@qvac/inference 0.19.1 → 0.20.0

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.
Files changed (137) hide show
  1. package/NOTICE +43 -11
  2. package/dist/api/audio-gen.d.ts +42 -1
  3. package/dist/api/audio-gen.js +137 -51
  4. package/dist/api/index.d.ts +2 -1
  5. package/dist/api/index.js +2 -1
  6. package/dist/api/text-to-speech.d.ts +31 -4
  7. package/dist/api/text-to-speech.js +117 -21
  8. package/dist/api/vector-index.d.ts +142 -0
  9. package/dist/api/vector-index.js +168 -0
  10. package/dist/api/world.d.ts +2 -2
  11. package/dist/errors/index.d.ts +12 -0
  12. package/dist/errors/index.js +21 -0
  13. package/dist/handlers/vector-index.d.ts +2 -0
  14. package/dist/handlers/vector-index.js +40 -0
  15. package/dist/index.d.ts +1 -1
  16. package/dist/index.js +1 -1
  17. package/dist/models/registry/models.d.ts +351 -12
  18. package/dist/models/registry/models.js +2271 -1957
  19. package/dist/models/registry/resource-profiles.js +135 -0
  20. package/dist/plugins/builtin/audiogen-ggml/ops/audio-edit-stream.d.ts +2 -0
  21. package/dist/plugins/builtin/audiogen-ggml/ops/audio-edit-stream.js +65 -0
  22. package/dist/plugins/builtin/audiogen-ggml/ops/audio-gen-input.d.ts +9 -1
  23. package/dist/plugins/builtin/audiogen-ggml/ops/audio-gen-input.js +24 -7
  24. package/dist/plugins/builtin/audiogen-ggml/ops/audio-gen-run.d.ts +46 -0
  25. package/dist/plugins/builtin/audiogen-ggml/ops/audio-gen-run.js +218 -0
  26. package/dist/plugins/builtin/audiogen-ggml/ops/audio-gen-stream.d.ts +1 -1
  27. package/dist/plugins/builtin/audiogen-ggml/ops/audio-gen-stream.js +56 -225
  28. package/dist/plugins/builtin/audiogen-ggml/ops/audio-understand.d.ts +8 -0
  29. package/dist/plugins/builtin/audiogen-ggml/ops/audio-understand.js +28 -0
  30. package/dist/plugins/builtin/audiogen-ggml/plugin.d.ts +248 -0
  31. package/dist/plugins/builtin/audiogen-ggml/plugin.js +17 -1
  32. package/dist/plugins/builtin/bci-whispercpp-transcription/plugin.d.ts +5 -2
  33. package/dist/plugins/builtin/bci-whispercpp-transcription/plugin.js +1 -2
  34. package/dist/plugins/builtin/ggml-ocr/plugin.d.ts +1 -1
  35. package/dist/plugins/builtin/ggml-ocr/plugin.js +1 -2
  36. package/dist/plugins/builtin/llamacpp-completion/ops/batch-completion-stream.js +6 -4
  37. package/dist/plugins/builtin/llamacpp-completion/ops/cache-logger.d.ts +1 -2
  38. package/dist/plugins/builtin/llamacpp-completion/ops/cache-logger.js +0 -4
  39. package/dist/plugins/builtin/llamacpp-completion/ops/completion-stats.d.ts +1 -0
  40. package/dist/plugins/builtin/llamacpp-completion/ops/completion-stats.js +3 -0
  41. package/dist/plugins/builtin/llamacpp-completion/ops/completion-stream.d.ts +9 -0
  42. package/dist/plugins/builtin/llamacpp-completion/ops/completion-stream.js +96 -89
  43. package/dist/plugins/builtin/llamacpp-completion/ops/context-overflow.d.ts +0 -1
  44. package/dist/plugins/builtin/llamacpp-completion/ops/context-overflow.js +0 -31
  45. package/dist/plugins/builtin/llamacpp-completion/ops/kv-cache-session.d.ts +40 -24
  46. package/dist/plugins/builtin/llamacpp-completion/ops/kv-cache-session.js +187 -123
  47. package/dist/plugins/builtin/llamacpp-completion/ops/kv-cache-state.d.ts +11 -4
  48. package/dist/plugins/builtin/llamacpp-completion/ops/kv-cache-state.js +18 -16
  49. package/dist/plugins/builtin/llamacpp-completion/plugin.d.ts +8 -4
  50. package/dist/plugins/builtin/llamacpp-completion/plugin.js +1 -2
  51. package/dist/plugins/builtin/llamacpp-embedding/plugin.d.ts +1 -2
  52. package/dist/plugins/builtin/llamacpp-embedding/plugin.js +2 -38
  53. package/dist/plugins/builtin/llamacpp-embedding/transform.d.ts +9 -0
  54. package/dist/plugins/builtin/llamacpp-embedding/transform.js +42 -0
  55. package/dist/plugins/builtin/nmtcpp-translation/plugin.d.ts +1 -1
  56. package/dist/plugins/builtin/nmtcpp-translation/plugin.js +1 -2
  57. package/dist/plugins/builtin/parakeet-transcription/plugin.d.ts +1 -1
  58. package/dist/plugins/builtin/parakeet-transcription/plugin.js +1 -2
  59. package/dist/plugins/builtin/sdcpp-generation/ops/video.d.ts +1 -0
  60. package/dist/plugins/builtin/sdcpp-generation/ops/video.js +8 -2
  61. package/dist/plugins/builtin/sdcpp-generation/plugin.d.ts +9 -4
  62. package/dist/plugins/builtin/sdcpp-generation/plugin.js +51 -19
  63. package/dist/plugins/builtin/tts-ggml/ops/cancel-binding.d.ts +29 -0
  64. package/dist/plugins/builtin/tts-ggml/ops/cancel-binding.js +66 -0
  65. package/dist/plugins/builtin/tts-ggml/ops/parler-options.js +10 -0
  66. package/dist/plugins/builtin/tts-ggml/ops/text-to-speech-stream.d.ts +3 -5
  67. package/dist/plugins/builtin/tts-ggml/ops/text-to-speech-stream.js +143 -28
  68. package/dist/plugins/builtin/tts-ggml/ops/text-to-speech.d.ts +5 -2
  69. package/dist/plugins/builtin/tts-ggml/ops/text-to-speech.js +182 -66
  70. package/dist/plugins/builtin/tts-ggml/plugin.d.ts +267 -22
  71. package/dist/plugins/builtin/tts-ggml/plugin.js +94 -35
  72. package/dist/plugins/builtin/whispercpp-transcription/plugin.d.ts +1 -1
  73. package/dist/plugins/builtin/whispercpp-transcription/plugin.js +1 -2
  74. package/dist/plugins/index.d.ts +1 -1
  75. package/dist/plugins/index.js +1 -1
  76. package/dist/plugins/ops/load-model.js +22 -2
  77. package/dist/plugins/ops/translate.js +58 -12
  78. package/dist/plugins/ops/translatepsy-utils.d.ts +17 -0
  79. package/dist/plugins/ops/translatepsy-utils.js +32 -0
  80. package/dist/plugins/registry.d.ts +15 -0
  81. package/dist/plugins/registry.js +133 -15
  82. package/dist/plugins/turbovec-backend.d.ts +17 -0
  83. package/dist/plugins/turbovec-backend.js +99 -0
  84. package/dist/registry.js +4 -0
  85. package/dist/resources/model-fit/native-probe/advisory-fit.d.ts +46 -0
  86. package/dist/resources/model-fit/native-probe/advisory-fit.js +214 -0
  87. package/dist/resources/model-fit/native-probe/create-llama-fit-request.d.ts +26 -0
  88. package/dist/resources/model-fit/native-probe/create-llama-fit-request.js +171 -0
  89. package/dist/resources/model-fit/native-probe/run-in-process-fit.d.ts +42 -0
  90. package/dist/resources/model-fit/native-probe/run-in-process-fit.js +192 -0
  91. package/dist/resources/model-fit/native-probe/run-isolated-fit.d.ts +70 -0
  92. package/dist/resources/model-fit/native-probe/run-isolated-fit.js +372 -0
  93. package/dist/runtime/env.d.ts +2 -0
  94. package/dist/runtime/env.js +9 -2
  95. package/dist/runtime/lifecycle.js +2 -0
  96. package/dist/runtime/model-registry.d.ts +3 -1
  97. package/dist/runtime/model-registry.js +2 -1
  98. package/dist/runtime/request-context.js +17 -0
  99. package/dist/runtime/vector-index-backend.d.ts +25 -0
  100. package/dist/runtime/vector-index-backend.js +1 -0
  101. package/dist/runtime/vector-index-registry.d.ts +76 -0
  102. package/dist/runtime/vector-index-registry.js +161 -0
  103. package/dist/schemas/assess-model-fit.d.ts +44 -0
  104. package/dist/schemas/assess-model-fit.js +40 -0
  105. package/dist/schemas/audio-gen.d.ts +479 -4
  106. package/dist/schemas/audio-gen.js +376 -17
  107. package/dist/schemas/batch-completion-stream.d.ts +7 -0
  108. package/dist/schemas/batch-completion-stream.js +5 -3
  109. package/dist/schemas/common.d.ts +749 -15
  110. package/dist/schemas/common.js +8 -1
  111. package/dist/schemas/completion-event.d.ts +11 -1
  112. package/dist/schemas/completion-event.js +3 -0
  113. package/dist/schemas/completion-stream.d.ts +20 -0
  114. package/dist/schemas/completion-stream.js +41 -4
  115. package/dist/schemas/config.d.ts +48 -30
  116. package/dist/schemas/errors.d.ts +4 -0
  117. package/dist/schemas/errors.js +22 -0
  118. package/dist/schemas/index.d.ts +1 -0
  119. package/dist/schemas/index.js +1 -0
  120. package/dist/schemas/llamacpp-config.d.ts +4 -8
  121. package/dist/schemas/llamacpp-config.js +4 -4
  122. package/dist/schemas/load-model.d.ts +2503 -139
  123. package/dist/schemas/model-config-schemas.d.ts +228 -5
  124. package/dist/schemas/plugin.d.ts +10 -0
  125. package/dist/schemas/sdcpp-config.d.ts +221 -4
  126. package/dist/schemas/sdcpp-config.js +128 -22
  127. package/dist/schemas/text-to-speech.d.ts +780 -19
  128. package/dist/schemas/text-to-speech.js +298 -44
  129. package/dist/schemas/transcription-config.js +7 -4
  130. package/dist/schemas/vector-index.d.ts +194 -0
  131. package/dist/schemas/vector-index.js +185 -0
  132. package/dist/utils/addon-responses.d.ts +29 -3
  133. package/dist/utils/aggregate-events.d.ts +3 -1
  134. package/dist/utils/aggregate-events.js +7 -1
  135. package/dist/utils/tts-stats.d.ts +15 -11
  136. package/dist/utils/tts-stats.js +33 -11
  137. package/package.json +28 -25
@@ -15,6 +15,7 @@ import { loggingStreamRequestSchema, loggingStreamResponseSchema } from './loggi
15
15
  import { ttsRequestSchema, ttsResponseSchema, textToSpeechStreamRequestSchema, textToSpeechStreamResponseSchema } from './text-to-speech.js';
16
16
  import { errorResponseSchema } from './error.js';
17
17
  import { ragRequestSchema, ragResponseSchema, ragProgressUpdateSchema } from './rag.js';
18
+ import { vectorIndexRequestSchema, vectorIndexResponseSchema } from './vector-index.js';
18
19
  import { getModelInfoRequestSchema, getModelInfoResponseSchema, getLoadedModelInfoRequestSchema, getLoadedModelInfoResponseSchema } from './model-info.js';
19
20
  import { getSystemResourcesRequestSchema, getSystemResourcesResponseSchema } from './system-resources.js';
20
21
  import { assessModelFitRequestSchema, assessModelFitResponseSchema } from './assess-model-fit.js';
@@ -25,9 +26,11 @@ import { pluginInvokeRequestSchema, pluginInvokeResponseSchema, pluginInvokeStre
25
26
  import { modelRegistryListRequestSchema, modelRegistryListResponseSchema, modelRegistrySearchRequestSchema, modelRegistrySearchResponseSchema, modelRegistryGetModelRequestSchema, modelRegistryGetModelResponseSchema } from './registry.js';
26
27
  import { suspendRequestSchema, suspendResponseSchema, resumeRequestSchema, resumeResponseSchema, stateRequestSchema, stateResponseSchema } from './lifecycle.js';
27
28
  import { classifyRequestSchema, classifyResponseSchema } from './classification.js';
28
- import { audioGenStreamRequestSchema, audioGenStreamResponseSchema } from './audio-gen.js';
29
+ import { audioEditStreamRequestSchema, audioEditStreamResponseSchema, audioGenStreamRequestSchema, audioGenStreamResponseSchema, audioUnderstandRequestSchema, audioUnderstandResponseSchema } from './audio-gen.js';
29
30
  export const requestSchema = z.union([
30
31
  audioGenStreamRequestSchema,
32
+ audioEditStreamRequestSchema,
33
+ audioUnderstandRequestSchema,
31
34
  heartbeatRequestSchema,
32
35
  loadModelRequestSchema,
33
36
  downloadAssetRequestSchema,
@@ -46,6 +49,7 @@ export const requestSchema = z.union([
46
49
  textToSpeechStreamRequestSchema,
47
50
  cancelRequestSchema,
48
51
  ragRequestSchema,
52
+ vectorIndexRequestSchema,
49
53
  deleteCacheRequestSchema,
50
54
  getModelInfoRequestSchema,
51
55
  getLoadedModelInfoRequestSchema,
@@ -70,6 +74,8 @@ export const requestSchema = z.union([
70
74
  ]);
71
75
  export const responseSchema = z.discriminatedUnion('type', [
72
76
  audioGenStreamResponseSchema,
77
+ audioEditStreamResponseSchema,
78
+ audioUnderstandResponseSchema,
73
79
  heartbeatResponseSchema,
74
80
  loadModelResponseSchema,
75
81
  downloadAssetResponseSchema,
@@ -91,6 +97,7 @@ export const responseSchema = z.discriminatedUnion('type', [
91
97
  errorResponseSchema,
92
98
  ragResponseSchema,
93
99
  ragProgressUpdateSchema,
100
+ vectorIndexResponseSchema,
94
101
  deleteCacheResponseSchema,
95
102
  getModelInfoResponseSchema,
96
103
  getLoadedModelInfoResponseSchema,
@@ -1,5 +1,5 @@
1
1
  import { z } from 'zod';
2
- import { type Tool, type ToolCallWithCall } from './tools.js';
2
+ import { type Tool, type ToolCallError, type ToolCallWithCall } from './tools.js';
3
3
  import type { ToolDialect } from './completion-stream.js';
4
4
  export declare const completionStatsSchema: z.ZodObject<{
5
5
  timeToFirstToken: z.ZodOptional<z.ZodNumber>;
@@ -10,6 +10,7 @@ export declare const completionStatsSchema: z.ZodObject<{
10
10
  generatedTokens: z.ZodOptional<z.ZodNumber>;
11
11
  emittedTokens: z.ZodOptional<z.ZodNumber>;
12
12
  avgConcurrentSeq: z.ZodOptional<z.ZodNumber>;
13
+ toolDefinitionsDropped: z.ZodOptional<z.ZodNumber>;
13
14
  backendDevice: z.ZodOptional<z.ZodEnum<{
14
15
  gpu: "gpu";
15
16
  cpu: "cpu";
@@ -67,6 +68,7 @@ export declare const statsEventSchema: z.ZodObject<{
67
68
  generatedTokens: z.ZodOptional<z.ZodNumber>;
68
69
  emittedTokens: z.ZodOptional<z.ZodNumber>;
69
70
  avgConcurrentSeq: z.ZodOptional<z.ZodNumber>;
71
+ toolDefinitionsDropped: z.ZodOptional<z.ZodNumber>;
70
72
  backendDevice: z.ZodOptional<z.ZodEnum<{
71
73
  gpu: "gpu";
72
74
  cpu: "cpu";
@@ -150,6 +152,7 @@ export declare const completionEventSchema: z.ZodUnion<readonly [z.ZodObject<{
150
152
  generatedTokens: z.ZodOptional<z.ZodNumber>;
151
153
  emittedTokens: z.ZodOptional<z.ZodNumber>;
152
154
  avgConcurrentSeq: z.ZodOptional<z.ZodNumber>;
155
+ toolDefinitionsDropped: z.ZodOptional<z.ZodNumber>;
153
156
  backendDevice: z.ZodOptional<z.ZodEnum<{
154
157
  gpu: "gpu";
155
158
  cpu: "cpu";
@@ -192,6 +195,13 @@ export type CompletionFinal = {
192
195
  contentText: string;
193
196
  thinkingText?: string;
194
197
  toolCalls: ToolCallWithCall[];
198
+ /**
199
+ * Tool-call regions the model emitted that could not be turned into a
200
+ * `toolCalls` entry — malformed markup, arguments failing the tool's
201
+ * schema, or a call to an undeclared tool. Omitted when there were none.
202
+ * Lets a caller tell "the model made a mistake" from "no tool was called".
203
+ */
204
+ toolErrors?: ToolCallError[];
195
205
  stats?: CompletionStats;
196
206
  stopReason?: StopReason;
197
207
  raw: {
@@ -11,6 +11,9 @@ export const completionStatsSchema = z.object({
11
11
  // Non-empty addon stream pieces — prefer for usage reporting when present.
12
12
  emittedTokens: z.number().optional(),
13
13
  avgConcurrentSeq: z.number().optional(),
14
+ // Prompt renders that provably left the request's tool definitions out.
15
+ // Non-zero means the model never saw them; 0 is not proof it saw them all.
16
+ toolDefinitionsDropped: z.number().optional(),
14
17
  backendDevice: z.enum(['cpu', 'gpu']).optional()
15
18
  });
16
19
  export const seqSchema = z.number().int().nonnegative();
@@ -37,7 +37,22 @@ export declare const generationParamsSchema: z.ZodObject<{
37
37
  repeat_penalty: z.ZodOptional<z.ZodNumber>;
38
38
  reasoning_budget: z.ZodOptional<z.ZodNumber>;
39
39
  remove_thinking_from_context: z.ZodOptional<z.ZodBoolean>;
40
+ tool_choice: z.ZodOptional<z.ZodString>;
40
41
  }, z.core.$strict>;
42
+ /**
43
+ * Whether a `tool_choice` demands a tool call (`required` or a named tool),
44
+ * as opposed to `auto` / `none`, which only shape how the model may call one.
45
+ */
46
+ export declare function toolChoiceDemandsCall(toolChoice: string | undefined): boolean;
47
+ export declare function refineToolChoiceMatchesTools(data: {
48
+ tools?: {
49
+ type: 'function';
50
+ name: string;
51
+ }[] | undefined;
52
+ generationParams?: {
53
+ tool_choice?: string | undefined;
54
+ } | undefined;
55
+ }, ctx: z.RefinementCtx): void;
41
56
  export declare const responseFormatSchema: z.ZodDiscriminatedUnion<[z.ZodObject<{
42
57
  type: z.ZodLiteral<"text">;
43
58
  }, z.core.$strict>, z.ZodObject<{
@@ -105,6 +120,7 @@ export declare const completionClientParamsSchema: z.ZodObject<{
105
120
  repeat_penalty: z.ZodOptional<z.ZodNumber>;
106
121
  reasoning_budget: z.ZodOptional<z.ZodNumber>;
107
122
  remove_thinking_from_context: z.ZodOptional<z.ZodBoolean>;
123
+ tool_choice: z.ZodOptional<z.ZodString>;
108
124
  }, z.core.$strict>>;
109
125
  captureThinking: z.ZodOptional<z.ZodBoolean>;
110
126
  emitRawDeltas: z.ZodOptional<z.ZodBoolean>;
@@ -175,6 +191,7 @@ export declare const completionStreamRequestSchema: z.ZodObject<{
175
191
  repeat_penalty: z.ZodOptional<z.ZodNumber>;
176
192
  reasoning_budget: z.ZodOptional<z.ZodNumber>;
177
193
  remove_thinking_from_context: z.ZodOptional<z.ZodBoolean>;
194
+ tool_choice: z.ZodOptional<z.ZodString>;
178
195
  }, z.core.$strict>>;
179
196
  captureThinking: z.ZodOptional<z.ZodBoolean>;
180
197
  emitRawDeltas: z.ZodOptional<z.ZodBoolean>;
@@ -251,6 +268,7 @@ export declare const completionStreamResponseSchema: z.ZodObject<{
251
268
  generatedTokens: z.ZodOptional<z.ZodNumber>;
252
269
  emittedTokens: z.ZodOptional<z.ZodNumber>;
253
270
  avgConcurrentSeq: z.ZodOptional<z.ZodNumber>;
271
+ toolDefinitionsDropped: z.ZodOptional<z.ZodNumber>;
254
272
  backendDevice: z.ZodOptional<z.ZodEnum<{
255
273
  gpu: "gpu";
256
274
  cpu: "cpu";
@@ -323,6 +341,7 @@ export declare const completionOrchestrateRequestSchema: z.ZodObject<{
323
341
  repeat_penalty: z.ZodOptional<z.ZodNumber>;
324
342
  reasoning_budget: z.ZodOptional<z.ZodNumber>;
325
343
  remove_thinking_from_context: z.ZodOptional<z.ZodBoolean>;
344
+ tool_choice: z.ZodOptional<z.ZodString>;
326
345
  }, z.core.$strict>>;
327
346
  captureThinking: z.ZodOptional<z.ZodBoolean>;
328
347
  emitRawDeltas: z.ZodOptional<z.ZodBoolean>;
@@ -413,6 +432,7 @@ export declare const completionOrchestrateResponseSchema: z.ZodObject<{
413
432
  generatedTokens: z.ZodOptional<z.ZodNumber>;
414
433
  emittedTokens: z.ZodOptional<z.ZodNumber>;
415
434
  avgConcurrentSeq: z.ZodOptional<z.ZodNumber>;
435
+ toolDefinitionsDropped: z.ZodOptional<z.ZodNumber>;
416
436
  backendDevice: z.ZodOptional<z.ZodEnum<{
417
437
  gpu: "gpu";
418
438
  cpu: "cpu";
@@ -64,9 +64,42 @@ export const generationParamsSchema = z
64
64
  remove_thinking_from_context: z
65
65
  .boolean()
66
66
  .optional()
67
- .describe('When the model emits a reasoning block during generation (e.g. `<think>...</think>` for the Qwen3 family, `<|channel>thought ... <channel|>` for Gemma 4), drop those tokens from the KV cache at end-of-generation so subsequent turns do not accumulate reasoning history. Defaults to `false`, except the Qwen3 reasoning family (Qwen3, Qwen3.5, Qwen3.6, including MoE variants), which defaults to `true`. No-op for models without a recognised reasoning channel. Supported on recurrent / hybrid-SSM models (e.g. Qwen3.5) via a state snapshot and replay when the reasoning close marker is a single token; on such a model with a multi-token close marker, enabling this fails with an error.')
67
+ .describe('When the model emits a reasoning block during generation (e.g. `<think>...</think>` for the Qwen3 family, `<|channel>thought ... <channel|>` for Gemma 4), drop those tokens from the KV cache at end-of-generation so subsequent turns do not accumulate reasoning history. Defaults to `false`, except the Qwen3 reasoning family (Qwen3, Qwen3.5, Qwen3.6, including MoE variants), which defaults to `true`. No-op for models without a recognised reasoning channel. Supported on recurrent / hybrid-SSM models (e.g. Qwen3.5) via a state snapshot and replay when the reasoning close marker is a single token; on such a model with a multi-token close marker, enabling this fails with an error.'),
68
+ tool_choice: z
69
+ .string()
70
+ .min(1)
71
+ .optional()
72
+ .describe('Controls tool calling for a request that declares `tools`, in the OpenAI style. `"auto"` (default) lets the model decide and constrains output to the tool-call grammar only once it starts a call; `"required"` forces a tool call; `"none"` leaves the tool definitions in the prompt but disables the tool-call grammar; any other value names one declared tool and forces a call to it. `"required"` and a tool name are rejected when the request declares no tools, and fail the request rather than answering in prose when the model cannot honour them. Only honoured by llama.cpp-backed models; other backends ignore it.')
68
73
  })
69
74
  .strict();
75
+ const TOOL_CHOICE_MODES = new Set(['auto', 'none', 'required']);
76
+ /**
77
+ * Whether a `tool_choice` demands a tool call (`required` or a named tool),
78
+ * as opposed to `auto` / `none`, which only shape how the model may call one.
79
+ */
80
+ export function toolChoiceDemandsCall(toolChoice) {
81
+ return toolChoice !== undefined && toolChoice !== 'auto' && toolChoice !== 'none';
82
+ }
83
+ export function refineToolChoiceMatchesTools(data, ctx) {
84
+ const toolChoice = data.generationParams?.tool_choice;
85
+ if (toolChoice === undefined || TOOL_CHOICE_MODES.has(toolChoice)) {
86
+ if (toolChoice === 'required' && !(data.tools && data.tools.length > 0)) {
87
+ ctx.addIssue({
88
+ code: 'custom',
89
+ message: 'generationParams.tool_choice "required" needs at least one tool.',
90
+ path: ['generationParams', 'tool_choice']
91
+ });
92
+ }
93
+ return;
94
+ }
95
+ if (!data.tools?.some((tool) => tool.name === toolChoice)) {
96
+ ctx.addIssue({
97
+ code: 'custom',
98
+ message: `generationParams.tool_choice names "${toolChoice}", which is not one of the declared tools.`,
99
+ path: ['generationParams', 'tool_choice']
100
+ });
101
+ }
102
+ }
70
103
  const jsonSchemaObjectSchema = z.record(z.string(), z.unknown());
71
104
  export const responseFormatSchema = z.discriminatedUnion('type', [
72
105
  z.object({ type: z.literal('text') }).strict(),
@@ -154,12 +187,15 @@ function refineNoToolsWithStructuredOutput(data, ctx) {
154
187
  });
155
188
  }
156
189
  }
157
- export const completionClientParamsSchema = completionClientParamsBaseSchema.superRefine(refineNoToolsWithStructuredOutput);
190
+ export const completionClientParamsSchema = completionClientParamsBaseSchema
191
+ .superRefine(refineNoToolsWithStructuredOutput)
192
+ .superRefine(refineToolChoiceMatchesTools);
158
193
  export const completionStreamRequestSchema = completionClientParamsBaseSchema
159
194
  .extend({
160
195
  type: z.literal('completionStream')
161
196
  })
162
- .superRefine(refineNoToolsWithStructuredOutput);
197
+ .superRefine(refineNoToolsWithStructuredOutput)
198
+ .superRefine(refineToolChoiceMatchesTools);
163
199
  export const completionStreamResponseSchema = z
164
200
  .object({
165
201
  type: z.literal('completionStream'),
@@ -184,7 +220,8 @@ export const completionOrchestrateRequestSchema = completionClientParamsBaseSche
184
220
  'pass; a turn that requests tool calls consumes callbacks and starts ' +
185
221
  'the next turn. Defaults to 8.')
186
222
  })
187
- .superRefine(refineNoToolsWithStructuredOutput);
223
+ .superRefine(refineNoToolsWithStructuredOutput)
224
+ .superRefine(refineToolChoiceMatchesTools);
188
225
  /**
189
226
  * Downstream frame of the orchestrated completion duplex stream. Exactly one
190
227
  * payload field is set per frame:
@@ -61,7 +61,6 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
61
61
  'split-mode': z.ZodOptional<z.ZodEnum<{
62
62
  none: "none";
63
63
  layer: "layer";
64
- row: "row";
65
64
  tensor: "tensor";
66
65
  }>>;
67
66
  'flash-attn': z.ZodOptional<z.ZodEnum<{
@@ -153,7 +152,6 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
153
152
  splitMode: z.ZodOptional<z.ZodEnum<{
154
153
  none: "none";
155
154
  layer: "layer";
156
- row: "row";
157
155
  }>>;
158
156
  tensorSplit: z.ZodOptional<z.ZodString>;
159
157
  verbosity: z.ZodOptional<z.ZodEnum<{
@@ -379,6 +377,7 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
379
377
  auto: "auto";
380
378
  f32: "f32";
381
379
  f16: "f16";
380
+ q8_0: "q8_0";
382
381
  bf16: "bf16";
383
382
  q2_k: "q2_k";
384
383
  q3_k: "q3_k";
@@ -389,7 +388,6 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
389
388
  q5_1: "q5_1";
390
389
  q5_k: "q5_k";
391
390
  q6_k: "q6_k";
392
- q8_0: "q8_0";
393
391
  }>>>;
394
392
  rng: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
395
393
  cpu: "cpu";
@@ -401,12 +399,17 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
401
399
  cuda: "cuda";
402
400
  std_default: "std_default";
403
401
  }>>>;
404
- clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
405
- vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
406
402
  vae_auto_cpu_fallback: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
407
403
  vae_auto_cpu_fallback_memory_ratio: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
408
404
  vae_tiling: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
409
405
  offload_to_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
406
+ control_net_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
407
+ clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
408
+ vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
409
+ backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
410
+ params_backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
411
+ max_vram: z.ZodOptional<z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>>;
412
+ stream_layers: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
410
413
  flash_attn: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
411
414
  diffusion_fa: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
412
415
  lora_apply_mode: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
@@ -992,7 +995,6 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
992
995
  'split-mode': z.ZodOptional<z.ZodEnum<{
993
996
  none: "none";
994
997
  layer: "layer";
995
- row: "row";
996
998
  tensor: "tensor";
997
999
  }>>;
998
1000
  'flash-attn': z.ZodOptional<z.ZodEnum<{
@@ -1084,7 +1086,6 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
1084
1086
  splitMode: z.ZodOptional<z.ZodEnum<{
1085
1087
  none: "none";
1086
1088
  layer: "layer";
1087
- row: "row";
1088
1089
  }>>;
1089
1090
  tensorSplit: z.ZodOptional<z.ZodString>;
1090
1091
  verbosity: z.ZodOptional<z.ZodEnum<{
@@ -1310,6 +1311,7 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
1310
1311
  auto: "auto";
1311
1312
  f32: "f32";
1312
1313
  f16: "f16";
1314
+ q8_0: "q8_0";
1313
1315
  bf16: "bf16";
1314
1316
  q2_k: "q2_k";
1315
1317
  q3_k: "q3_k";
@@ -1320,7 +1322,6 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
1320
1322
  q5_1: "q5_1";
1321
1323
  q5_k: "q5_k";
1322
1324
  q6_k: "q6_k";
1323
- q8_0: "q8_0";
1324
1325
  }>>>;
1325
1326
  rng: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
1326
1327
  cpu: "cpu";
@@ -1332,12 +1333,17 @@ export declare const deviceConfigDefaultsSchema: z.ZodObject<{
1332
1333
  cuda: "cuda";
1333
1334
  std_default: "std_default";
1334
1335
  }>>>;
1335
- clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1336
- vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1337
1336
  vae_auto_cpu_fallback: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1338
1337
  vae_auto_cpu_fallback_memory_ratio: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
1339
1338
  vae_tiling: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1340
1339
  offload_to_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1340
+ control_net_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
1341
+ clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
1342
+ vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
1343
+ backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
1344
+ params_backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
1345
+ max_vram: z.ZodOptional<z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>>;
1346
+ stream_layers: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1341
1347
  flash_attn: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1342
1348
  diffusion_fa: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
1343
1349
  lora_apply_mode: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
@@ -1943,7 +1949,6 @@ export declare const devicePatternSchema: z.ZodObject<{
1943
1949
  'split-mode': z.ZodOptional<z.ZodEnum<{
1944
1950
  none: "none";
1945
1951
  layer: "layer";
1946
- row: "row";
1947
1952
  tensor: "tensor";
1948
1953
  }>>;
1949
1954
  'flash-attn': z.ZodOptional<z.ZodEnum<{
@@ -2035,7 +2040,6 @@ export declare const devicePatternSchema: z.ZodObject<{
2035
2040
  splitMode: z.ZodOptional<z.ZodEnum<{
2036
2041
  none: "none";
2037
2042
  layer: "layer";
2038
- row: "row";
2039
2043
  }>>;
2040
2044
  tensorSplit: z.ZodOptional<z.ZodString>;
2041
2045
  verbosity: z.ZodOptional<z.ZodEnum<{
@@ -2261,6 +2265,7 @@ export declare const devicePatternSchema: z.ZodObject<{
2261
2265
  auto: "auto";
2262
2266
  f32: "f32";
2263
2267
  f16: "f16";
2268
+ q8_0: "q8_0";
2264
2269
  bf16: "bf16";
2265
2270
  q2_k: "q2_k";
2266
2271
  q3_k: "q3_k";
@@ -2271,7 +2276,6 @@ export declare const devicePatternSchema: z.ZodObject<{
2271
2276
  q5_1: "q5_1";
2272
2277
  q5_k: "q5_k";
2273
2278
  q6_k: "q6_k";
2274
- q8_0: "q8_0";
2275
2279
  }>>>;
2276
2280
  rng: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
2277
2281
  cpu: "cpu";
@@ -2283,12 +2287,17 @@ export declare const devicePatternSchema: z.ZodObject<{
2283
2287
  cuda: "cuda";
2284
2288
  std_default: "std_default";
2285
2289
  }>>>;
2286
- clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
2287
- vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
2288
2290
  vae_auto_cpu_fallback: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
2289
2291
  vae_auto_cpu_fallback_memory_ratio: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
2290
2292
  vae_tiling: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
2291
2293
  offload_to_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
2294
+ control_net_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
2295
+ clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
2296
+ vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
2297
+ backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
2298
+ params_backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
2299
+ max_vram: z.ZodOptional<z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>>;
2300
+ stream_layers: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
2292
2301
  flash_attn: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
2293
2302
  diffusion_fa: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
2294
2303
  lora_apply_mode: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
@@ -2874,7 +2883,6 @@ export declare const devicePatternSchema: z.ZodObject<{
2874
2883
  'split-mode': z.ZodOptional<z.ZodEnum<{
2875
2884
  none: "none";
2876
2885
  layer: "layer";
2877
- row: "row";
2878
2886
  tensor: "tensor";
2879
2887
  }>>;
2880
2888
  'flash-attn': z.ZodOptional<z.ZodEnum<{
@@ -2966,7 +2974,6 @@ export declare const devicePatternSchema: z.ZodObject<{
2966
2974
  splitMode: z.ZodOptional<z.ZodEnum<{
2967
2975
  none: "none";
2968
2976
  layer: "layer";
2969
- row: "row";
2970
2977
  }>>;
2971
2978
  tensorSplit: z.ZodOptional<z.ZodString>;
2972
2979
  verbosity: z.ZodOptional<z.ZodEnum<{
@@ -3192,6 +3199,7 @@ export declare const devicePatternSchema: z.ZodObject<{
3192
3199
  auto: "auto";
3193
3200
  f32: "f32";
3194
3201
  f16: "f16";
3202
+ q8_0: "q8_0";
3195
3203
  bf16: "bf16";
3196
3204
  q2_k: "q2_k";
3197
3205
  q3_k: "q3_k";
@@ -3202,7 +3210,6 @@ export declare const devicePatternSchema: z.ZodObject<{
3202
3210
  q5_1: "q5_1";
3203
3211
  q5_k: "q5_k";
3204
3212
  q6_k: "q6_k";
3205
- q8_0: "q8_0";
3206
3213
  }>>>;
3207
3214
  rng: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
3208
3215
  cpu: "cpu";
@@ -3214,12 +3221,17 @@ export declare const devicePatternSchema: z.ZodObject<{
3214
3221
  cuda: "cuda";
3215
3222
  std_default: "std_default";
3216
3223
  }>>>;
3217
- clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
3218
- vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
3219
3224
  vae_auto_cpu_fallback: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
3220
3225
  vae_auto_cpu_fallback_memory_ratio: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
3221
3226
  vae_tiling: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
3222
3227
  offload_to_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
3228
+ control_net_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
3229
+ clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
3230
+ vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
3231
+ backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
3232
+ params_backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
3233
+ max_vram: z.ZodOptional<z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>>;
3234
+ stream_layers: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
3223
3235
  flash_attn: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
3224
3236
  diffusion_fa: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
3225
3237
  lora_apply_mode: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
@@ -3848,7 +3860,6 @@ export declare const qvacConfigSchema: z.ZodObject<{
3848
3860
  'split-mode': z.ZodOptional<z.ZodEnum<{
3849
3861
  none: "none";
3850
3862
  layer: "layer";
3851
- row: "row";
3852
3863
  tensor: "tensor";
3853
3864
  }>>;
3854
3865
  'flash-attn': z.ZodOptional<z.ZodEnum<{
@@ -3940,7 +3951,6 @@ export declare const qvacConfigSchema: z.ZodObject<{
3940
3951
  splitMode: z.ZodOptional<z.ZodEnum<{
3941
3952
  none: "none";
3942
3953
  layer: "layer";
3943
- row: "row";
3944
3954
  }>>;
3945
3955
  tensorSplit: z.ZodOptional<z.ZodString>;
3946
3956
  verbosity: z.ZodOptional<z.ZodEnum<{
@@ -4166,6 +4176,7 @@ export declare const qvacConfigSchema: z.ZodObject<{
4166
4176
  auto: "auto";
4167
4177
  f32: "f32";
4168
4178
  f16: "f16";
4179
+ q8_0: "q8_0";
4169
4180
  bf16: "bf16";
4170
4181
  q2_k: "q2_k";
4171
4182
  q3_k: "q3_k";
@@ -4176,7 +4187,6 @@ export declare const qvacConfigSchema: z.ZodObject<{
4176
4187
  q5_1: "q5_1";
4177
4188
  q5_k: "q5_k";
4178
4189
  q6_k: "q6_k";
4179
- q8_0: "q8_0";
4180
4190
  }>>>;
4181
4191
  rng: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
4182
4192
  cpu: "cpu";
@@ -4188,12 +4198,17 @@ export declare const qvacConfigSchema: z.ZodObject<{
4188
4198
  cuda: "cuda";
4189
4199
  std_default: "std_default";
4190
4200
  }>>>;
4191
- clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
4192
- vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
4193
4201
  vae_auto_cpu_fallback: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
4194
4202
  vae_auto_cpu_fallback_memory_ratio: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
4195
4203
  vae_tiling: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
4196
4204
  offload_to_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
4205
+ control_net_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
4206
+ clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
4207
+ vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
4208
+ backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
4209
+ params_backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
4210
+ max_vram: z.ZodOptional<z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>>;
4211
+ stream_layers: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
4197
4212
  flash_attn: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
4198
4213
  diffusion_fa: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
4199
4214
  lora_apply_mode: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
@@ -4779,7 +4794,6 @@ export declare const qvacConfigSchema: z.ZodObject<{
4779
4794
  'split-mode': z.ZodOptional<z.ZodEnum<{
4780
4795
  none: "none";
4781
4796
  layer: "layer";
4782
- row: "row";
4783
4797
  tensor: "tensor";
4784
4798
  }>>;
4785
4799
  'flash-attn': z.ZodOptional<z.ZodEnum<{
@@ -4871,7 +4885,6 @@ export declare const qvacConfigSchema: z.ZodObject<{
4871
4885
  splitMode: z.ZodOptional<z.ZodEnum<{
4872
4886
  none: "none";
4873
4887
  layer: "layer";
4874
- row: "row";
4875
4888
  }>>;
4876
4889
  tensorSplit: z.ZodOptional<z.ZodString>;
4877
4890
  verbosity: z.ZodOptional<z.ZodEnum<{
@@ -5097,6 +5110,7 @@ export declare const qvacConfigSchema: z.ZodObject<{
5097
5110
  auto: "auto";
5098
5111
  f32: "f32";
5099
5112
  f16: "f16";
5113
+ q8_0: "q8_0";
5100
5114
  bf16: "bf16";
5101
5115
  q2_k: "q2_k";
5102
5116
  q3_k: "q3_k";
@@ -5107,7 +5121,6 @@ export declare const qvacConfigSchema: z.ZodObject<{
5107
5121
  q5_1: "q5_1";
5108
5122
  q5_k: "q5_k";
5109
5123
  q6_k: "q6_k";
5110
- q8_0: "q8_0";
5111
5124
  }>>>;
5112
5125
  rng: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
5113
5126
  cpu: "cpu";
@@ -5119,12 +5132,17 @@ export declare const qvacConfigSchema: z.ZodObject<{
5119
5132
  cuda: "cuda";
5120
5133
  std_default: "std_default";
5121
5134
  }>>>;
5122
- clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
5123
- vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
5124
5135
  vae_auto_cpu_fallback: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
5125
5136
  vae_auto_cpu_fallback_memory_ratio: z.ZodOptional<z.ZodOptional<z.ZodNumber>>;
5126
5137
  vae_tiling: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
5127
5138
  offload_to_cpu: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
5139
+ control_net_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
5140
+ clip_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
5141
+ vae_on_cpu: z.ZodOptional<z.ZodOptional<z.ZodCustom<never, never>>>;
5142
+ backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
5143
+ params_backend: z.ZodOptional<z.ZodOptional<z.ZodString>>;
5144
+ max_vram: z.ZodOptional<z.ZodOptional<z.ZodUnion<readonly [z.ZodNumber, z.ZodString]>>>;
5145
+ stream_layers: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
5128
5146
  flash_attn: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
5129
5147
  diffusion_fa: z.ZodOptional<z.ZodOptional<z.ZodBoolean>>;
5130
5148
  lora_apply_mode: z.ZodOptional<z.ZodOptional<z.ZodEnum<{
@@ -62,6 +62,10 @@ export declare const ERROR_CODES: {
62
62
  readonly RAG_LIST_WORKSPACES_FAILED: 52809;
63
63
  readonly RAG_CHUNK_FAILED: 52810;
64
64
  readonly RAG_WORKSPACE_NOT_OPEN: 52811;
65
+ readonly VECTOR_INDEX_PROVIDER_UNAVAILABLE: 52850;
66
+ readonly VECTOR_INDEX_NOT_FOUND: 52851;
67
+ readonly VECTOR_INDEX_INVALID_VECTORS: 52852;
68
+ readonly VECTOR_INDEX_FAILED: 52853;
65
69
  readonly FILE_NOT_FOUND: 53000;
66
70
  readonly DOWNLOAD_CANCELLED: 53001;
67
71
  readonly CHECKSUM_VALIDATION_FAILED: 53002;
@@ -72,6 +72,11 @@ export const ERROR_CODES = {
72
72
  RAG_LIST_WORKSPACES_FAILED: 52809,
73
73
  RAG_CHUNK_FAILED: 52810,
74
74
  RAG_WORKSPACE_NOT_OPEN: 52811,
75
+ // Vector index (52,850-52,899)
76
+ VECTOR_INDEX_PROVIDER_UNAVAILABLE: 52850,
77
+ VECTOR_INDEX_NOT_FOUND: 52851,
78
+ VECTOR_INDEX_INVALID_VECTORS: 52852,
79
+ VECTOR_INDEX_FAILED: 52853,
75
80
  // Download / resource (53,000-53,199)
76
81
  FILE_NOT_FOUND: 53000,
77
82
  DOWNLOAD_CANCELLED: 53001,
@@ -409,6 +414,23 @@ const errorDefinitions = {
409
414
  name: 'RAG_WORKSPACE_NOT_OPEN',
410
415
  message: (workspace) => `RAG workspace '${workspace}' is not open`
411
416
  },
417
+ // Vector index
418
+ [ERROR_CODES.VECTOR_INDEX_PROVIDER_UNAVAILABLE]: {
419
+ name: 'VECTOR_INDEX_PROVIDER_UNAVAILABLE',
420
+ message: () => 'No vector index provider is registered. Register a plugin that exposes the turbovecIndexProvider capability, such as the built-in llamacpp-embedding plugin'
421
+ },
422
+ [ERROR_CODES.VECTOR_INDEX_NOT_FOUND]: {
423
+ name: 'VECTOR_INDEX_NOT_FOUND',
424
+ message: (indexId) => `Vector index '${indexId}' is not open. It was disposed or the worker restarted; create a new index or load a snapshot`
425
+ },
426
+ [ERROR_CODES.VECTOR_INDEX_INVALID_VECTORS]: {
427
+ name: 'VECTOR_INDEX_INVALID_VECTORS',
428
+ message: (details) => `Invalid vector input: ${details}`
429
+ },
430
+ [ERROR_CODES.VECTOR_INDEX_FAILED]: {
431
+ name: 'VECTOR_INDEX_FAILED',
432
+ message: (details) => `Vector index operation failed: ${details}`
433
+ },
412
434
  // Download / resource
413
435
  [ERROR_CODES.FILE_NOT_FOUND]: {
414
436
  name: 'FILE_NOT_FOUND',
@@ -23,6 +23,7 @@ export * from './transcription-config.js';
23
23
  export * from './text-to-speech.js';
24
24
  export * from './error.js';
25
25
  export * from './rag.js';
26
+ export * from './vector-index.js';
26
27
  export * from './ocr.js';
27
28
  export * from './sdcpp-config.js';
28
29
  export * from './vla.js';
@@ -24,6 +24,7 @@ export * from './transcription-config.js';
24
24
  export * from './text-to-speech.js';
25
25
  export * from './error.js';
26
26
  export * from './rag.js';
27
+ export * from './vector-index.js';
27
28
  export * from './ocr.js';
28
29
  export * from './sdcpp-config.js';
29
30
  export * from './vla.js';