modelmix 4.7.4 → 5.0.0

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.
package/index.d.ts CHANGED
@@ -10,6 +10,20 @@ export type DebugLevel = 0 | 1 | 2 | 3 | 4;
10
10
  /** Unified effort: -1 = adaptive, 0–100 = intensity. */
11
11
  export type EffortValue = number;
12
12
 
13
+ export interface PromptCacheOptions {
14
+ mode?: 'implicit' | 'explicit';
15
+ ttl?: '30m';
16
+ }
17
+
18
+ export interface CacheBreakpoint {
19
+ breakpoint: true;
20
+ }
21
+
22
+ export interface AnthropicCacheControl {
23
+ type: 'ephemeral';
24
+ ttl?: '5m' | '1h';
25
+ }
26
+
13
27
  export interface BottleneckConfig {
14
28
  maxConcurrent?: number;
15
29
  minTime?: number;
@@ -35,6 +49,13 @@ export interface ModelMixOptions {
35
49
  model?: string;
36
50
  messages?: ChatMessage[];
37
51
  response_format?: { type: string; [key: string]: unknown };
52
+ prompt_cache_key?: string;
53
+ /** OpenAI models before GPT-5.6. GPT-5.6 uses prompt_cache_options.ttl. */
54
+ prompt_cache_retention?: string;
55
+ /** GPT-5.6 Responses API prompt caching controls. */
56
+ prompt_cache_options?: PromptCacheOptions;
57
+ /** Anthropic automatic caching, or the policy used by neutral explicit breakpoints. */
58
+ cache_control?: AnthropicCacheControl;
38
59
  [key: string]: unknown;
39
60
  }
40
61
 
@@ -49,7 +70,7 @@ export interface ModelMixConfig {
49
70
  roundRobin?: boolean;
50
71
  /** Unified effort (-1 adaptive, or 0–100). Not a native provider field. */
51
72
  effort?: EffortValue | null;
52
- replace?: Record<string, string>;
73
+ replace?: Record<string, unknown>;
53
74
  schema?: Record<string, unknown>;
54
75
  [key: string]: unknown;
55
76
  }
@@ -82,11 +103,15 @@ export interface ModelAttachArgs {
82
103
 
83
104
  export interface RoleOptions {
84
105
  role?: MessageRole;
106
+ /** Provider-neutral explicit cache breakpoint, translated by each adapter. */
107
+ cache?: CacheBreakpoint;
85
108
  }
86
109
 
87
110
  export interface TextContentPart {
88
111
  type: 'text';
89
112
  text: string;
113
+ cache?: CacheBreakpoint;
114
+ cache_control?: AnthropicCacheControl;
90
115
  }
91
116
 
92
117
  export interface ImageContentPart {
@@ -96,6 +121,8 @@ export interface ImageContentPart {
96
121
  media_type?: string;
97
122
  data: string | Buffer;
98
123
  };
124
+ cache?: CacheBreakpoint;
125
+ cache_control?: AnthropicCacheControl;
99
126
  }
100
127
 
101
128
  export type ContentPart = TextContentPart | ImageContentPart | Record<string, unknown>;
@@ -121,12 +148,34 @@ export interface ToolCall {
121
148
  };
122
149
  }
123
150
 
151
+ export interface TokenCostBreakdown {
152
+ uncachedInput: number;
153
+ cachedInput: number;
154
+ cacheWrite: number;
155
+ cacheWrite5m: number;
156
+ cacheWrite1h: number;
157
+ output: number;
158
+ total: number;
159
+ }
160
+
124
161
  export interface TokenUsage {
125
162
  input: number;
126
163
  output: number;
127
164
  total: number;
128
165
  cached: number;
129
- cost?: number | null;
166
+ cacheWrite: number;
167
+ cacheWrite5m: number;
168
+ cacheWrite1h: number;
169
+ uncachedInput: number;
170
+ cacheHitRate: number;
171
+ /** USD avoided versus billing cache reads at the uncached input rate. */
172
+ cacheSavings: number;
173
+ /** Extra USD paid for cache writes versus ordinary uncached input. */
174
+ cacheWritePremium: number;
175
+ /** Full future cache hits needed to recover the current write premium. */
176
+ breakEvenHits: number;
177
+ cost: number;
178
+ costBreakdown: TokenCostBreakdown;
130
179
  speed?: number;
131
180
  }
132
181
 
@@ -244,9 +293,51 @@ export declare class ModelMix {
244
293
  static truncate(str: string, maxLen?: number): string;
245
294
  static calculateCost(
246
295
  modelKey: string,
247
- tokens: { input: number; output: number }
296
+ tokens: {
297
+ input: number;
298
+ output: number;
299
+ total?: number;
300
+ cached?: number;
301
+ cacheWrite?: number;
302
+ cacheWrite5m?: number;
303
+ cacheWrite1h?: number;
304
+ }
248
305
  ): number | null;
306
+ static calculateCostBreakdown(
307
+ modelKey: string,
308
+ tokens: {
309
+ input: number;
310
+ output: number;
311
+ total?: number;
312
+ cached?: number;
313
+ cacheWrite?: number;
314
+ cacheWrite5m?: number;
315
+ cacheWrite1h?: number;
316
+ }
317
+ ): TokenCostBreakdown;
318
+ static calculateCacheMetrics(
319
+ modelKey: string,
320
+ tokens: {
321
+ input: number;
322
+ output: number;
323
+ total?: number;
324
+ cached?: number;
325
+ cacheWrite?: number;
326
+ cacheWrite5m?: number;
327
+ cacheWrite1h?: number;
328
+ }
329
+ ): Pick<TokenUsage, 'cacheSavings' | 'cacheWritePremium' | 'breakEvenHits'>;
330
+ static normalizeTokenUsage(tokens?: {
331
+ input?: number;
332
+ output?: number;
333
+ total?: number;
334
+ cached?: number;
335
+ cacheWrite?: number;
336
+ cacheWrite5m?: number;
337
+ cacheWrite1h?: number;
338
+ }): TokenUsage;
249
339
  static extractCacheTokens(usage?: Record<string, unknown>): number;
340
+ static extractCacheWriteTokens(usage?: Record<string, unknown>): number;
250
341
  static formatInputSummary(
251
342
  messages: ChatMessage[],
252
343
  system: string,
@@ -256,7 +347,7 @@ export declare class ModelMix {
256
347
  static hasToolInteraction(message: ChatMessage | null | undefined): boolean;
257
348
 
258
349
  new(setup?: ModelMixSetup): ModelMix;
259
- replace(keyValues: Record<string, string>): this;
350
+ replace(keyValues: Record<string, unknown>): this;
260
351
  effort(value: EffortValue): this;
261
352
  attach(key: string, provider: MixCustom): this;
262
353
 
@@ -285,11 +376,14 @@ export declare class ModelMix {
285
376
  gptOss(args?: ModelAttachArgs): this;
286
377
 
287
378
  // Anthropic
379
+ fable50(args?: ModelAttachArgs): this;
288
380
  fable5(args?: ModelAttachArgs): this;
381
+ opus50(args?: ModelAttachArgs): this;
289
382
  opus5(args?: ModelAttachArgs): this;
290
383
  opus48(args?: ModelAttachArgs): this;
291
384
  opus47(args?: ModelAttachArgs): this;
292
385
  opus46(args?: ModelAttachArgs): this;
386
+ sonnet50(args?: ModelAttachArgs): this;
293
387
  sonnet5(args?: ModelAttachArgs): this;
294
388
  sonnet46(args?: ModelAttachArgs): this;
295
389
  sonnet45(args?: ModelAttachArgs): this;
@@ -359,7 +453,6 @@ export declare class ModelMix {
359
453
 
360
454
  replaceKeyFromFile(key: string, filePath: string): this;
361
455
  groupByRoles(messages: ChatMessage[]): ChatMessage[];
362
- applyTemplate(): void;
363
456
  prepareMessages(): Promise<void>;
364
457
  readFile(filePath: string, options?: { encoding?: BufferEncoding | null }): string | Buffer;
365
458
  execute(args?: CreateArgs): Promise<ModelMixResult>;
@@ -386,6 +479,7 @@ export declare class MixCustom {
386
479
  getDefaultConfig(customConfig?: Record<string, unknown>): Record<string, unknown>;
387
480
  getDefaultHeaders(customHeaders?: Record<string, string>): Record<string, string>;
388
481
  convertMessages(messages: ChatMessage[], config?: ModelMixConfig): ChatMessage[];
482
+ sanitizeCacheOptions(options: ModelMixOptions): void;
389
483
 
390
484
  static stripContentTypeHeader(headers?: Record<string, string>): Record<string, string>;
391
485
  static createMultipartFormData(args?: {
@@ -402,16 +496,11 @@ export declare class MixCustom {
402
496
  static extractTokens(data: unknown): TokenUsage;
403
497
 
404
498
  create(args?: CreateArgs): Promise<ModelMixResult>;
405
- handleError(
406
- error: unknown,
407
- context: CreateArgs
408
- ): {
499
+ handleError(error: unknown): {
409
500
  message: string;
410
501
  statusCode: number | null;
411
502
  details: unknown;
412
503
  stack?: string;
413
- config?: ModelMixConfig;
414
- options?: ModelMixOptions;
415
504
  };
416
505
  processStream(response: { data: NodeJS.ReadableStream }): Promise<ModelMixResult>;
417
506
  extractDelta(data: unknown): string;