@akasecurity/ai-tc-codex 0.9.17 → 0.9.18

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.
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "aka-codex",
3
- "version": "0.9.17",
3
+ "version": "0.9.18",
4
4
  "description": "AI Traffic Control \u2014 inspect and govern AI prompts in Codex CLI. Detection runs locally; events are recorded to a local SQLite store on your machine.",
5
5
  "hooks": "./hooks/hooks.json",
6
6
  "skills": "./skills",
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@akasecurity/ai-tc-codex",
3
- "version": "0.9.17",
3
+ "version": "0.9.18",
4
4
  "description": "AI Traffic Control — inspect and govern AI prompts in Codex CLI",
5
5
  "license": "Apache-2.0",
6
6
  "type": "module",
@@ -28,12 +28,12 @@
28
28
  "typescript": "^5.8.0",
29
29
  "vitest": "^4.1.11",
30
30
  "@akasecurity/eslint-config": "0.8.0",
31
- "@akasecurity/plugin-runtime": "0.8.0",
32
- "@akasecurity/scanner": "0.8.0",
33
- "@akasecurity/persistence": "0.8.0",
34
31
  "@akasecurity/plugin-sdk": "0.8.0",
35
- "@akasecurity/setup-wizard": "0.8.0",
36
- "@akasecurity/schema": "0.8.0"
32
+ "@akasecurity/persistence": "0.8.0",
33
+ "@akasecurity/schema": "0.8.0",
34
+ "@akasecurity/scanner": "0.8.0",
35
+ "@akasecurity/plugin-runtime": "0.8.0",
36
+ "@akasecurity/setup-wizard": "0.8.0"
37
37
  },
38
38
  "scripts": {
39
39
  "build": "tsup",
@@ -19858,7 +19858,10 @@ function date4(params) {
19858
19858
  var LongContextTier = external_exports.object({
19859
19859
  thresholdInputTokens: external_exports.number().int().positive(),
19860
19860
  input: external_exports.number().nonnegative().nullable(),
19861
- output: external_exports.number().nonnegative().nullable()
19861
+ output: external_exports.number().nonnegative().nullable(),
19862
+ cacheRead: external_exports.number().nonnegative().nullable().optional(),
19863
+ cacheWrite5m: external_exports.number().nonnegative().nullable().optional(),
19864
+ cacheWrite1h: external_exports.number().nonnegative().nullable().optional()
19862
19865
  });
19863
19866
  var ModelPrice = external_exports.object({
19864
19867
  /** USD per 1M uncached input tokens, at or below any long-context threshold. */
@@ -19867,7 +19870,8 @@ var ModelPrice = external_exports.object({
19867
19870
  output: external_exports.number().nonnegative(),
19868
19871
  /**
19869
19872
  * Rate band above a size threshold, or `null` when the model is flat-rated.
19870
- * Anthropic is flat across its 1M window; OpenAI, Google and xAI are not.
19873
+ * OpenAI, Google, xAI and some Anthropic models (Haiku 5.5) band; most
19874
+ * Claude models are flat across their window.
19871
19875
  */
19872
19876
  longContext: LongContextTier.nullable(),
19873
19877
  /** USD per 1M tokens written to a 5-minute ephemeral cache, or null if unpriced. */
@@ -19880,18 +19884,35 @@ var ModelPrice = external_exports.object({
19880
19884
  webSearch: external_exports.number().nonnegative().nullable()
19881
19885
  });
19882
19886
  var WEB_SEARCH_PER_REQUEST = 0.01;
19883
- function anthropicPrice(input2, output2, cacheRead) {
19887
+ function anthropicPrice(input2, output2, cacheRead, longContext = null) {
19884
19888
  return Object.freeze({
19885
19889
  input: input2,
19886
19890
  output: output2,
19887
- longContext: null,
19888
- // flat across the full window on every current Claude
19891
+ longContext: longContext === null ? null : Object.freeze(longContext),
19889
19892
  cacheWrite5m: input2 * 1.25,
19890
19893
  cacheWrite1h: input2 * 2,
19891
19894
  cacheRead,
19892
19895
  webSearch: WEB_SEARCH_PER_REQUEST
19893
19896
  });
19894
19897
  }
19898
+ function anthropicBand(thresholdInputTokens, input2, output2, cacheRead) {
19899
+ return {
19900
+ thresholdInputTokens,
19901
+ input: input2,
19902
+ output: output2,
19903
+ cacheRead,
19904
+ cacheWrite5m: input2 * 1.25,
19905
+ cacheWrite1h: input2 * 2
19906
+ };
19907
+ }
19908
+ function scaledBand(thresholdInputTokens, base, scale) {
19909
+ return {
19910
+ thresholdInputTokens,
19911
+ input: base.input * scale.input,
19912
+ output: base.output * scale.output,
19913
+ ...base.cacheRead === void 0 ? {} : { cacheRead: base.cacheRead * scale.cacheRead }
19914
+ };
19915
+ }
19895
19916
  function tokenPrice(input2, output2, options = {}) {
19896
19917
  return Object.freeze({
19897
19918
  input: input2,
@@ -19924,17 +19945,19 @@ function serviceTierMultiplier(tier) {
19924
19945
  function promptTokensOf(usage) {
19925
19946
  return usage.promptTokens ?? (usage.inputTokens ?? 0) + (usage.cacheReadTokens ?? 0) + (usage.cacheWrite1hTokens ?? 0) + (usage.cacheWrite5mTokens ?? 0);
19926
19947
  }
19948
+ function bandRate(column, base) {
19949
+ return column === void 0 ? base : column;
19950
+ }
19927
19951
  function costOf(price, usage) {
19952
+ const band = price.longContext !== null && promptTokensOf(usage) > price.longContext.thresholdInputTokens ? price.longContext : null;
19928
19953
  const metered = [
19929
- [usage.cacheWrite1hTokens ?? 0, price.cacheWrite1h],
19930
- [usage.cacheWrite5mTokens ?? 0, price.cacheWrite5m],
19931
- [usage.cacheReadTokens ?? 0, price.cacheRead]
19954
+ [usage.inputTokens ?? 0, bandRate(band?.input, price.input)],
19955
+ [usage.outputTokens ?? 0, bandRate(band?.output, price.output)],
19956
+ [usage.cacheWrite1hTokens ?? 0, bandRate(band?.cacheWrite1h, price.cacheWrite1h)],
19957
+ [usage.cacheWrite5mTokens ?? 0, bandRate(band?.cacheWrite5m, price.cacheWrite5m)],
19958
+ [usage.cacheReadTokens ?? 0, bandRate(band?.cacheRead, price.cacheRead)]
19932
19959
  ];
19933
- const band = price.longContext !== null && promptTokensOf(usage) > price.longContext.thresholdInputTokens ? price.longContext : null;
19934
- if (band !== null && (band.input === null || band.output === null)) return null;
19935
- const inputRate = band?.input ?? price.input;
19936
- const outputRate = band?.output ?? price.output;
19937
- let tokenCost = (usage.inputTokens ?? 0) * inputRate + (usage.outputTokens ?? 0) * outputRate;
19960
+ let tokenCost = 0;
19938
19961
  for (const [tokens, rate] of metered) {
19939
19962
  if (tokens === 0) continue;
19940
19963
  if (rate === null) return null;
@@ -20224,10 +20247,20 @@ anthropic.model("claude-fable-5-1", {
20224
20247
  });
20225
20248
  anthropic.model("claude-fable-5", {
20226
20249
  name: "Claude Fable 5",
20250
+ ctx: 1e6,
20251
+ maxOut: 128e3,
20227
20252
  capability: "reasoning",
20228
20253
  price: anthropicPrice(10, 50, 1),
20229
20254
  on: [...CLAUDE_RESELLERS]
20230
20255
  });
20256
+ anthropic.model("claude-opus-5-5", {
20257
+ name: "Claude Opus 5.5",
20258
+ ctx: 1e6,
20259
+ maxOut: 128e3,
20260
+ capability: "reasoning",
20261
+ price: anthropicPrice(4, 20, 0.2),
20262
+ on: [...CLAUDE_RESELLERS]
20263
+ });
20231
20264
  anthropic.model("claude-opus-5", {
20232
20265
  name: "Claude Opus 5",
20233
20266
  ctx: 1e6,
@@ -20238,24 +20271,32 @@ anthropic.model("claude-opus-5", {
20238
20271
  });
20239
20272
  anthropic.model("claude-opus-4-8", {
20240
20273
  name: "Claude Opus 4.8",
20274
+ ctx: 1e6,
20275
+ maxOut: 128e3,
20241
20276
  capability: "reasoning",
20242
20277
  price: anthropicPrice(5, 25, 0.5),
20243
20278
  on: [...CLAUDE_RESELLERS]
20244
20279
  });
20245
20280
  anthropic.model("claude-opus-4-7", {
20246
20281
  name: "Claude Opus 4.7",
20282
+ ctx: 1e6,
20283
+ maxOut: 128e3,
20247
20284
  capability: "reasoning",
20248
20285
  price: anthropicPrice(5, 25, 0.5),
20249
20286
  on: [...CLAUDE_RESELLERS]
20250
20287
  });
20251
20288
  anthropic.model("claude-opus-4-6", {
20252
20289
  name: "Claude Opus 4.6",
20290
+ ctx: 1e6,
20291
+ maxOut: 128e3,
20253
20292
  capability: "reasoning",
20254
20293
  price: anthropicPrice(5, 25, 0.5),
20255
20294
  on: [...CLAUDE_RESELLERS]
20256
20295
  });
20257
20296
  anthropic.model("claude-opus-4-5", {
20258
20297
  name: "Claude Opus 4.5",
20298
+ ctx: 2e5,
20299
+ maxOut: 64e3,
20259
20300
  capability: "reasoning",
20260
20301
  price: anthropicPrice(5, 25, 0.5),
20261
20302
  on: [...CLAUDE_RESELLERS]
@@ -20266,6 +20307,14 @@ anthropic.model("claude-opus-4-1", {
20266
20307
  price: anthropicPrice(15, 75, 1.5),
20267
20308
  on: [...CLAUDE_RESELLERS]
20268
20309
  });
20310
+ anthropic.model("claude-sonnet-5-5", {
20311
+ name: "Claude Sonnet 5.5",
20312
+ ctx: 1e6,
20313
+ maxOut: 128e3,
20314
+ capability: "reasoning",
20315
+ price: anthropicPrice(2, 10, 0.1),
20316
+ on: [...CLAUDE_RESELLERS]
20317
+ });
20269
20318
  anthropic.model("claude-sonnet-5", {
20270
20319
  name: "Claude Sonnet 5",
20271
20320
  ctx: 1e6,
@@ -20276,16 +20325,30 @@ anthropic.model("claude-sonnet-5", {
20276
20325
  });
20277
20326
  anthropic.model("claude-sonnet-4-6", {
20278
20327
  name: "Claude Sonnet 4.6",
20328
+ ctx: 1e6,
20329
+ maxOut: 128e3,
20279
20330
  capability: "reasoning",
20280
20331
  price: anthropicPrice(3, 15, 0.3),
20281
20332
  on: [...CLAUDE_RESELLERS]
20282
20333
  });
20283
20334
  anthropic.model("claude-sonnet-4-5", {
20284
20335
  name: "Claude Sonnet 4.5",
20336
+ ctx: 2e5,
20337
+ maxOut: 64e3,
20285
20338
  capability: "reasoning",
20286
20339
  price: anthropicPrice(3, 15, 0.3),
20287
20340
  on: [...CLAUDE_RESELLERS]
20288
20341
  });
20342
+ anthropic.model("claude-haiku-5-5", {
20343
+ name: "Claude Haiku 5.5",
20344
+ ctx: 1e6,
20345
+ maxOut: 128e3,
20346
+ capability: null,
20347
+ // A prompt over 100,000 tokens, cache reads and writes included, bills the
20348
+ // whole request at the higher band.
20349
+ price: anthropicPrice(0.1, 0.5, 0.01, anthropicBand(1e5, 0.5, 2.5, 0.05)),
20350
+ on: [...CLAUDE_RESELLERS]
20351
+ });
20289
20352
  anthropic.model("claude-haiku-4-5", {
20290
20353
  name: "Claude Haiku 4.5",
20291
20354
  ctx: 2e5,
@@ -20308,7 +20371,13 @@ var OPENAI_TRAINING = {
20308
20371
  consumerControl: "opt-out",
20309
20372
  source: "https://developers.openai.com/api/docs/guides/your-data"
20310
20373
  };
20311
- var LONG_CONTEXT_272K = { thresholdInputTokens: 272e3, input: null, output: null };
20374
+ function gptPrice(input2, output2, cacheRead) {
20375
+ const base = { input: input2, output: output2, ...cacheRead === void 0 ? {} : { cacheRead } };
20376
+ return tokenPrice(input2, output2, {
20377
+ ...base,
20378
+ longContext: scaledBand(272e3, base, { input: 2, output: 1.5, cacheRead: 2 })
20379
+ });
20380
+ }
20312
20381
  var openai = vendor("openai", {
20313
20382
  family: { id: "gpt", name: "GPT" },
20314
20383
  region: "us",
@@ -20316,12 +20385,45 @@ var openai = vendor("openai", {
20316
20385
  training: OPENAI_TRAINING
20317
20386
  });
20318
20387
  var OPENAI_RESELLERS = ["azure", "foundry"];
20388
+ openai.model("gpt-6-astra", {
20389
+ name: "GPT-6 Astra",
20390
+ ctx: 105e4,
20391
+ maxOut: 128e3,
20392
+ capability: "reasoning",
20393
+ price: gptPrice(10, 50, 1),
20394
+ on: [...OPENAI_RESELLERS]
20395
+ });
20396
+ openai.model("gpt-6.1-sol", {
20397
+ name: "GPT-6.1 Sol",
20398
+ ctx: 105e4,
20399
+ maxOut: 128e3,
20400
+ capability: "reasoning",
20401
+ price: gptPrice(2, 10, 0.1),
20402
+ on: [...OPENAI_RESELLERS]
20403
+ });
20404
+ openai.model("gpt-6-sol", {
20405
+ name: "GPT-6 Sol",
20406
+ ctx: 105e4,
20407
+ maxOut: 128e3,
20408
+ capability: "reasoning",
20409
+ price: gptPrice(2, 10, 0.2),
20410
+ on: [...OPENAI_RESELLERS]
20411
+ });
20412
+ openai.model("gpt-6-luna", {
20413
+ name: "GPT-6 Luna",
20414
+ ctx: 105e4,
20415
+ maxOut: 128e3,
20416
+ capability: null,
20417
+ price: gptPrice(0.1, 0.5, 0.01),
20418
+ on: [...OPENAI_RESELLERS]
20419
+ });
20319
20420
  openai.model("gpt-5.6-sol", {
20320
20421
  name: "GPT-5.6 Sol",
20422
+ // Promotional rate, published as available at least through 2026-11-21.
20321
20423
  ctx: 105e4,
20322
20424
  maxOut: 128e3,
20323
20425
  capability: "reasoning",
20324
- price: tokenPrice(4, 20, { cacheRead: 0.4, longContext: LONG_CONTEXT_272K }),
20426
+ price: gptPrice(4, 20, 0.4),
20325
20427
  on: [...OPENAI_RESELLERS]
20326
20428
  });
20327
20429
  openai.model("gpt-5.6-terra", {
@@ -20329,7 +20431,7 @@ openai.model("gpt-5.6-terra", {
20329
20431
  ctx: 105e4,
20330
20432
  maxOut: 128e3,
20331
20433
  capability: "reasoning",
20332
- price: tokenPrice(2, 12, { cacheRead: 0.2, longContext: LONG_CONTEXT_272K }),
20434
+ price: gptPrice(2, 12, 0.2),
20333
20435
  on: [...OPENAI_RESELLERS]
20334
20436
  });
20335
20437
  openai.model("gpt-5.6-luna", {
@@ -20337,7 +20439,7 @@ openai.model("gpt-5.6-luna", {
20337
20439
  ctx: 105e4,
20338
20440
  maxOut: 128e3,
20339
20441
  capability: null,
20340
- price: tokenPrice(0.2, 1.2, { cacheRead: 0.02, longContext: LONG_CONTEXT_272K }),
20442
+ price: gptPrice(0.2, 1.2, 0.02),
20341
20443
  on: [...OPENAI_RESELLERS]
20342
20444
  });
20343
20445
  openai.model("gpt-5.5", {
@@ -20345,7 +20447,15 @@ openai.model("gpt-5.5", {
20345
20447
  ctx: 105e4,
20346
20448
  maxOut: 128e3,
20347
20449
  capability: "reasoning",
20348
- price: tokenPrice(5, 30, { cacheRead: 0.5, longContext: LONG_CONTEXT_272K }),
20450
+ price: gptPrice(5, 30, 0.5),
20451
+ on: [...OPENAI_RESELLERS]
20452
+ });
20453
+ openai.model("gpt-5.5-pro", {
20454
+ name: "GPT-5.5 Pro",
20455
+ ctx: 105e4,
20456
+ maxOut: 128e3,
20457
+ capability: "reasoning",
20458
+ price: gptPrice(30, 180),
20349
20459
  on: [...OPENAI_RESELLERS]
20350
20460
  });
20351
20461
  openai.model("gpt-5.4", {
@@ -20353,7 +20463,7 @@ openai.model("gpt-5.4", {
20353
20463
  ctx: 105e4,
20354
20464
  maxOut: 128e3,
20355
20465
  capability: "reasoning",
20356
- price: tokenPrice(2.5, 15, { cacheRead: 0.25, longContext: LONG_CONTEXT_272K }),
20466
+ price: gptPrice(2.5, 15, 0.25),
20357
20467
  on: [...OPENAI_RESELLERS]
20358
20468
  });
20359
20469
  openai.model("gpt-5.4-mini", {
@@ -20399,6 +20509,7 @@ openai.model("gpt-5", {
20399
20509
  openai.model("gpt-5-mini", {
20400
20510
  name: "GPT-5 mini",
20401
20511
  ctx: 4e5,
20512
+ maxOut: 128e3,
20402
20513
  capability: null,
20403
20514
  price: tokenPrice(0.25, 2, { cacheRead: 0.025 }),
20404
20515
  on: [...OPENAI_RESELLERS]
@@ -20413,11 +20524,15 @@ openai.model("gpt-5-nano", {
20413
20524
  });
20414
20525
  openai.model("gpt-4.1", {
20415
20526
  name: "GPT-4.1",
20527
+ ctx: 1047576,
20528
+ maxOut: 32768,
20416
20529
  price: tokenPrice(2, 8, { cacheRead: 0.5 }),
20417
20530
  on: [...OPENAI_RESELLERS]
20418
20531
  });
20419
20532
  openai.model("gpt-4o", {
20420
20533
  name: "GPT-4o",
20534
+ ctx: 128e3,
20535
+ maxOut: 16384,
20421
20536
  price: tokenPrice(2.5, 10, { cacheRead: 1.25 }),
20422
20537
  on: [...OPENAI_RESELLERS]
20423
20538
  });
@@ -20441,6 +20556,8 @@ openai.model("gpt-4-turbo", {
20441
20556
  });
20442
20557
  openai.model("o3", {
20443
20558
  name: "o3",
20559
+ ctx: 2e5,
20560
+ maxOut: 1e5,
20444
20561
  capability: "reasoning",
20445
20562
  price: tokenPrice(2, 8, { cacheRead: 0.5 }),
20446
20563
  on: [...OPENAI_RESELLERS]
@@ -20493,57 +20610,100 @@ var google = vendor("google", {
20493
20610
  retention: "configurable",
20494
20611
  training: GOOGLE_TRAINING
20495
20612
  });
20613
+ google.model("gemini-3.8-flash", {
20614
+ name: "Gemini 3.8 Flash",
20615
+ ctx: 1048576,
20616
+ maxOut: 65536,
20617
+ capability: null,
20618
+ price: tokenPrice(0.75, 3.75, { cacheRead: 0.075 }),
20619
+ on: ["vertex"]
20620
+ });
20496
20621
  google.model("gemini-3.7-flash", {
20497
20622
  name: "Gemini 3.7 Flash",
20498
20623
  capability: null,
20499
- // Promotional rate; the published post-promotion rate is 1.50 / 7.50.
20500
20624
  price: tokenPrice(0.75, 3.75),
20501
20625
  on: ["vertex"]
20502
20626
  });
20503
20627
  google.model("gemini-3.6-flash", {
20504
20628
  name: "Gemini 3.6 Flash",
20629
+ ctx: 1048576,
20630
+ maxOut: 65536,
20505
20631
  capability: null,
20506
- price: tokenPrice(0.75, 3.75),
20632
+ price: tokenPrice(0.75, 3.75, { cacheRead: 0.075 }),
20507
20633
  on: ["vertex"]
20508
20634
  });
20509
20635
  google.model("gemini-3.5-flash", {
20510
20636
  name: "Gemini 3.5 Flash",
20511
20637
  capability: null,
20512
- price: tokenPrice(1.5, 9),
20513
20638
  on: ["vertex"]
20514
20639
  });
20515
20640
  google.model("gemini-3.5-flash-lite", {
20516
20641
  name: "Gemini 3.5 Flash Lite",
20642
+ ctx: 1048576,
20643
+ maxOut: 65536,
20644
+ capability: null,
20645
+ price: tokenPrice(0.3, 2.5, { cacheRead: 0.03 }),
20646
+ on: ["vertex"]
20647
+ });
20648
+ google.model("gemini-3.1-flash-lite", {
20649
+ name: "Gemini 3.1 Flash Lite",
20650
+ ctx: 1048576,
20651
+ maxOut: 65536,
20652
+ capability: null,
20653
+ price: tokenPrice(0.25, 1.5, { cacheRead: 0.025 }),
20654
+ on: ["vertex"]
20655
+ });
20656
+ google.model("gemini-3-flash-preview", {
20657
+ name: "Gemini 3 Flash Preview",
20658
+ ctx: 1048576,
20659
+ maxOut: 65536,
20517
20660
  capability: null,
20518
- price: tokenPrice(0.3, 2.5),
20661
+ price: tokenPrice(0.5, 3, { cacheRead: 0.05 }),
20519
20662
  on: ["vertex"]
20520
20663
  });
20521
20664
  google.model("gemini-3.1-pro-preview", {
20522
20665
  name: "Gemini 3.1 Pro Preview",
20666
+ ctx: 1048576,
20667
+ maxOut: 65536,
20523
20668
  capability: "reasoning",
20524
20669
  price: tokenPrice(2, 12, {
20525
- longContext: { thresholdInputTokens: 2e5, input: 4, output: 18 }
20670
+ cacheRead: 0.2,
20671
+ longContext: { thresholdInputTokens: 2e5, input: 4, output: 18, cacheRead: 0.4 }
20526
20672
  }),
20673
+ aliases: ["gemini-3.1-pro-preview-customtools"],
20527
20674
  on: ["vertex"]
20528
20675
  });
20529
20676
  google.model("gemini-2.5-pro", {
20530
20677
  name: "Gemini 2.5 Pro",
20678
+ ctx: 1048576,
20679
+ maxOut: 65536,
20531
20680
  capability: "reasoning",
20532
20681
  price: tokenPrice(1.25, 10, {
20533
- longContext: { thresholdInputTokens: 2e5, input: 2.5, output: 15 }
20682
+ cacheRead: 0.125,
20683
+ longContext: { thresholdInputTokens: 2e5, input: 2.5, output: 15, cacheRead: 0.25 }
20534
20684
  }),
20535
20685
  on: ["vertex"]
20536
20686
  });
20537
20687
  google.model("gemini-2.5-flash", {
20538
20688
  name: "Gemini 2.5 Flash",
20689
+ ctx: 1048576,
20690
+ maxOut: 65536,
20539
20691
  capability: null,
20540
- price: tokenPrice(0.3, 2.5),
20692
+ price: tokenPrice(0.3, 2.5, { cacheRead: 0.03 }),
20541
20693
  on: ["vertex"]
20542
20694
  });
20543
20695
  google.model("gemini-2.5-flash-lite", {
20544
20696
  name: "Gemini 2.5 Flash Lite",
20697
+ ctx: 1048576,
20698
+ maxOut: 65536,
20545
20699
  capability: null,
20546
- price: tokenPrice(0.1, 0.4),
20700
+ price: tokenPrice(0.1, 0.4, { cacheRead: 0.01 }),
20701
+ on: ["vertex"]
20702
+ });
20703
+ google.model("gemini-embedding-2", {
20704
+ name: "Gemini Embedding 2",
20705
+ capability: "embedding",
20706
+ price: tokenPrice(0.2, 0),
20547
20707
  on: ["vertex"]
20548
20708
  });
20549
20709
  google.model("gemini-1-5-pro", { name: "Gemini 1.5 Pro", on: ["vertex"] });
@@ -20559,43 +20719,62 @@ var xai = vendor("xai", {
20559
20719
  source: "https://docs.x.ai/developers/faq/security"
20560
20720
  }
20561
20721
  });
20722
+ function grokPrice(input2, output2, cacheRead) {
20723
+ const base = { input: input2, output: output2, cacheRead };
20724
+ return tokenPrice(input2, output2, {
20725
+ cacheRead,
20726
+ longContext: scaledBand(2e5, base, { input: 2, output: 2, cacheRead: 2 })
20727
+ });
20728
+ }
20729
+ xai.model("grok-4.7", {
20730
+ name: "Grok 4.7",
20731
+ ctx: 5e5,
20732
+ capability: "reasoning",
20733
+ price: grokPrice(2, 6, 0.5)
20734
+ });
20562
20735
  xai.model("grok-4.6", {
20563
20736
  name: "Grok 4.6",
20564
20737
  ctx: 5e5,
20565
20738
  capability: "reasoning",
20566
- price: tokenPrice(2, 6, {
20567
- cacheRead: 0.5,
20568
- longContext: { thresholdInputTokens: 2e5, input: 4, output: 12 }
20569
- }),
20739
+ price: grokPrice(2, 6, 0.5),
20570
20740
  on: ["bedrock"]
20571
20741
  });
20572
20742
  xai.model("grok-4.5", {
20573
20743
  name: "Grok 4.5",
20574
20744
  ctx: 5e5,
20575
20745
  capability: "reasoning",
20576
- price: tokenPrice(2, 6, {
20577
- cacheRead: 0.3,
20578
- longContext: { thresholdInputTokens: 2e5, input: 4, output: 12 }
20579
- })
20746
+ price: grokPrice(2, 6, 0.3)
20580
20747
  });
20581
20748
  xai.model("grok-4.3", {
20582
20749
  name: "Grok 4.3",
20583
20750
  ctx: 1e6,
20584
20751
  capability: "reasoning",
20585
- price: tokenPrice(1.25, 2.5, {
20586
- cacheRead: 0.2,
20587
- longContext: { thresholdInputTokens: 2e5, input: 2.5, output: 5 }
20588
- }),
20752
+ price: grokPrice(1.25, 2.5, 0.2),
20589
20753
  on: ["bedrock"]
20590
20754
  });
20755
+ xai.model("grok-4.20-0309-reasoning", {
20756
+ name: "Grok 4.20 Reasoning",
20757
+ capability: "reasoning",
20758
+ ctx: 1e6,
20759
+ price: grokPrice(1.25, 2.5, 0.2)
20760
+ });
20761
+ xai.model("grok-4.20-0309-non-reasoning", {
20762
+ name: "Grok 4.20",
20763
+ capability: null,
20764
+ ctx: 1e6,
20765
+ price: grokPrice(1.25, 2.5, 0.2)
20766
+ });
20767
+ xai.model("grok-4.20-multi-agent-0309", {
20768
+ name: "Grok 4.20 Multi-Agent",
20769
+ capability: "reasoning",
20770
+ ctx: 1e6,
20771
+ price: grokPrice(1.25, 2.5, 0.2)
20772
+ });
20591
20773
  xai.model("grok-build-0.1", {
20592
20774
  name: "Grok Build 0.1",
20593
20775
  ctx: 256e3,
20594
20776
  capability: "code",
20595
- price: tokenPrice(1, 2, {
20596
- cacheRead: 0.2,
20597
- longContext: { thresholdInputTokens: 2e5, input: 2, output: 4 }
20598
- })
20777
+ price: grokPrice(1, 2, 0.2)
20599
20778
  });
20600
20779
  var deepseek = vendor("deepseek", {
20601
20780
  family: { id: "deepseek", name: "DeepSeek" },
@@ -20608,16 +20787,25 @@ var deepseek = vendor("deepseek", {
20608
20787
  source: "https://cdn.deepseek.com/policies/en-US/deepseek-privacy-policy.html"
20609
20788
  }
20610
20789
  });
20790
+ deepseek.model("deepseek-flash", {
20791
+ name: "DeepSeek V4.1 Flash",
20792
+ ctx: 1e6,
20793
+ maxOut: 384e3,
20794
+ capability: null,
20795
+ price: tokenPrice(0.3, 1.2, { cacheRead: 6e-3 })
20796
+ });
20611
20797
  deepseek.model("deepseek-v4-flash", {
20612
20798
  name: "DeepSeek V4 Flash",
20613
20799
  ctx: 1e6,
20800
+ maxOut: 384e3,
20614
20801
  capability: null,
20615
- price: tokenPrice(0.44, 1.32, { cacheRead: 0.014 }),
20802
+ price: tokenPrice(0.3, 1.2, { cacheRead: 6e-3 }),
20616
20803
  on: ["bedrock", "together", "fireworks", "openrouter", "ollama", "local"]
20617
20804
  });
20618
20805
  deepseek.model("deepseek-v4-pro", {
20619
20806
  name: "DeepSeek V4 Pro",
20620
20807
  ctx: 1e6,
20808
+ maxOut: 384e3,
20621
20809
  capability: "reasoning",
20622
20810
  price: tokenPrice(1.32, 3.96, { cacheRead: 0.044 }),
20623
20811
  on: ["bedrock", "together", "fireworks", "openrouter"]
@@ -20664,7 +20852,14 @@ cohere.model("command-r-plus-08-2024", {
20664
20852
  ctx: 128e3,
20665
20853
  maxOut: 4e3,
20666
20854
  price: tokenPrice(2.5, 10),
20667
- // The undated form is what deployments report; the live id carries the date.
20855
+ on: ["bedrock"]
20856
+ });
20857
+ cohere.model("command-r-plus-04-2024", {
20858
+ name: "Command R+ (04-2024)",
20859
+ seen: "discovered",
20860
+ ctx: 128e3,
20861
+ maxOut: 4e3,
20862
+ price: tokenPrice(3, 15),
20668
20863
  aliases: ["command-r-plus"],
20669
20864
  on: ["bedrock"]
20670
20865
  });
@@ -20732,6 +20927,65 @@ var mistral = vendor("mistral", {
20732
20927
  retention: "unknown",
20733
20928
  defaultHosting: "local"
20734
20929
  });
20930
+ var MISTRAL_API = { defaultHosting: "api" };
20931
+ mistral.model("mistral-large-4", {
20932
+ ...MISTRAL_API,
20933
+ name: "Mistral Large 4",
20934
+ ctx: 1e6,
20935
+ capability: null,
20936
+ // Public preview at a sale rate; the published original is 1.36 / 4.18 / 0.14.
20937
+ price: tokenPrice(0.68, 2.09, { cacheRead: 0.07 })
20938
+ });
20939
+ mistral.model("mistral-large-2512", {
20940
+ ...MISTRAL_API,
20941
+ name: "Mistral Large 3",
20942
+ ctx: 256e3,
20943
+ capability: null,
20944
+ price: tokenPrice(0.5, 1.5, { cacheRead: 0.05 })
20945
+ });
20946
+ mistral.model("mistral-medium-3-5", {
20947
+ ...MISTRAL_API,
20948
+ name: "Mistral Medium 3.5",
20949
+ ctx: 256e3,
20950
+ capability: null,
20951
+ price: tokenPrice(1.5, 7.5, { cacheRead: 0.15 })
20952
+ });
20953
+ mistral.model("mistral-small-2603", {
20954
+ ...MISTRAL_API,
20955
+ name: "Mistral Small 4",
20956
+ ctx: 256e3,
20957
+ capability: null,
20958
+ price: tokenPrice(0.15, 0.6, { cacheRead: 0.015 })
20959
+ });
20960
+ mistral.model("ministral-14b-2512", {
20961
+ ...MISTRAL_API,
20962
+ name: "Ministral 3 14B",
20963
+ ctx: 256e3,
20964
+ capability: null,
20965
+ price: tokenPrice(0.2, 0.2, { cacheRead: 0.02 })
20966
+ });
20967
+ mistral.model("ministral-8b-2512", {
20968
+ ...MISTRAL_API,
20969
+ name: "Ministral 3 8B",
20970
+ ctx: 256e3,
20971
+ capability: null,
20972
+ price: tokenPrice(0.15, 0.15, { cacheRead: 0.015 })
20973
+ });
20974
+ mistral.model("ministral-3b-2512", {
20975
+ ...MISTRAL_API,
20976
+ name: "Ministral 3 3B",
20977
+ ctx: 256e3,
20978
+ capability: null,
20979
+ price: tokenPrice(0.1, 0.1, { cacheRead: 0.01 })
20980
+ });
20981
+ mistral.model("codestral-2508", {
20982
+ ...MISTRAL_API,
20983
+ name: "Codestral",
20984
+ ctx: 128e3,
20985
+ capability: "code",
20986
+ openWeights: false,
20987
+ price: tokenPrice(0.3, 0.9, { cacheRead: 0.03 })
20988
+ });
20735
20989
  mistral.model("mixtral-8x7b", {
20736
20990
  name: "Mixtral 8x7B",
20737
20991
  seen: "discovered",
@@ -20895,11 +21149,14 @@ var COMPACT = new Intl.NumberFormat("en-US", {
20895
21149
  // ../../packages/schema/src/token/token-report.ts
20896
21150
  var num = (value) => value ?? 0;
20897
21151
  function costUsageOf(a, promptTokens) {
21152
+ const write1h = num(a.ephemeral_1h_input_tokens);
21153
+ const write5m = num(a.ephemeral_5m_input_tokens);
21154
+ const unsplit = Math.max(num(a.cache_creation_input_tokens) - write1h - write5m, 0);
20898
21155
  const usage = {
20899
21156
  inputTokens: num(a.input_tokens),
20900
21157
  outputTokens: num(a.output_tokens),
20901
- cacheWrite1hTokens: num(a.ephemeral_1h_input_tokens),
20902
- cacheWrite5mTokens: num(a.ephemeral_5m_input_tokens),
21158
+ cacheWrite1hTokens: write1h,
21159
+ cacheWrite5mTokens: write5m + unsplit,
20903
21160
  cacheReadTokens: num(a.cache_read_input_tokens),
20904
21161
  webSearchRequests: num(a.web_search_requests)
20905
21162
  };
@@ -27547,7 +27804,8 @@ var TIMELINE_COLUMNS = `
27547
27804
  json_extract(attributes, '$.internal') AS internal,
27548
27805
  json_extract(attributes, '$.flagged') AS flagged`;
27549
27806
  var CALL_PROMPT_TOKENS = `(coalesce(input_tokens, 0) + coalesce(cache_read_input_tokens, 0)
27550
- + coalesce(ephemeral_1h_input_tokens, 0) + coalesce(ephemeral_5m_input_tokens, 0))`;
27807
+ + max(coalesce(cache_creation_input_tokens, 0),
27808
+ coalesce(ephemeral_1h_input_tokens, 0) + coalesce(ephemeral_5m_input_tokens, 0)))`;
27551
27809
  var CALL_BAND = LONG_CONTEXT_THRESHOLDS.length === 0 ? "0" : LONG_CONTEXT_THRESHOLDS.map((t) => `(${CALL_PROMPT_TOKENS} > ${String(Math.trunc(t))})`).join(
27552
27810
  " + "
27553
27811
  );
@@ -36220,6 +36478,10 @@ var GROWTH_CORPORA = [
36220
36478
  build: (length) => joinTo(length, "[", "]", (i) => `"${SQL_WORDS[i % SQL_WORDS.length] ?? ""} ${String(i)}"`)
36221
36479
  },
36222
36480
  { name: "single-quoted keyword run ('where)", build: (length) => repeatTo("'where", length) },
36481
+ {
36482
+ name: 'double-quoted keyword literal run ("where" )',
36483
+ build: (length) => repeatTo('"where" ', length)
36484
+ },
36223
36485
  { name: "double-quote run", build: (length) => repeatTo('"', length) },
36224
36486
  { name: "space run", build: (length) => repeatTo(" ", length) },
36225
36487
  { name: "newline run", build: (length) => repeatTo("\n", length) },
@@ -38591,7 +38853,7 @@ var BUNDLED_PACKS = [
38591
38853
  {
38592
38854
  packId: "code-flaws",
38593
38855
  name: "Code Security Flaws",
38594
- version: "0.1.1",
38856
+ version: "0.1.2",
38595
38857
  rawRules: [
38596
38858
  sql_inject_concat_default,
38597
38859
  sql_inject_concat_dot_default,