeaders }),
597
+ },
598
+ },
599
+ content: messageInfo.content,
600
+ reasoning: messageInfo.reasoning,
601
+ error: serializeError(capturedTransportError),
602
+ ...(failureRawBody === undefined ? {} : { rawBody: failureRawBody }),
603
+ outputObserved: outputObserved || messageInfo.content.length > 0 || messageInfo.reasoning.length > 0,
604
+ });
605
+ throw capturedTransportError;
606
+ }
607
+ let parsedErrorData;
608
+ let errorMessage = streamError ?? `HTTP ${statusCode ?? "error"}`;
609
+ if (httpErrorBody !== undefined && httpErrorBody.length > 0) {
610
+ try {
611
+ parsedErrorData = JSON.parse(httpErrorBody);
612
+ if (isRecord(parsedErrorData)) {
613
+ const errField = parsedErrorData.error;
614
+ if (isRecord(errField) && typeof errField.message === "string" && errField.message.length > 0) {
615
+ errorMessage = errField.message;
616
+ }
617
+ else if (typeof errField === "string" && errField.length > 0) {
618
+ errorMessage = errField;
619
+ }
620
+ else if (typeof parsedErrorData.message === "string" && parsedErrorData.message.length > 0) {
621
+ errorMessage = parsedErrorData.message;
622
+ }
623
+ else {
624
+ errorMessage = httpErrorBody;
625
+ }
626
+ }
627
+ }
628
+ catch {
629
+ errorMessage = httpErrorBody;
630
+ }
631
+ }
632
+ const apiError = new APICallError({
633
+ message: errorMessage,
634
+ url: requestUrl,
635
+ requestBodyValues,
636
+ statusCode,
637
+ responseHeaders,
638
+ responseBody: httpErrorBody,
639
+ data: parsedErrorData,
640
+ });
641
+ const usage = toUsageFromPiMessage(lastErrorMessage, forensics.rawUsage);
642
+ let charge = forensics.charge;
643
+ if (parsedErrorData !== undefined && isRecord(parsedErrorData) && isRecord(parsedErrorData.usage)) {
644
+ charge = parsedErrorData.usage.cost ?? charge;
645
+ }
646
+ attachFailureEvidence(apiError, {
647
+ status: statusCode,
648
+ usage,
649
+ chargeEvidence: {
650
+ ...(charge === undefined ? {} : { charge }),
651
+ ...(forensics.rawUsage === undefined ? {} : { usage: forensics.rawUsage }),
652
+ response: {
653
+ ...(responseId === undefined ? {} : { id: responseId }),
654
+ ...(responseHeaders === undefined ? {} : { headers: responseHeaders }),
655
+ },
656
+ },
657
+ content: messageInfo.content,
658
+ reasoning: messageInfo.reasoning,
659
+ error: serializeError(apiError),
660
+ ...(failureRawBody === undefined ? {} : { rawBody: failureRawBody }),
661
+ outputObserved: outputObserved || messageInfo.content.length > 0 || messageInfo.reasoning.length > 0,
662
+ });
663
+ throw apiError;
664
+ }
665
+ const usage = toUsageFromPiMessage(finalMessage, forensics.rawUsage);
666
+ const chargeEvidence = {
667
+ ...(forensics.charge === undefined ? {} : { charge: forensics.charge }),
668
+ ...(forensics.rawUsage === undefined ? {} : { usage: forensics.rawUsage }),
669
+ response: {
670
+ ...(responseId === undefined ? {} : { id: responseId }),
671
+ ...(responseHeaders === undefined ? {} : { headers: responseHeaders }),
672
+ },
673
+ };
674
+ return {
675
+ id: responseId,
676
+ model: forensics.model ?? finalMessage.model ?? effectiveModel.id,
677
+ content: messageInfo.content,
678
+ reasoning: messageInfo.reasoning,
679
+ reasoningProjected: messageInfo.reasoningProjected,
680
+ finishReason: messageInfo.finishReason,
681
+ ...(messageInfo.rawFinishReason === undefined ? {} : { rawFinishReason: messageInfo.rawFinishReason }),
682
+ usage,
683
+ chargeEvidence,
684
+ logprobs: forensics.logprobs,
685
+ metadata: forensics.metadata,
686
+ wire: {
687
+ chunks: rawChunks.length,
688
+ emptyChunks: 0,
689
+ fields: {},
690
+ channels: forensics.channels,
691
+ toolCalls: messageInfo.toolCalls,
692
+ unmappedChunks: [],
693
+ },
694
+ ...(rawBody ? { rawBody: rawChunks } : {}),
695
+ };
696
+ };
697
+ //# sourceMappingURL=pi-transport.js.map