visual-ai-assertions 0.25.0 → 0.27.0

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 (8) hide show
  1. package/README.md +146 -76
  2. package/dist/index.cjs +225 -137
  3. package/dist/index.cjs.map +1 -1
  4. package/dist/index.d.cts +47 -13
  5. package/dist/index.d.ts +47 -13
  6. package/dist/index.js +225 -137
  7. package/dist/index.js.map +1 -1
  8. package/package.json +2 -5
package/dist/index.d.cts CHANGED
@@ -1,7 +1,21 @@
1
1
  import { z } from 'zod';
2
2
 
3
3
  /** Supported reasoning effort levels. */
4
+ /**
5
+ * Abstract reasoning-effort hint. Each provider driver maps it to its native
6
+ * mechanism; levels a provider lacks clamp to its nearest tier.
7
+ *
8
+ * `minimal` is **not universally accepted**. OpenAI rejects it per-model with
9
+ * HTTP 400 ("Unsupported value: 'minimal' is not supported with the
10
+ * '<model>' model"), observed on `gpt-6-astra` and `gpt-6.1-sol`, whose
11
+ * supported set is low/medium/high/xhigh/max, and on `gpt-6-sol` and
12
+ * `gpt-6-luna`, which accept none/low/medium/high/xhigh/max. Gemini defines
13
+ * the tier but some models (e.g. Gemini 3.1 Pro) reject it, so Google and
14
+ * OpenRouter clamp it to `low` rather than send it. Use `minimal` only
15
+ * against an OpenAI model known to accept it; `low` is the portable floor.
16
+ */
4
17
  declare const ReasoningEffort: {
18
+ readonly MINIMAL: "minimal";
5
19
  readonly LOW: "low";
6
20
  readonly MEDIUM: "medium";
7
21
  readonly HIGH: "high";
@@ -34,16 +48,22 @@ declare const Model: {
34
48
  readonly Anthropic: {
35
49
  readonly FABLE_5_1: "claude-fable-5-1";
36
50
  readonly FABLE_5: "claude-fable-5";
51
+ readonly OPUS_5_5: "claude-opus-5-5";
37
52
  readonly OPUS_5: "claude-opus-5";
38
53
  readonly OPUS_4_8: "claude-opus-4-8";
39
54
  readonly OPUS_4_7: "claude-opus-4-7";
40
55
  readonly OPUS_4_6: "claude-opus-4-6";
56
+ readonly SONNET_5_5: "claude-sonnet-5-5";
41
57
  readonly SONNET_5: "claude-sonnet-5";
42
58
  readonly SONNET_4_6: "claude-sonnet-4-6";
59
+ readonly HAIKU_5_5: "claude-haiku-5-5";
43
60
  readonly HAIKU_4_5: "claude-haiku-4-5";
44
61
  };
45
62
  readonly OpenAI: {
46
63
  readonly GPT_6_ASTRA: "gpt-6-astra";
64
+ readonly GPT_6_1_SOL: "gpt-6.1-sol";
65
+ readonly GPT_6_SOL: "gpt-6-sol";
66
+ readonly GPT_6_LUNA: "gpt-6-luna";
47
67
  readonly GPT_5_6_SOL: "gpt-5.6-sol";
48
68
  readonly GPT_5_6_TERRA: "gpt-5.6-terra";
49
69
  readonly GPT_5_6_LUNA: "gpt-5.6-luna";
@@ -72,6 +92,7 @@ declare const Model: {
72
92
  */
73
93
  readonly OpenRouter: {
74
94
  readonly MUSE_SPARK_1_3: "meta/muse-spark-1.3";
95
+ readonly GROK_4_7: "x-ai/grok-4.7";
75
96
  readonly GROK_4_6: "x-ai/grok-4.6";
76
97
  readonly GROK_4_5: "x-ai/grok-4.5";
77
98
  readonly KIMI_K3: "moonshotai/kimi-k3";
@@ -80,16 +101,28 @@ declare const Model: {
80
101
  readonly QWEN_3_7_PLUS: "qwen/qwen3.7-plus";
81
102
  readonly QWEN_3_6_FLASH: "qwen/qwen3.6-flash";
82
103
  readonly GLM_5_3_FLASH: "z-ai/glm-5.3-flash";
104
+ readonly MIMO_V2_6_PRO: "xiaomi/mimo-v2.6-pro";
83
105
  };
84
106
  };
85
107
  /** Union of all built-in model name literals exposed by `Model`. */
86
108
  type KnownModelName = (typeof Model.Anthropic)[keyof typeof Model.Anthropic] | (typeof Model.OpenAI)[keyof typeof Model.OpenAI] | (typeof Model.Google)[keyof typeof Model.Google] | (typeof Model.OpenRouter)[keyof typeof Model.OpenRouter];
87
- /** Default model selection used when a caller omits `config.model`. */
109
+ /**
110
+ * Default model selection used when a caller omits `config.model`.
111
+ *
112
+ * Two of these carry constraints the older defaults did not:
113
+ * - `gpt-6.1-sol` rejects `reasoningEffort: "minimal"` (and `none`) with HTTP
114
+ * 400; `low` is the floor. See the `ReasoningEffort` note above.
115
+ * - `meta/muse-spark-1.3` is age-gated on OpenRouter and returns HTTP 403 until
116
+ * the account completes the 18+ confirmation at
117
+ * openrouter.ai/settings/preferences. It also reasons by default (~370-814
118
+ * reasoning tokens per call) even with no effort configured, so cost per call
119
+ * runs above what its headline rate suggests.
120
+ */
88
121
  declare const DEFAULT_MODELS: {
89
- readonly anthropic: "claude-sonnet-4-6";
90
- readonly openai: "gpt-5.6-luna";
91
- readonly google: "gemini-3-flash-preview";
92
- readonly openrouter: "qwen/qwen3.6-flash";
122
+ readonly anthropic: "claude-sonnet-5-5";
123
+ readonly openai: "gpt-6.1-sol";
124
+ readonly google: "gemini-3.8-flash";
125
+ readonly openrouter: "meta/muse-spark-1.3";
93
126
  };
94
127
  /** Built-in content checks available through `client.content()`. */
95
128
  declare const Content: {
@@ -178,13 +211,13 @@ declare const StatementResultSchema: z.ZodObject<{
178
211
  statement: string;
179
212
  pass: boolean;
180
213
  reasoning: string;
181
- confidence?: "high" | "medium" | "low" | undefined;
214
+ confidence?: "low" | "medium" | "high" | undefined;
182
215
  timestampSeconds?: number | null | undefined;
183
216
  }, {
184
217
  statement: string;
185
218
  pass: boolean;
186
219
  reasoning: string;
187
- confidence?: "high" | "medium" | "low" | undefined;
220
+ confidence?: "low" | "medium" | "high" | undefined;
188
221
  timestampSeconds?: number | null | undefined;
189
222
  }>;
190
223
  /** Outcome of a single statement evaluated by `check()`. */
@@ -303,13 +336,13 @@ declare const CheckResultSchema: z.ZodObject<{
303
336
  statement: string;
304
337
  pass: boolean;
305
338
  reasoning: string;
306
- confidence?: "high" | "medium" | "low" | undefined;
339
+ confidence?: "low" | "medium" | "high" | undefined;
307
340
  timestampSeconds?: number | null | undefined;
308
341
  }, {
309
342
  statement: string;
310
343
  pass: boolean;
311
344
  reasoning: string;
312
- confidence?: "high" | "medium" | "low" | undefined;
345
+ confidence?: "low" | "medium" | "high" | undefined;
313
346
  timestampSeconds?: number | null | undefined;
314
347
  }>, "many">;
315
348
  }, "strip", z.ZodTypeAny, {
@@ -325,7 +358,7 @@ declare const CheckResultSchema: z.ZodObject<{
325
358
  statement: string;
326
359
  pass: boolean;
327
360
  reasoning: string;
328
- confidence?: "high" | "medium" | "low" | undefined;
361
+ confidence?: "low" | "medium" | "high" | undefined;
329
362
  timestampSeconds?: number | null | undefined;
330
363
  }[];
331
364
  usage?: {
@@ -350,7 +383,7 @@ declare const CheckResultSchema: z.ZodObject<{
350
383
  statement: string;
351
384
  pass: boolean;
352
385
  reasoning: string;
353
- confidence?: "high" | "medium" | "low" | undefined;
386
+ confidence?: "low" | "medium" | "high" | undefined;
354
387
  timestampSeconds?: number | null | undefined;
355
388
  }[];
356
389
  usage?: {
@@ -968,8 +1001,9 @@ interface VisualAIClient {
968
1001
  * @param imageA Baseline image source.
969
1002
  * @param imageB Candidate image source.
970
1003
  * @param options Optional comparison prompt, instructions, and diff-image settings.
971
- * `gemini-3-flash-preview` generates an annotated diff image by default;
972
- * pass `{ diffImage: false }` to opt out.
1004
+ * Gemini flash models (`DIFF_ALLOWED_MODELS`) generate an annotated diff image
1005
+ * by default; pass `{ diffImage: false }` to opt out. Flash-Lite and Pro tiers
1006
+ * produce none even when asked.
973
1007
  * @returns A structured comparison result with optional diff image metadata.
974
1008
  * @throws {VisualAIImageError} When either image cannot be loaded or decoded.
975
1009
  * @throws {VisualAIError} When the provider rejects the request or returns invalid output.
package/dist/index.d.ts CHANGED
@@ -1,7 +1,21 @@
1
1
  import { z } from 'zod';
2
2
 
3
3
  /** Supported reasoning effort levels. */
4
+ /**
5
+ * Abstract reasoning-effort hint. Each provider driver maps it to its native
6
+ * mechanism; levels a provider lacks clamp to its nearest tier.
7
+ *
8
+ * `minimal` is **not universally accepted**. OpenAI rejects it per-model with
9
+ * HTTP 400 ("Unsupported value: 'minimal' is not supported with the
10
+ * '<model>' model"), observed on `gpt-6-astra` and `gpt-6.1-sol`, whose
11
+ * supported set is low/medium/high/xhigh/max, and on `gpt-6-sol` and
12
+ * `gpt-6-luna`, which accept none/low/medium/high/xhigh/max. Gemini defines
13
+ * the tier but some models (e.g. Gemini 3.1 Pro) reject it, so Google and
14
+ * OpenRouter clamp it to `low` rather than send it. Use `minimal` only
15
+ * against an OpenAI model known to accept it; `low` is the portable floor.
16
+ */
4
17
  declare const ReasoningEffort: {
18
+ readonly MINIMAL: "minimal";
5
19
  readonly LOW: "low";
6
20
  readonly MEDIUM: "medium";
7
21
  readonly HIGH: "high";
@@ -34,16 +48,22 @@ declare const Model: {
34
48
  readonly Anthropic: {
35
49
  readonly FABLE_5_1: "claude-fable-5-1";
36
50
  readonly FABLE_5: "claude-fable-5";
51
+ readonly OPUS_5_5: "claude-opus-5-5";
37
52
  readonly OPUS_5: "claude-opus-5";
38
53
  readonly OPUS_4_8: "claude-opus-4-8";
39
54
  readonly OPUS_4_7: "claude-opus-4-7";
40
55
  readonly OPUS_4_6: "claude-opus-4-6";
56
+ readonly SONNET_5_5: "claude-sonnet-5-5";
41
57
  readonly SONNET_5: "claude-sonnet-5";
42
58
  readonly SONNET_4_6: "claude-sonnet-4-6";
59
+ readonly HAIKU_5_5: "claude-haiku-5-5";
43
60
  readonly HAIKU_4_5: "claude-haiku-4-5";
44
61
  };
45
62
  readonly OpenAI: {
46
63
  readonly GPT_6_ASTRA: "gpt-6-astra";
64
+ readonly GPT_6_1_SOL: "gpt-6.1-sol";
65
+ readonly GPT_6_SOL: "gpt-6-sol";
66
+ readonly GPT_6_LUNA: "gpt-6-luna";
47
67
  readonly GPT_5_6_SOL: "gpt-5.6-sol";
48
68
  readonly GPT_5_6_TERRA: "gpt-5.6-terra";
49
69
  readonly GPT_5_6_LUNA: "gpt-5.6-luna";
@@ -72,6 +92,7 @@ declare const Model: {
72
92
  */
73
93
  readonly OpenRouter: {
74
94
  readonly MUSE_SPARK_1_3: "meta/muse-spark-1.3";
95
+ readonly GROK_4_7: "x-ai/grok-4.7";
75
96
  readonly GROK_4_6: "x-ai/grok-4.6";
76
97
  readonly GROK_4_5: "x-ai/grok-4.5";
77
98
  readonly KIMI_K3: "moonshotai/kimi-k3";
@@ -80,16 +101,28 @@ declare const Model: {
80
101
  readonly QWEN_3_7_PLUS: "qwen/qwen3.7-plus";
81
102
  readonly QWEN_3_6_FLASH: "qwen/qwen3.6-flash";
82
103
  readonly GLM_5_3_FLASH: "z-ai/glm-5.3-flash";
104
+ readonly MIMO_V2_6_PRO: "xiaomi/mimo-v2.6-pro";
83
105
  };
84
106
  };
85
107
  /** Union of all built-in model name literals exposed by `Model`. */
86
108
  type KnownModelName = (typeof Model.Anthropic)[keyof typeof Model.Anthropic] | (typeof Model.OpenAI)[keyof typeof Model.OpenAI] | (typeof Model.Google)[keyof typeof Model.Google] | (typeof Model.OpenRouter)[keyof typeof Model.OpenRouter];
87
- /** Default model selection used when a caller omits `config.model`. */
109
+ /**
110
+ * Default model selection used when a caller omits `config.model`.
111
+ *
112
+ * Two of these carry constraints the older defaults did not:
113
+ * - `gpt-6.1-sol` rejects `reasoningEffort: "minimal"` (and `none`) with HTTP
114
+ * 400; `low` is the floor. See the `ReasoningEffort` note above.
115
+ * - `meta/muse-spark-1.3` is age-gated on OpenRouter and returns HTTP 403 until
116
+ * the account completes the 18+ confirmation at
117
+ * openrouter.ai/settings/preferences. It also reasons by default (~370-814
118
+ * reasoning tokens per call) even with no effort configured, so cost per call
119
+ * runs above what its headline rate suggests.
120
+ */
88
121
  declare const DEFAULT_MODELS: {
89
- readonly anthropic: "claude-sonnet-4-6";
90
- readonly openai: "gpt-5.6-luna";
91
- readonly google: "gemini-3-flash-preview";
92
- readonly openrouter: "qwen/qwen3.6-flash";
122
+ readonly anthropic: "claude-sonnet-5-5";
123
+ readonly openai: "gpt-6.1-sol";
124
+ readonly google: "gemini-3.8-flash";
125
+ readonly openrouter: "meta/muse-spark-1.3";
93
126
  };
94
127
  /** Built-in content checks available through `client.content()`. */
95
128
  declare const Content: {
@@ -178,13 +211,13 @@ declare const StatementResultSchema: z.ZodObject<{
178
211
  statement: string;
179
212
  pass: boolean;
180
213
  reasoning: string;
181
- confidence?: "high" | "medium" | "low" | undefined;
214
+ confidence?: "low" | "medium" | "high" | undefined;
182
215
  timestampSeconds?: number | null | undefined;
183
216
  }, {
184
217
  statement: string;
185
218
  pass: boolean;
186
219
  reasoning: string;
187
- confidence?: "high" | "medium" | "low" | undefined;
220
+ confidence?: "low" | "medium" | "high" | undefined;
188
221
  timestampSeconds?: number | null | undefined;
189
222
  }>;
190
223
  /** Outcome of a single statement evaluated by `check()`. */
@@ -303,13 +336,13 @@ declare const CheckResultSchema: z.ZodObject<{
303
336
  statement: string;
304
337
  pass: boolean;
305
338
  reasoning: string;
306
- confidence?: "high" | "medium" | "low" | undefined;
339
+ confidence?: "low" | "medium" | "high" | undefined;
307
340
  timestampSeconds?: number | null | undefined;
308
341
  }, {
309
342
  statement: string;
310
343
  pass: boolean;
311
344
  reasoning: string;
312
- confidence?: "high" | "medium" | "low" | undefined;
345
+ confidence?: "low" | "medium" | "high" | undefined;
313
346
  timestampSeconds?: number | null | undefined;
314
347
  }>, "many">;
315
348
  }, "strip", z.ZodTypeAny, {
@@ -325,7 +358,7 @@ declare const CheckResultSchema: z.ZodObject<{
325
358
  statement: string;
326
359
  pass: boolean;
327
360
  reasoning: string;
328
- confidence?: "high" | "medium" | "low" | undefined;
361
+ confidence?: "low" | "medium" | "high" | undefined;
329
362
  timestampSeconds?: number | null | undefined;
330
363
  }[];
331
364
  usage?: {
@@ -350,7 +383,7 @@ declare const CheckResultSchema: z.ZodObject<{
350
383
  statement: string;
351
384
  pass: boolean;
352
385
  reasoning: string;
353
- confidence?: "high" | "medium" | "low" | undefined;
386
+ confidence?: "low" | "medium" | "high" | undefined;
354
387
  timestampSeconds?: number | null | undefined;
355
388
  }[];
356
389
  usage?: {
@@ -968,8 +1001,9 @@ interface VisualAIClient {
968
1001
  * @param imageA Baseline image source.
969
1002
  * @param imageB Candidate image source.
970
1003
  * @param options Optional comparison prompt, instructions, and diff-image settings.
971
- * `gemini-3-flash-preview` generates an annotated diff image by default;
972
- * pass `{ diffImage: false }` to opt out.
1004
+ * Gemini flash models (`DIFF_ALLOWED_MODELS`) generate an annotated diff image
1005
+ * by default; pass `{ diffImage: false }` to opt out. Flash-Lite and Pro tiers
1006
+ * produce none even when asked.
973
1007
  * @returns A structured comparison result with optional diff image metadata.
974
1008
  * @throws {VisualAIImageError} When either image cannot be loaded or decoded.
975
1009
  * @throws {VisualAIError} When the provider rejects the request or returns invalid output.