@probeo/anymodel 0.3.0 → 0.3.1

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 (115) hide show
  1. package/README.md +1 -10
  2. package/dist/cli.cjs +827 -108
  3. package/dist/cli.cjs.map +1 -1
  4. package/dist/cli.d.ts +0 -2
  5. package/dist/cli.js +2785 -30
  6. package/dist/cli.js.map +1 -1
  7. package/dist/index.cjs +848 -110
  8. package/dist/index.cjs.map +1 -1
  9. package/dist/index.d.cts +106 -22
  10. package/dist/index.d.ts +632 -14
  11. package/dist/index.js +2782 -15
  12. package/dist/index.js.map +1 -1
  13. package/package.json +1 -1
  14. package/dist/batch/index.d.ts +0 -4
  15. package/dist/batch/index.d.ts.map +0 -1
  16. package/dist/batch/index.js +0 -3
  17. package/dist/batch/index.js.map +0 -1
  18. package/dist/batch/manager.d.ts +0 -72
  19. package/dist/batch/manager.d.ts.map +0 -1
  20. package/dist/batch/manager.js +0 -328
  21. package/dist/batch/manager.js.map +0 -1
  22. package/dist/batch/store.d.ts +0 -54
  23. package/dist/batch/store.d.ts.map +0 -1
  24. package/dist/batch/store.js +0 -109
  25. package/dist/batch/store.js.map +0 -1
  26. package/dist/cli.d.ts.map +0 -1
  27. package/dist/client.d.ts +0 -42
  28. package/dist/client.d.ts.map +0 -1
  29. package/dist/client.js +0 -181
  30. package/dist/client.js.map +0 -1
  31. package/dist/config.d.ts +0 -6
  32. package/dist/config.d.ts.map +0 -1
  33. package/dist/config.js +0 -120
  34. package/dist/config.js.map +0 -1
  35. package/dist/index.d.ts.map +0 -1
  36. package/dist/providers/adapter.d.ts +0 -33
  37. package/dist/providers/adapter.d.ts.map +0 -1
  38. package/dist/providers/adapter.js +0 -2
  39. package/dist/providers/adapter.js.map +0 -1
  40. package/dist/providers/anthropic-batch.d.ts +0 -3
  41. package/dist/providers/anthropic-batch.d.ts.map +0 -1
  42. package/dist/providers/anthropic-batch.js +0 -228
  43. package/dist/providers/anthropic-batch.js.map +0 -1
  44. package/dist/providers/anthropic.d.ts +0 -3
  45. package/dist/providers/anthropic.d.ts.map +0 -1
  46. package/dist/providers/anthropic.js +0 -358
  47. package/dist/providers/anthropic.js.map +0 -1
  48. package/dist/providers/custom.d.ts +0 -8
  49. package/dist/providers/custom.d.ts.map +0 -1
  50. package/dist/providers/custom.js +0 -61
  51. package/dist/providers/custom.js.map +0 -1
  52. package/dist/providers/google.d.ts +0 -3
  53. package/dist/providers/google.d.ts.map +0 -1
  54. package/dist/providers/google.js +0 -331
  55. package/dist/providers/google.js.map +0 -1
  56. package/dist/providers/index.d.ts +0 -6
  57. package/dist/providers/index.d.ts.map +0 -1
  58. package/dist/providers/index.js +0 -5
  59. package/dist/providers/index.js.map +0 -1
  60. package/dist/providers/openai-batch.d.ts +0 -3
  61. package/dist/providers/openai-batch.d.ts.map +0 -1
  62. package/dist/providers/openai-batch.js +0 -208
  63. package/dist/providers/openai-batch.js.map +0 -1
  64. package/dist/providers/openai.d.ts +0 -3
  65. package/dist/providers/openai.d.ts.map +0 -1
  66. package/dist/providers/openai.js +0 -221
  67. package/dist/providers/openai.js.map +0 -1
  68. package/dist/providers/registry.d.ts +0 -10
  69. package/dist/providers/registry.d.ts.map +0 -1
  70. package/dist/providers/registry.js +0 -27
  71. package/dist/providers/registry.js.map +0 -1
  72. package/dist/router.d.ts +0 -29
  73. package/dist/router.d.ts.map +0 -1
  74. package/dist/router.js +0 -212
  75. package/dist/router.js.map +0 -1
  76. package/dist/server.d.ts +0 -10
  77. package/dist/server.d.ts.map +0 -1
  78. package/dist/server.js +0 -149
  79. package/dist/server.js.map +0 -1
  80. package/dist/types.d.ts +0 -283
  81. package/dist/types.d.ts.map +0 -1
  82. package/dist/types.js +0 -21
  83. package/dist/types.js.map +0 -1
  84. package/dist/utils/fs-io.d.ts +0 -40
  85. package/dist/utils/fs-io.d.ts.map +0 -1
  86. package/dist/utils/fs-io.js +0 -203
  87. package/dist/utils/fs-io.js.map +0 -1
  88. package/dist/utils/generation-stats.d.ts +0 -25
  89. package/dist/utils/generation-stats.d.ts.map +0 -1
  90. package/dist/utils/generation-stats.js +0 -46
  91. package/dist/utils/generation-stats.js.map +0 -1
  92. package/dist/utils/id.d.ts +0 -2
  93. package/dist/utils/id.d.ts.map +0 -1
  94. package/dist/utils/id.js +0 -6
  95. package/dist/utils/id.js.map +0 -1
  96. package/dist/utils/model-parser.d.ts +0 -6
  97. package/dist/utils/model-parser.d.ts.map +0 -1
  98. package/dist/utils/model-parser.js +0 -21
  99. package/dist/utils/model-parser.js.map +0 -1
  100. package/dist/utils/rate-limiter.d.ts +0 -36
  101. package/dist/utils/rate-limiter.d.ts.map +0 -1
  102. package/dist/utils/rate-limiter.js +0 -80
  103. package/dist/utils/rate-limiter.js.map +0 -1
  104. package/dist/utils/retry.d.ts +0 -7
  105. package/dist/utils/retry.d.ts.map +0 -1
  106. package/dist/utils/retry.js +0 -54
  107. package/dist/utils/retry.js.map +0 -1
  108. package/dist/utils/transforms.d.ts +0 -7
  109. package/dist/utils/transforms.d.ts.map +0 -1
  110. package/dist/utils/transforms.js +0 -66
  111. package/dist/utils/transforms.js.map +0 -1
  112. package/dist/utils/validate.d.ts +0 -3
  113. package/dist/utils/validate.d.ts.map +0 -1
  114. package/dist/utils/validate.js +0 -31
  115. package/dist/utils/validate.js.map +0 -1
package/dist/cli.cjs CHANGED
@@ -1,5 +1,27 @@
1
1
  #!/usr/bin/env node
2
2
  "use strict";
3
+ var __create = Object.create;
4
+ var __defProp = Object.defineProperty;
5
+ var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
6
+ var __getOwnPropNames = Object.getOwnPropertyNames;
7
+ var __getProtoOf = Object.getPrototypeOf;
8
+ var __hasOwnProp = Object.prototype.hasOwnProperty;
9
+ var __copyProps = (to, from, except, desc) => {
10
+ if (from && typeof from === "object" || typeof from === "function") {
11
+ for (let key of __getOwnPropNames(from))
12
+ if (!__hasOwnProp.call(to, key) && key !== except)
13
+ __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
14
+ }
15
+ return to;
16
+ };
17
+ var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
18
+ // If the importer is in node compatibility mode or this is not an ESM
19
+ // file that has been converted to a CommonJS file using a Babel-
20
+ // compatible transform (i.e. "__esModule" has not been set), then set
21
+ // "default" to the CommonJS "module.exports" for node compatibility.
22
+ isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
23
+ mod
24
+ ));
3
25
 
4
26
  // src/server.ts
5
27
  var import_node_http = require("http");
@@ -148,7 +170,7 @@ async function withRetry(fn, options = {}) {
148
170
  throw error;
149
171
  }
150
172
  const delay = computeDelay(attempt, opts, error);
151
- await new Promise((resolve3) => setTimeout(resolve3, delay));
173
+ await new Promise((resolve2) => setTimeout(resolve2, delay));
152
174
  }
153
175
  }
154
176
  throw lastError;
@@ -507,8 +529,8 @@ var SUPPORTED_PARAMS = /* @__PURE__ */ new Set([
507
529
  ]);
508
530
  function createOpenAIAdapter(apiKey, baseURL) {
509
531
  const base = baseURL || OPENAI_API_BASE;
510
- async function makeRequest(path, body, method = "POST") {
511
- const res = await fetch(`${base}${path}`, {
532
+ async function makeRequest(path2, body, method = "POST") {
533
+ const res = await fetch(`${base}${path2}`, {
512
534
  method,
513
535
  headers: {
514
536
  "Content-Type": "application/json",
@@ -623,7 +645,19 @@ function createOpenAIAdapter(apiKey, baseURL) {
623
645
  async listModels() {
624
646
  const res = await makeRequest("/models", void 0, "GET");
625
647
  const data = await res.json();
626
- return (data.data || []).filter((m) => m.id.startsWith("gpt-") || m.id.startsWith("o1") || m.id.startsWith("o3") || m.id.startsWith("o4") || m.id.startsWith("chatgpt-")).map((m) => ({
648
+ return (data.data || []).filter((m) => {
649
+ const id = m.id;
650
+ if (id.includes("embedding")) return false;
651
+ if (id.includes("whisper")) return false;
652
+ if (id.includes("tts")) return false;
653
+ if (id.includes("dall-e")) return false;
654
+ if (id.includes("davinci")) return false;
655
+ if (id.includes("babbage")) return false;
656
+ if (id.includes("moderation")) return false;
657
+ if (id.includes("realtime")) return false;
658
+ if (id.startsWith("ft:")) return false;
659
+ return id.startsWith("gpt-") || id.startsWith("o1") || id.startsWith("o3") || id.startsWith("o4") || id.startsWith("chatgpt-");
660
+ }).map((m) => ({
627
661
  id: `openai/${m.id}`,
628
662
  name: m.id,
629
663
  created: m.created,
@@ -647,6 +681,9 @@ function createOpenAIAdapter(apiKey, baseURL) {
647
681
  supportsParameter(param) {
648
682
  return SUPPORTED_PARAMS.has(param);
649
683
  },
684
+ supportsBatch() {
685
+ return true;
686
+ },
650
687
  async sendRequest(request) {
651
688
  const body = buildRequestBody(request);
652
689
  const res = await makeRequest("/chat/completions", body);
@@ -690,14 +727,19 @@ var SUPPORTED_PARAMS2 = /* @__PURE__ */ new Set([
690
727
  "response_format"
691
728
  ]);
692
729
  var FALLBACK_MODELS = [
693
- { id: "anthropic/claude-opus-4-6", name: "Claude Opus 4.6", created: 0, description: "Most capable model", context_length: 2e5, pricing: { prompt: "0.000005", completion: "0.000025" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image"], output_modalities: ["text"], tokenizer: "claude" }, top_provider: { context_length: 2e5, max_completion_tokens: 32768, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS2) },
730
+ // Claude 4.6
731
+ { id: "anthropic/claude-opus-4-6", name: "Claude Opus 4.6", created: 0, description: "Most capable model", context_length: 2e5, pricing: { prompt: "0.000015", completion: "0.000075" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image"], output_modalities: ["text"], tokenizer: "claude" }, top_provider: { context_length: 2e5, max_completion_tokens: 32768, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS2) },
694
732
  { id: "anthropic/claude-sonnet-4-6", name: "Claude Sonnet 4.6", created: 0, description: "Best balance of speed and capability", context_length: 2e5, pricing: { prompt: "0.000003", completion: "0.000015" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image"], output_modalities: ["text"], tokenizer: "claude" }, top_provider: { context_length: 2e5, max_completion_tokens: 16384, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS2) },
733
+ // Claude 4.5
695
734
  { id: "anthropic/claude-sonnet-4-5-20251022", name: "Claude Sonnet 4.5", created: 0, description: "Previous generation balanced model", context_length: 2e5, pricing: { prompt: "0.000003", completion: "0.000015" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image"], output_modalities: ["text"], tokenizer: "claude" }, top_provider: { context_length: 2e5, max_completion_tokens: 16384, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS2) },
696
- { id: "anthropic/claude-haiku-4-5", name: "Claude Haiku 4.5", created: 0, description: "Fastest and most compact", context_length: 2e5, pricing: { prompt: "0.000001", completion: "0.000005" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image"], output_modalities: ["text"], tokenizer: "claude" }, top_provider: { context_length: 2e5, max_completion_tokens: 8192, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS2) }
735
+ { id: "anthropic/claude-haiku-4-5", name: "Claude Haiku 4.5", created: 0, description: "Fast and compact", context_length: 2e5, pricing: { prompt: "0.000001", completion: "0.000005" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image"], output_modalities: ["text"], tokenizer: "claude" }, top_provider: { context_length: 2e5, max_completion_tokens: 8192, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS2) },
736
+ // Claude 3.5
737
+ { id: "anthropic/claude-3-5-sonnet-20241022", name: "Claude 3.5 Sonnet", created: 0, description: "Legacy balanced model", context_length: 2e5, pricing: { prompt: "0.000003", completion: "0.000015" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image"], output_modalities: ["text"], tokenizer: "claude" }, top_provider: { context_length: 2e5, max_completion_tokens: 8192, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS2) },
738
+ { id: "anthropic/claude-3-5-haiku-20241022", name: "Claude 3.5 Haiku", created: 0, description: "Legacy fast model", context_length: 2e5, pricing: { prompt: "0.0000008", completion: "0.000004" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image"], output_modalities: ["text"], tokenizer: "claude" }, top_provider: { context_length: 2e5, max_completion_tokens: 8192, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS2) }
697
739
  ];
698
740
  function createAnthropicAdapter(apiKey) {
699
- async function makeRequest(path, body, stream = false) {
700
- const res = await fetch(`${ANTHROPIC_API_BASE}${path}`, {
741
+ async function makeRequest(path2, body, stream = false) {
742
+ const res = await fetch(`${ANTHROPIC_API_BASE}${path2}`, {
701
743
  method: "POST",
702
744
  headers: {
703
745
  "Content-Type": "application/json",
@@ -991,6 +1033,9 @@ ${body.system}` : jsonInstruction;
991
1033
  supportsParameter(param) {
992
1034
  return SUPPORTED_PARAMS2.has(param);
993
1035
  },
1036
+ supportsBatch() {
1037
+ return true;
1038
+ },
994
1039
  async sendRequest(request) {
995
1040
  const body = translateRequest(request);
996
1041
  const res = await makeRequest("/messages", body);
@@ -1025,8 +1070,15 @@ var SUPPORTED_PARAMS3 = /* @__PURE__ */ new Set([
1025
1070
  "response_format"
1026
1071
  ]);
1027
1072
  var FALLBACK_MODELS2 = [
1073
+ // Gemini 2.5
1028
1074
  { id: "google/gemini-2.5-pro", name: "Gemini 2.5 Pro", created: 0, description: "Most capable Gemini model", context_length: 1048576, pricing: { prompt: "0.00000125", completion: "0.000005" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image", "video", "audio"], output_modalities: ["text"], tokenizer: "gemini" }, top_provider: { context_length: 1048576, max_completion_tokens: 65536, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS3) },
1029
- { id: "google/gemini-2.5-flash", name: "Gemini 2.5 Flash", created: 0, description: "Fast and efficient", context_length: 1048576, pricing: { prompt: "0.00000015", completion: "0.0000006" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image", "video", "audio"], output_modalities: ["text"], tokenizer: "gemini" }, top_provider: { context_length: 1048576, max_completion_tokens: 65536, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS3) }
1075
+ { id: "google/gemini-2.5-flash", name: "Gemini 2.5 Flash", created: 0, description: "Fast and efficient", context_length: 1048576, pricing: { prompt: "0.00000015", completion: "0.0000006" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image", "video", "audio"], output_modalities: ["text"], tokenizer: "gemini" }, top_provider: { context_length: 1048576, max_completion_tokens: 65536, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS3) },
1076
+ // Gemini 2.0
1077
+ { id: "google/gemini-2.0-flash", name: "Gemini 2.0 Flash", created: 0, description: "Fast multimodal model", context_length: 1048576, pricing: { prompt: "0.0000001", completion: "0.0000004" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image", "video", "audio"], output_modalities: ["text"], tokenizer: "gemini" }, top_provider: { context_length: 1048576, max_completion_tokens: 65536, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS3) },
1078
+ { id: "google/gemini-2.0-flash-lite", name: "Gemini 2.0 Flash Lite", created: 0, description: "Lightweight and fast", context_length: 1048576, pricing: { prompt: "0.00000005", completion: "0.0000002" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image", "video", "audio"], output_modalities: ["text"], tokenizer: "gemini" }, top_provider: { context_length: 1048576, max_completion_tokens: 65536, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS3) },
1079
+ // Gemini 1.5
1080
+ { id: "google/gemini-1.5-pro", name: "Gemini 1.5 Pro", created: 0, description: "Previous generation pro model", context_length: 2097152, pricing: { prompt: "0.00000125", completion: "0.000005" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image", "video", "audio"], output_modalities: ["text"], tokenizer: "gemini" }, top_provider: { context_length: 2097152, max_completion_tokens: 8192, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS3) },
1081
+ { id: "google/gemini-1.5-flash", name: "Gemini 1.5 Flash", created: 0, description: "Previous generation flash model", context_length: 1048576, pricing: { prompt: "0.000000075", completion: "0.0000003" }, architecture: { modality: "text+image->text", input_modalities: ["text", "image", "video", "audio"], output_modalities: ["text"], tokenizer: "gemini" }, top_provider: { context_length: 1048576, max_completion_tokens: 8192, is_moderated: false }, supported_parameters: Array.from(SUPPORTED_PARAMS3) }
1030
1082
  ];
1031
1083
  function createGoogleAdapter(apiKey) {
1032
1084
  function getModelEndpoint(model, stream) {
@@ -1263,6 +1315,9 @@ function createGoogleAdapter(apiKey) {
1263
1315
  supportsParameter(param) {
1264
1316
  return SUPPORTED_PARAMS3.has(param);
1265
1317
  },
1318
+ supportsBatch() {
1319
+ return false;
1320
+ },
1266
1321
  async sendRequest(request) {
1267
1322
  const body = translateRequest(request);
1268
1323
  const url = getModelEndpoint(request.model, false);
@@ -1321,6 +1376,9 @@ function createCustomAdapter(name, config) {
1321
1376
  return {
1322
1377
  ...openaiAdapter,
1323
1378
  name,
1379
+ supportsBatch() {
1380
+ return false;
1381
+ },
1324
1382
  async listModels() {
1325
1383
  if (config.models && config.models.length > 0) {
1326
1384
  return config.models.map((modelId) => ({
@@ -1386,10 +1444,10 @@ function interpolateDeep(obj) {
1386
1444
  }
1387
1445
  return obj;
1388
1446
  }
1389
- function loadJsonFile(path) {
1390
- if (!(0, import_node_fs.existsSync)(path)) return null;
1447
+ function loadJsonFile(path2) {
1448
+ if (!(0, import_node_fs.existsSync)(path2)) return null;
1391
1449
  try {
1392
- const raw = (0, import_node_fs.readFileSync)(path, "utf-8");
1450
+ const raw = (0, import_node_fs.readFileSync)(path2, "utf-8");
1393
1451
  const parsed = JSON.parse(raw);
1394
1452
  return interpolateDeep(parsed);
1395
1453
  } catch {
@@ -1490,93 +1548,228 @@ var GenerationStatsStore = class {
1490
1548
  }
1491
1549
  };
1492
1550
 
1493
- // src/batch/store.ts
1551
+ // src/utils/fs-io.ts
1552
+ var import_promises = require("fs/promises");
1494
1553
  var import_node_fs2 = require("fs");
1495
- var import_node_path2 = require("path");
1496
- var import_node_os2 = require("os");
1497
- var DEFAULT_BATCH_DIR = (0, import_node_path2.join)((0, import_node_os2.homedir)(), ".anymodel", "batches");
1554
+ var import_node_path2 = __toESM(require("path"), 1);
1555
+ var import_p_queue = __toESM(require("p-queue"), 1);
1556
+ var writeQueue = new import_p_queue.default({ concurrency: 10 });
1557
+ var readQueue = new import_p_queue.default({ concurrency: 20 });
1558
+ function configureFsIO(options) {
1559
+ if (options.readConcurrency !== void 0) {
1560
+ readQueue.concurrency = options.readConcurrency;
1561
+ }
1562
+ if (options.writeConcurrency !== void 0) {
1563
+ writeQueue.concurrency = options.writeConcurrency;
1564
+ }
1565
+ }
1566
+ var ensuredDirs = /* @__PURE__ */ new Set();
1567
+ var joinPathCache = /* @__PURE__ */ new Map();
1568
+ var dirnameCache = /* @__PURE__ */ new Map();
1569
+ var resolvePathCache = /* @__PURE__ */ new Map();
1570
+ async function ensureDir(dir) {
1571
+ if (!dir) return;
1572
+ if (ensuredDirs.has(dir)) return;
1573
+ await (0, import_promises.mkdir)(dir, { recursive: true });
1574
+ ensuredDirs.add(dir);
1575
+ }
1576
+ async function readFileQueued(filePath, encoding = "utf8") {
1577
+ return readQueue.add(async () => {
1578
+ return (0, import_promises.readFile)(filePath, encoding);
1579
+ });
1580
+ }
1581
+ async function readJsonQueued(filePath) {
1582
+ const raw = await readFileQueued(filePath, "utf8");
1583
+ return JSON.parse(raw);
1584
+ }
1585
+ async function readDirQueued(dirPath) {
1586
+ return readQueue.add(async () => {
1587
+ return (0, import_promises.readdir)(dirPath, { withFileTypes: true });
1588
+ });
1589
+ }
1590
+ async function pathExistsQueued(p) {
1591
+ return readQueue.add(async () => {
1592
+ try {
1593
+ await (0, import_promises.stat)(p);
1594
+ return true;
1595
+ } catch {
1596
+ return false;
1597
+ }
1598
+ });
1599
+ }
1600
+ async function fileExistsQueued(filePath) {
1601
+ return readQueue.add(async () => {
1602
+ try {
1603
+ const s = await (0, import_promises.stat)(filePath);
1604
+ return s.isFile();
1605
+ } catch {
1606
+ return false;
1607
+ }
1608
+ });
1609
+ }
1610
+ async function writeFileQueued(filePath, data) {
1611
+ await writeQueue.add(async () => {
1612
+ const dir = dirnameOf(filePath);
1613
+ await ensureDir(dir);
1614
+ await (0, import_promises.writeFile)(filePath, data);
1615
+ });
1616
+ }
1617
+ async function appendFileQueued(filePath, data) {
1618
+ await writeQueue.add(async () => {
1619
+ const dir = dirnameOf(filePath);
1620
+ await ensureDir(dir);
1621
+ await (0, import_promises.writeFile)(filePath, data, { flag: "a" });
1622
+ });
1623
+ }
1624
+ async function writeFileFlushedQueued(filePath, data) {
1625
+ await writeQueue.add(async () => {
1626
+ const dir = dirnameOf(filePath);
1627
+ await ensureDir(dir);
1628
+ const tmpPath = joinPath(
1629
+ dir,
1630
+ `.${import_node_path2.default.basename(filePath)}.${Date.now()}.${Math.random().toString(16).slice(2)}.tmp`
1631
+ );
1632
+ const fh = await (0, import_promises.open)(tmpPath, "w");
1633
+ try {
1634
+ await fh.writeFile(data);
1635
+ await fh.sync();
1636
+ } finally {
1637
+ await fh.close();
1638
+ }
1639
+ await (0, import_promises.rename)(tmpPath, filePath);
1640
+ try {
1641
+ const dh = await (0, import_promises.open)(dir, "r");
1642
+ try {
1643
+ await dh.sync();
1644
+ } finally {
1645
+ await dh.close();
1646
+ }
1647
+ } catch {
1648
+ }
1649
+ });
1650
+ }
1651
+ function joinPath(...segments) {
1652
+ const key = segments.join("\0");
1653
+ const cached = joinPathCache.get(key);
1654
+ if (cached !== void 0) return cached;
1655
+ const out = import_node_path2.default.join(...segments);
1656
+ joinPathCache.set(key, out);
1657
+ return out;
1658
+ }
1659
+ function dirnameOf(p) {
1660
+ const cached = dirnameCache.get(p);
1661
+ if (cached !== void 0) return cached;
1662
+ const out = import_node_path2.default.dirname(p);
1663
+ dirnameCache.set(p, out);
1664
+ return out;
1665
+ }
1666
+ function resolvePath(...segments) {
1667
+ const key = segments.join("\0");
1668
+ const cached = resolvePathCache.get(key);
1669
+ if (cached !== void 0) return cached;
1670
+ const out = import_node_path2.default.resolve(...segments);
1671
+ resolvePathCache.set(key, out);
1672
+ return out;
1673
+ }
1674
+
1675
+ // src/batch/store.ts
1676
+ var DEFAULT_BATCH_DIR = joinPath(process.cwd(), ".anymodel", "batches");
1498
1677
  var BatchStore = class {
1499
1678
  dir;
1679
+ initialized = false;
1500
1680
  constructor(dir) {
1501
- this.dir = (0, import_node_path2.resolve)(dir || DEFAULT_BATCH_DIR);
1502
- (0, import_node_fs2.mkdirSync)(this.dir, { recursive: true });
1681
+ this.dir = resolvePath(dir || DEFAULT_BATCH_DIR);
1682
+ }
1683
+ async init() {
1684
+ if (this.initialized) return;
1685
+ await ensureDir(this.dir);
1686
+ this.initialized = true;
1503
1687
  }
1504
1688
  batchDir(id) {
1505
- return (0, import_node_path2.join)(this.dir, id);
1689
+ return joinPath(this.dir, id);
1506
1690
  }
1507
1691
  /**
1508
1692
  * Create a new batch directory and save initial metadata.
1509
1693
  */
1510
- create(batch) {
1694
+ async create(batch) {
1695
+ await this.init();
1511
1696
  const dir = this.batchDir(batch.id);
1512
- (0, import_node_fs2.mkdirSync)(dir, { recursive: true });
1513
- (0, import_node_fs2.writeFileSync)((0, import_node_path2.join)(dir, "meta.json"), JSON.stringify(batch, null, 2));
1697
+ await ensureDir(dir);
1698
+ await writeFileFlushedQueued(joinPath(dir, "meta.json"), JSON.stringify(batch, null, 2));
1514
1699
  }
1515
1700
  /**
1516
- * Update batch metadata.
1701
+ * Update batch metadata (atomic write).
1517
1702
  */
1518
- updateMeta(batch) {
1519
- const dir = this.batchDir(batch.id);
1520
- (0, import_node_fs2.writeFileSync)((0, import_node_path2.join)(dir, "meta.json"), JSON.stringify(batch, null, 2));
1703
+ async updateMeta(batch) {
1704
+ await writeFileFlushedQueued(
1705
+ joinPath(this.batchDir(batch.id), "meta.json"),
1706
+ JSON.stringify(batch, null, 2)
1707
+ );
1521
1708
  }
1522
1709
  /**
1523
1710
  * Save requests as JSONL.
1524
1711
  */
1525
- saveRequests(id, requests) {
1526
- const dir = this.batchDir(id);
1712
+ async saveRequests(id, requests) {
1527
1713
  const lines = requests.map((r) => JSON.stringify(r)).join("\n") + "\n";
1528
- (0, import_node_fs2.writeFileSync)((0, import_node_path2.join)(dir, "requests.jsonl"), lines);
1714
+ await writeFileQueued(joinPath(this.batchDir(id), "requests.jsonl"), lines);
1529
1715
  }
1530
1716
  /**
1531
1717
  * Append a result to results.jsonl.
1532
1718
  */
1533
- appendResult(id, result) {
1534
- const dir = this.batchDir(id);
1535
- (0, import_node_fs2.appendFileSync)((0, import_node_path2.join)(dir, "results.jsonl"), JSON.stringify(result) + "\n");
1719
+ async appendResult(id, result) {
1720
+ await appendFileQueued(
1721
+ joinPath(this.batchDir(id), "results.jsonl"),
1722
+ JSON.stringify(result) + "\n"
1723
+ );
1536
1724
  }
1537
1725
  /**
1538
1726
  * Save provider-specific state (e.g., provider batch ID).
1539
1727
  */
1540
- saveProviderState(id, state) {
1541
- const dir = this.batchDir(id);
1542
- (0, import_node_fs2.writeFileSync)((0, import_node_path2.join)(dir, "provider.json"), JSON.stringify(state, null, 2));
1728
+ async saveProviderState(id, state) {
1729
+ await writeFileFlushedQueued(
1730
+ joinPath(this.batchDir(id), "provider.json"),
1731
+ JSON.stringify(state, null, 2)
1732
+ );
1543
1733
  }
1544
1734
  /**
1545
1735
  * Load provider state.
1546
1736
  */
1547
- loadProviderState(id) {
1548
- const path = (0, import_node_path2.join)(this.batchDir(id), "provider.json");
1549
- if (!(0, import_node_fs2.existsSync)(path)) return null;
1550
- return JSON.parse((0, import_node_fs2.readFileSync)(path, "utf-8"));
1737
+ async loadProviderState(id) {
1738
+ const p = joinPath(this.batchDir(id), "provider.json");
1739
+ if (!await fileExistsQueued(p)) return null;
1740
+ return readJsonQueued(p);
1551
1741
  }
1552
1742
  /**
1553
1743
  * Get batch metadata.
1554
1744
  */
1555
- getMeta(id) {
1556
- const path = (0, import_node_path2.join)(this.batchDir(id), "meta.json");
1557
- if (!(0, import_node_fs2.existsSync)(path)) return null;
1558
- return JSON.parse((0, import_node_fs2.readFileSync)(path, "utf-8"));
1745
+ async getMeta(id) {
1746
+ const p = joinPath(this.batchDir(id), "meta.json");
1747
+ if (!await fileExistsQueued(p)) return null;
1748
+ return readJsonQueued(p);
1559
1749
  }
1560
1750
  /**
1561
1751
  * Get all results for a batch.
1562
1752
  */
1563
- getResults(id) {
1564
- const path = (0, import_node_path2.join)(this.batchDir(id), "results.jsonl");
1565
- if (!(0, import_node_fs2.existsSync)(path)) return [];
1566
- return (0, import_node_fs2.readFileSync)(path, "utf-8").trim().split("\n").filter(Boolean).map((line) => JSON.parse(line));
1753
+ async getResults(id) {
1754
+ const p = joinPath(this.batchDir(id), "results.jsonl");
1755
+ if (!await fileExistsQueued(p)) return [];
1756
+ const raw = await readFileQueued(p, "utf8");
1757
+ return raw.trim().split("\n").filter(Boolean).map((line) => JSON.parse(line));
1567
1758
  }
1568
1759
  /**
1569
1760
  * List all batch IDs.
1570
1761
  */
1571
- listBatches() {
1572
- if (!(0, import_node_fs2.existsSync)(this.dir)) return [];
1573
- return (0, import_node_fs2.readdirSync)(this.dir, { withFileTypes: true }).filter((d) => d.isDirectory()).map((d) => d.name).sort();
1762
+ async listBatches() {
1763
+ await this.init();
1764
+ if (!await pathExistsQueued(this.dir)) return [];
1765
+ const entries = await readDirQueued(this.dir);
1766
+ return entries.filter((d) => d.isDirectory()).map((d) => d.name).sort();
1574
1767
  }
1575
1768
  /**
1576
1769
  * Check if a batch exists.
1577
1770
  */
1578
- exists(id) {
1579
- return (0, import_node_fs2.existsSync)((0, import_node_path2.join)(this.batchDir(id), "meta.json"));
1771
+ async exists(id) {
1772
+ return fileExistsQueued(joinPath(this.batchDir(id), "meta.json"));
1580
1773
  }
1581
1774
  };
1582
1775
 
@@ -1585,10 +1778,27 @@ var BatchManager = class {
1585
1778
  store;
1586
1779
  router;
1587
1780
  concurrencyLimit;
1781
+ defaultPollInterval;
1782
+ batchAdapters = /* @__PURE__ */ new Map();
1588
1783
  constructor(router, options) {
1589
1784
  this.store = new BatchStore(options?.dir);
1590
1785
  this.router = router;
1591
1786
  this.concurrencyLimit = options?.concurrency ?? 5;
1787
+ this.defaultPollInterval = options?.pollInterval ?? 5e3;
1788
+ }
1789
+ /**
1790
+ * Register a native batch adapter for a provider.
1791
+ */
1792
+ registerBatchAdapter(providerName, adapter) {
1793
+ this.batchAdapters.set(providerName, adapter);
1794
+ }
1795
+ /**
1796
+ * Check if a provider has native batch support.
1797
+ */
1798
+ getNativeBatchAdapter(model) {
1799
+ const providerName = model.split("/")[0];
1800
+ const adapter = this.batchAdapters.get(providerName);
1801
+ return adapter ? { adapter, providerName } : null;
1592
1802
  }
1593
1803
  /**
1594
1804
  * Create a batch and return immediately (no polling).
@@ -1596,13 +1806,16 @@ var BatchManager = class {
1596
1806
  async create(request) {
1597
1807
  const id = generateId("batch");
1598
1808
  const now = (/* @__PURE__ */ new Date()).toISOString();
1809
+ const providerName = request.model.split("/")[0] || "unknown";
1810
+ const native = this.getNativeBatchAdapter(request.model);
1811
+ const batchMode = native ? "native" : "concurrent";
1599
1812
  const batch = {
1600
1813
  id,
1601
1814
  object: "batch",
1602
1815
  status: "pending",
1603
1816
  model: request.model,
1604
- provider_name: request.model.split("/")[0] || "unknown",
1605
- batch_mode: "concurrent",
1817
+ provider_name: providerName,
1818
+ batch_mode: batchMode,
1606
1819
  total: request.requests.length,
1607
1820
  completed: 0,
1608
1821
  failed: 0,
@@ -1610,10 +1823,15 @@ var BatchManager = class {
1610
1823
  completed_at: null,
1611
1824
  expires_at: null
1612
1825
  };
1613
- this.store.create(batch);
1614
- this.store.saveRequests(id, request.requests);
1615
- this.processBatch(id, request).catch(() => {
1616
- });
1826
+ await this.store.create(batch);
1827
+ await this.store.saveRequests(id, request.requests);
1828
+ if (native) {
1829
+ this.processNativeBatch(id, request, native.adapter).catch(() => {
1830
+ });
1831
+ } else {
1832
+ this.processConcurrentBatch(id, request).catch(() => {
1833
+ });
1834
+ }
1617
1835
  return batch;
1618
1836
  }
1619
1837
  /**
@@ -1627,14 +1845,19 @@ var BatchManager = class {
1627
1845
  * Poll an existing batch until completion.
1628
1846
  */
1629
1847
  async poll(id, options = {}) {
1630
- const interval = options.interval ?? 5e3;
1848
+ const interval = options.interval ?? this.defaultPollInterval;
1631
1849
  const timeout = options.timeout ?? 0;
1632
1850
  const startTime = Date.now();
1633
1851
  while (true) {
1634
- const batch = this.store.getMeta(id);
1852
+ let batch = await this.store.getMeta(id);
1635
1853
  if (!batch) {
1636
1854
  throw new AnyModelError(404, `Batch ${id} not found`);
1637
1855
  }
1856
+ if (batch.batch_mode === "native" && batch.status === "processing") {
1857
+ await this.syncNativeBatchStatus(id);
1858
+ batch = await this.store.getMeta(id);
1859
+ if (!batch) throw new AnyModelError(404, `Batch ${id} not found`);
1860
+ }
1638
1861
  if (options.onProgress) {
1639
1862
  options.onProgress(batch);
1640
1863
  }
@@ -1644,24 +1867,24 @@ var BatchManager = class {
1644
1867
  if (timeout > 0 && Date.now() - startTime > timeout) {
1645
1868
  throw new AnyModelError(408, `Batch ${id} timed out after ${timeout}ms`);
1646
1869
  }
1647
- await new Promise((resolve3) => setTimeout(resolve3, interval));
1870
+ await new Promise((resolve2) => setTimeout(resolve2, interval));
1648
1871
  }
1649
1872
  }
1650
1873
  /**
1651
1874
  * Get the current status of a batch.
1652
1875
  */
1653
- get(id) {
1876
+ async get(id) {
1654
1877
  return this.store.getMeta(id);
1655
1878
  }
1656
1879
  /**
1657
1880
  * Get results for a completed batch.
1658
1881
  */
1659
- getResults(id) {
1660
- const batch = this.store.getMeta(id);
1882
+ async getResults(id) {
1883
+ const batch = await this.store.getMeta(id);
1661
1884
  if (!batch) {
1662
1885
  throw new AnyModelError(404, `Batch ${id} not found`);
1663
1886
  }
1664
- const results = this.store.getResults(id);
1887
+ const results = await this.store.getResults(id);
1665
1888
  const usage = {
1666
1889
  total_prompt_tokens: 0,
1667
1890
  total_completion_tokens: 0,
@@ -1683,37 +1906,119 @@ var BatchManager = class {
1683
1906
  /**
1684
1907
  * List all batches.
1685
1908
  */
1686
- list() {
1687
- return this.store.listBatches().map((id) => this.store.getMeta(id)).filter((b) => b !== null);
1909
+ async list() {
1910
+ const ids = await this.store.listBatches();
1911
+ const batches = [];
1912
+ for (const id of ids) {
1913
+ const meta = await this.store.getMeta(id);
1914
+ if (meta) batches.push(meta);
1915
+ }
1916
+ return batches;
1688
1917
  }
1689
1918
  /**
1690
1919
  * Cancel a batch.
1691
1920
  */
1692
- cancel(id) {
1693
- const batch = this.store.getMeta(id);
1921
+ async cancel(id) {
1922
+ const batch = await this.store.getMeta(id);
1694
1923
  if (!batch) {
1695
1924
  throw new AnyModelError(404, `Batch ${id} not found`);
1696
1925
  }
1697
1926
  if (batch.status === "completed" || batch.status === "cancelled") {
1698
1927
  return batch;
1699
1928
  }
1929
+ if (batch.batch_mode === "native") {
1930
+ const providerState = await this.store.loadProviderState(id);
1931
+ const adapter = this.batchAdapters.get(batch.provider_name);
1932
+ if (adapter && providerState?.providerBatchId) {
1933
+ try {
1934
+ await adapter.cancelBatch(providerState.providerBatchId);
1935
+ } catch {
1936
+ }
1937
+ }
1938
+ }
1700
1939
  batch.status = "cancelled";
1701
1940
  batch.completed_at = (/* @__PURE__ */ new Date()).toISOString();
1702
- this.store.updateMeta(batch);
1941
+ await this.store.updateMeta(batch);
1703
1942
  return batch;
1704
1943
  }
1705
1944
  /**
1706
- * Process batch requests concurrently.
1945
+ * Process batch via native provider batch API.
1707
1946
  */
1708
- async processBatch(batchId, request) {
1709
- const batch = this.store.getMeta(batchId);
1947
+ async processNativeBatch(batchId, request, adapter) {
1948
+ const batch = await this.store.getMeta(batchId);
1949
+ if (!batch) return;
1950
+ try {
1951
+ const model = request.model.includes("/") ? request.model.split("/").slice(1).join("/") : request.model;
1952
+ const { providerBatchId, metadata } = await adapter.createBatch(
1953
+ model,
1954
+ request.requests,
1955
+ request.options
1956
+ );
1957
+ await this.store.saveProviderState(batchId, {
1958
+ providerBatchId,
1959
+ providerName: batch.provider_name,
1960
+ ...metadata
1961
+ });
1962
+ batch.status = "processing";
1963
+ await this.store.updateMeta(batch);
1964
+ } catch (err) {
1965
+ batch.status = "failed";
1966
+ batch.completed_at = (/* @__PURE__ */ new Date()).toISOString();
1967
+ await this.store.updateMeta(batch);
1968
+ throw err;
1969
+ }
1970
+ }
1971
+ /**
1972
+ * Sync native batch status from provider.
1973
+ */
1974
+ async syncNativeBatchStatus(batchId) {
1975
+ const batch = await this.store.getMeta(batchId);
1976
+ if (!batch) return;
1977
+ const providerState = await this.store.loadProviderState(batchId);
1978
+ if (!providerState?.providerBatchId) return;
1979
+ const adapter = this.batchAdapters.get(batch.provider_name);
1980
+ if (!adapter) return;
1981
+ try {
1982
+ const status = await adapter.pollBatch(providerState.providerBatchId);
1983
+ batch.total = status.total || batch.total;
1984
+ batch.completed = status.completed;
1985
+ batch.failed = status.failed;
1986
+ if (status.status === "completed" || status.status === "failed" || status.status === "cancelled") {
1987
+ batch.status = status.status;
1988
+ batch.completed_at = (/* @__PURE__ */ new Date()).toISOString();
1989
+ if (status.status === "completed" || status.status === "failed") {
1990
+ try {
1991
+ const results = await adapter.getBatchResults(providerState.providerBatchId);
1992
+ for (const result of results) {
1993
+ await this.store.appendResult(batchId, result);
1994
+ }
1995
+ batch.completed = results.filter((r) => r.status === "success").length;
1996
+ batch.failed = results.filter((r) => r.status === "error").length;
1997
+ } catch {
1998
+ if (batch.status !== "failed") {
1999
+ batch.status = "failed";
2000
+ }
2001
+ }
2002
+ }
2003
+ } else {
2004
+ batch.status = "processing";
2005
+ }
2006
+ await this.store.updateMeta(batch);
2007
+ } catch {
2008
+ }
2009
+ }
2010
+ /**
2011
+ * Process batch requests concurrently (fallback path).
2012
+ */
2013
+ async processConcurrentBatch(batchId, request) {
2014
+ const batch = await this.store.getMeta(batchId);
2015
+ if (!batch) return;
1710
2016
  batch.status = "processing";
1711
- this.store.updateMeta(batch);
2017
+ await this.store.updateMeta(batch);
1712
2018
  const items = request.requests;
1713
- const queue = [...items];
1714
2019
  const active = /* @__PURE__ */ new Set();
1715
2020
  const processItem = async (item) => {
1716
- const current = this.store.getMeta(batchId);
2021
+ const current = await this.store.getMeta(batchId);
1717
2022
  if (current?.status === "cancelled") return;
1718
2023
  const chatRequest = {
1719
2024
  model: request.model,
@@ -1745,17 +2050,19 @@ var BatchManager = class {
1745
2050
  error: { code: error.code, message: error.message }
1746
2051
  };
1747
2052
  }
1748
- this.store.appendResult(batchId, result);
1749
- const meta = this.store.getMeta(batchId);
1750
- if (result.status === "success") {
1751
- meta.completed++;
1752
- } else {
1753
- meta.failed++;
2053
+ await this.store.appendResult(batchId, result);
2054
+ const meta = await this.store.getMeta(batchId);
2055
+ if (meta) {
2056
+ if (result.status === "success") {
2057
+ meta.completed++;
2058
+ } else {
2059
+ meta.failed++;
2060
+ }
2061
+ await this.store.updateMeta(meta);
1754
2062
  }
1755
- this.store.updateMeta(meta);
1756
2063
  };
1757
- for (const item of queue) {
1758
- const current = this.store.getMeta(batchId);
2064
+ for (const item of items) {
2065
+ const current = await this.store.getMeta(batchId);
1759
2066
  if (current?.status === "cancelled") break;
1760
2067
  if (active.size >= this.concurrencyLimit) {
1761
2068
  await Promise.race(active);
@@ -1766,15 +2073,411 @@ var BatchManager = class {
1766
2073
  active.add(promise);
1767
2074
  }
1768
2075
  await Promise.all(active);
1769
- const finalMeta = this.store.getMeta(batchId);
1770
- if (finalMeta.status !== "cancelled") {
2076
+ const finalMeta = await this.store.getMeta(batchId);
2077
+ if (finalMeta && finalMeta.status !== "cancelled") {
1771
2078
  finalMeta.status = finalMeta.failed === finalMeta.total ? "failed" : "completed";
1772
2079
  finalMeta.completed_at = (/* @__PURE__ */ new Date()).toISOString();
1773
- this.store.updateMeta(finalMeta);
2080
+ await this.store.updateMeta(finalMeta);
1774
2081
  }
1775
2082
  }
1776
2083
  };
1777
2084
 
2085
+ // src/providers/openai-batch.ts
2086
+ var OPENAI_API_BASE2 = "https://api.openai.com/v1";
2087
+ function createOpenAIBatchAdapter(apiKey) {
2088
+ async function apiRequest(path2, options = {}) {
2089
+ const headers = {
2090
+ "Authorization": `Bearer ${apiKey}`
2091
+ };
2092
+ let fetchBody;
2093
+ if (options.formData) {
2094
+ fetchBody = options.formData;
2095
+ } else if (options.body) {
2096
+ headers["Content-Type"] = "application/json";
2097
+ fetchBody = JSON.stringify(options.body);
2098
+ }
2099
+ const res = await fetch(`${OPENAI_API_BASE2}${path2}`, {
2100
+ method: options.method || "GET",
2101
+ headers,
2102
+ body: fetchBody
2103
+ });
2104
+ if (!res.ok) {
2105
+ let errorBody;
2106
+ try {
2107
+ errorBody = await res.json();
2108
+ } catch {
2109
+ errorBody = { message: res.statusText };
2110
+ }
2111
+ const msg = errorBody?.error?.message || errorBody?.message || res.statusText;
2112
+ throw new AnyModelError(res.status >= 500 ? 502 : res.status, msg, {
2113
+ provider_name: "openai",
2114
+ raw: errorBody
2115
+ });
2116
+ }
2117
+ return res;
2118
+ }
2119
+ function buildJSONL(model, requests) {
2120
+ return requests.map((req) => {
2121
+ const body = {
2122
+ model,
2123
+ messages: req.messages
2124
+ };
2125
+ if (req.max_tokens !== void 0) body.max_tokens = req.max_tokens;
2126
+ if (req.temperature !== void 0) body.temperature = req.temperature;
2127
+ if (req.top_p !== void 0) body.top_p = req.top_p;
2128
+ if (req.stop !== void 0) body.stop = req.stop;
2129
+ if (req.response_format !== void 0) body.response_format = req.response_format;
2130
+ if (req.tools !== void 0) body.tools = req.tools;
2131
+ if (req.tool_choice !== void 0) body.tool_choice = req.tool_choice;
2132
+ return JSON.stringify({
2133
+ custom_id: req.custom_id,
2134
+ method: "POST",
2135
+ url: "/v1/chat/completions",
2136
+ body
2137
+ });
2138
+ }).join("\n");
2139
+ }
2140
+ function rePrefixId(id) {
2141
+ if (id && id.startsWith("chatcmpl-")) {
2142
+ return `gen-${id.substring(9)}`;
2143
+ }
2144
+ return id.startsWith("gen-") ? id : `gen-${id}`;
2145
+ }
2146
+ function translateOpenAIResponse(body) {
2147
+ return {
2148
+ id: rePrefixId(body.id || generateId()),
2149
+ object: "chat.completion",
2150
+ created: body.created || Math.floor(Date.now() / 1e3),
2151
+ model: `openai/${body.model}`,
2152
+ choices: body.choices,
2153
+ usage: body.usage
2154
+ };
2155
+ }
2156
+ function mapStatus(openaiStatus) {
2157
+ switch (openaiStatus) {
2158
+ case "validating":
2159
+ case "finalizing":
2160
+ return "processing";
2161
+ case "in_progress":
2162
+ return "processing";
2163
+ case "completed":
2164
+ return "completed";
2165
+ case "failed":
2166
+ return "failed";
2167
+ case "expired":
2168
+ return "failed";
2169
+ case "cancelled":
2170
+ case "cancelling":
2171
+ return "cancelled";
2172
+ default:
2173
+ return "pending";
2174
+ }
2175
+ }
2176
+ return {
2177
+ async createBatch(model, requests, options) {
2178
+ const jsonlContent = buildJSONL(model, requests);
2179
+ const blob = new Blob([jsonlContent], { type: "application/jsonl" });
2180
+ const formData = new FormData();
2181
+ formData.append("purpose", "batch");
2182
+ formData.append("file", blob, "batch_input.jsonl");
2183
+ const uploadRes = await apiRequest("/files", { method: "POST", formData });
2184
+ const fileData = await uploadRes.json();
2185
+ const inputFileId = fileData.id;
2186
+ const batchRes = await apiRequest("/batches", {
2187
+ method: "POST",
2188
+ body: {
2189
+ input_file_id: inputFileId,
2190
+ endpoint: "/v1/chat/completions",
2191
+ completion_window: "24h",
2192
+ metadata: options?.metadata
2193
+ }
2194
+ });
2195
+ const batchData = await batchRes.json();
2196
+ return {
2197
+ providerBatchId: batchData.id,
2198
+ metadata: {
2199
+ input_file_id: inputFileId,
2200
+ openai_status: batchData.status
2201
+ }
2202
+ };
2203
+ },
2204
+ async pollBatch(providerBatchId) {
2205
+ const res = await apiRequest(`/batches/${providerBatchId}`);
2206
+ const data = await res.json();
2207
+ const requestCounts = data.request_counts || {};
2208
+ return {
2209
+ status: mapStatus(data.status),
2210
+ total: requestCounts.total || 0,
2211
+ completed: requestCounts.completed || 0,
2212
+ failed: requestCounts.failed || 0
2213
+ };
2214
+ },
2215
+ async getBatchResults(providerBatchId) {
2216
+ const batchRes = await apiRequest(`/batches/${providerBatchId}`);
2217
+ const batchData = await batchRes.json();
2218
+ const results = [];
2219
+ if (batchData.output_file_id) {
2220
+ const outputRes = await apiRequest(`/files/${batchData.output_file_id}/content`);
2221
+ const outputText = await outputRes.text();
2222
+ for (const line of outputText.trim().split("\n")) {
2223
+ if (!line) continue;
2224
+ const item = JSON.parse(line);
2225
+ if (item.response?.status_code === 200) {
2226
+ results.push({
2227
+ custom_id: item.custom_id,
2228
+ status: "success",
2229
+ response: translateOpenAIResponse(item.response.body),
2230
+ error: null
2231
+ });
2232
+ } else {
2233
+ results.push({
2234
+ custom_id: item.custom_id,
2235
+ status: "error",
2236
+ response: null,
2237
+ error: {
2238
+ code: item.response?.status_code || 500,
2239
+ message: item.error?.message || item.response?.body?.error?.message || "Unknown error"
2240
+ }
2241
+ });
2242
+ }
2243
+ }
2244
+ }
2245
+ if (batchData.error_file_id) {
2246
+ const errorRes = await apiRequest(`/files/${batchData.error_file_id}/content`);
2247
+ const errorText = await errorRes.text();
2248
+ for (const line of errorText.trim().split("\n")) {
2249
+ if (!line) continue;
2250
+ const item = JSON.parse(line);
2251
+ const existing = results.find((r) => r.custom_id === item.custom_id);
2252
+ if (!existing) {
2253
+ results.push({
2254
+ custom_id: item.custom_id,
2255
+ status: "error",
2256
+ response: null,
2257
+ error: {
2258
+ code: item.response?.status_code || 500,
2259
+ message: item.error?.message || "Batch item error"
2260
+ }
2261
+ });
2262
+ }
2263
+ }
2264
+ }
2265
+ return results;
2266
+ },
2267
+ async cancelBatch(providerBatchId) {
2268
+ await apiRequest(`/batches/${providerBatchId}/cancel`, { method: "POST" });
2269
+ }
2270
+ };
2271
+ }
2272
+
2273
+ // src/providers/anthropic-batch.ts
2274
+ var ANTHROPIC_API_BASE2 = "https://api.anthropic.com/v1";
2275
+ var ANTHROPIC_VERSION2 = "2023-06-01";
2276
+ var DEFAULT_MAX_TOKENS2 = 4096;
2277
+ function createAnthropicBatchAdapter(apiKey) {
2278
+ async function apiRequest(path2, options = {}) {
2279
+ const headers = {
2280
+ "x-api-key": apiKey,
2281
+ "anthropic-version": ANTHROPIC_VERSION2,
2282
+ "Content-Type": "application/json"
2283
+ };
2284
+ const res = await fetch(`${ANTHROPIC_API_BASE2}${path2}`, {
2285
+ method: options.method || "GET",
2286
+ headers,
2287
+ body: options.body ? JSON.stringify(options.body) : void 0
2288
+ });
2289
+ if (!res.ok) {
2290
+ let errorBody;
2291
+ try {
2292
+ errorBody = await res.json();
2293
+ } catch {
2294
+ errorBody = { message: res.statusText };
2295
+ }
2296
+ const msg = errorBody?.error?.message || errorBody?.message || res.statusText;
2297
+ throw new AnyModelError(res.status >= 500 ? 502 : res.status, msg, {
2298
+ provider_name: "anthropic",
2299
+ raw: errorBody
2300
+ });
2301
+ }
2302
+ return res;
2303
+ }
2304
+ function translateToAnthropicParams(model, req) {
2305
+ const params = {
2306
+ model,
2307
+ max_tokens: req.max_tokens || DEFAULT_MAX_TOKENS2
2308
+ };
2309
+ const systemMessages = req.messages.filter((m) => m.role === "system");
2310
+ const nonSystemMessages = req.messages.filter((m) => m.role !== "system");
2311
+ if (systemMessages.length > 0) {
2312
+ params.system = systemMessages.map((m) => typeof m.content === "string" ? m.content : "").join("\n");
2313
+ }
2314
+ params.messages = nonSystemMessages.map((m) => ({
2315
+ role: m.role === "tool" ? "user" : m.role,
2316
+ content: m.tool_call_id ? [{ type: "tool_result", tool_use_id: m.tool_call_id, content: typeof m.content === "string" ? m.content : "" }] : m.content
2317
+ }));
2318
+ if (req.temperature !== void 0) params.temperature = req.temperature;
2319
+ if (req.top_p !== void 0) params.top_p = req.top_p;
2320
+ if (req.top_k !== void 0) params.top_k = req.top_k;
2321
+ if (req.stop !== void 0) params.stop_sequences = Array.isArray(req.stop) ? req.stop : [req.stop];
2322
+ if (req.tools && req.tools.length > 0) {
2323
+ params.tools = req.tools.map((t) => ({
2324
+ name: t.function.name,
2325
+ description: t.function.description || "",
2326
+ input_schema: t.function.parameters || { type: "object", properties: {} }
2327
+ }));
2328
+ if (req.tool_choice) {
2329
+ if (req.tool_choice === "auto") {
2330
+ params.tool_choice = { type: "auto" };
2331
+ } else if (req.tool_choice === "required") {
2332
+ params.tool_choice = { type: "any" };
2333
+ } else if (req.tool_choice === "none") {
2334
+ delete params.tools;
2335
+ } else if (typeof req.tool_choice === "object") {
2336
+ params.tool_choice = { type: "tool", name: req.tool_choice.function.name };
2337
+ }
2338
+ }
2339
+ }
2340
+ if (req.response_format) {
2341
+ if (req.response_format.type === "json_object" || req.response_format.type === "json_schema") {
2342
+ const jsonInstruction = "Respond with valid JSON only. Do not include any text outside the JSON object.";
2343
+ params.system = params.system ? `${jsonInstruction}
2344
+
2345
+ ${params.system}` : jsonInstruction;
2346
+ }
2347
+ }
2348
+ return params;
2349
+ }
2350
+ function mapStopReason(reason) {
2351
+ switch (reason) {
2352
+ case "end_turn":
2353
+ return "stop";
2354
+ case "max_tokens":
2355
+ return "length";
2356
+ case "tool_use":
2357
+ return "tool_calls";
2358
+ case "stop_sequence":
2359
+ return "stop";
2360
+ default:
2361
+ return "stop";
2362
+ }
2363
+ }
2364
+ function translateAnthropicMessage(msg) {
2365
+ let content = "";
2366
+ const toolCalls = [];
2367
+ for (const block of msg.content || []) {
2368
+ if (block.type === "text") {
2369
+ content += block.text;
2370
+ } else if (block.type === "tool_use") {
2371
+ toolCalls.push({
2372
+ id: block.id,
2373
+ type: "function",
2374
+ function: {
2375
+ name: block.name,
2376
+ arguments: JSON.stringify(block.input)
2377
+ }
2378
+ });
2379
+ }
2380
+ }
2381
+ const message = { role: "assistant", content };
2382
+ if (toolCalls.length > 0) {
2383
+ message.tool_calls = toolCalls;
2384
+ }
2385
+ return {
2386
+ id: generateId(),
2387
+ object: "chat.completion",
2388
+ created: Math.floor(Date.now() / 1e3),
2389
+ model: `anthropic/${msg.model}`,
2390
+ choices: [{
2391
+ index: 0,
2392
+ message,
2393
+ finish_reason: mapStopReason(msg.stop_reason)
2394
+ }],
2395
+ usage: {
2396
+ prompt_tokens: msg.usage?.input_tokens || 0,
2397
+ completion_tokens: msg.usage?.output_tokens || 0,
2398
+ total_tokens: (msg.usage?.input_tokens || 0) + (msg.usage?.output_tokens || 0)
2399
+ }
2400
+ };
2401
+ }
2402
+ return {
2403
+ async createBatch(model, requests, _options) {
2404
+ const batchRequests = requests.map((req) => ({
2405
+ custom_id: req.custom_id,
2406
+ params: translateToAnthropicParams(model, req)
2407
+ }));
2408
+ const res = await apiRequest("/messages/batches", {
2409
+ method: "POST",
2410
+ body: { requests: batchRequests }
2411
+ });
2412
+ const data = await res.json();
2413
+ return {
2414
+ providerBatchId: data.id,
2415
+ metadata: {
2416
+ anthropic_type: data.type,
2417
+ created_at: data.created_at
2418
+ }
2419
+ };
2420
+ },
2421
+ async pollBatch(providerBatchId) {
2422
+ const res = await apiRequest(`/messages/batches/${providerBatchId}`);
2423
+ const data = await res.json();
2424
+ const counts = data.request_counts || {};
2425
+ const total = (counts.processing || 0) + (counts.succeeded || 0) + (counts.errored || 0) + (counts.canceled || 0) + (counts.expired || 0);
2426
+ let status;
2427
+ if (data.processing_status === "ended") {
2428
+ if (counts.succeeded === 0 && (counts.errored > 0 || counts.expired > 0 || counts.canceled > 0)) {
2429
+ status = "failed";
2430
+ } else if (data.cancel_initiated_at) {
2431
+ status = "cancelled";
2432
+ } else {
2433
+ status = "completed";
2434
+ }
2435
+ } else {
2436
+ status = "processing";
2437
+ }
2438
+ return {
2439
+ status,
2440
+ total,
2441
+ completed: counts.succeeded || 0,
2442
+ failed: (counts.errored || 0) + (counts.expired || 0) + (counts.canceled || 0)
2443
+ };
2444
+ },
2445
+ async getBatchResults(providerBatchId) {
2446
+ const res = await apiRequest(`/messages/batches/${providerBatchId}/results`);
2447
+ const text = await res.text();
2448
+ const results = [];
2449
+ for (const line of text.trim().split("\n")) {
2450
+ if (!line) continue;
2451
+ const item = JSON.parse(line);
2452
+ if (item.result?.type === "succeeded") {
2453
+ results.push({
2454
+ custom_id: item.custom_id,
2455
+ status: "success",
2456
+ response: translateAnthropicMessage(item.result.message),
2457
+ error: null
2458
+ });
2459
+ } else {
2460
+ const errorType = item.result?.type || "unknown";
2461
+ const errorMsg = item.result?.error?.message || `Batch item ${errorType}`;
2462
+ results.push({
2463
+ custom_id: item.custom_id,
2464
+ status: "error",
2465
+ response: null,
2466
+ error: {
2467
+ code: errorType === "expired" ? 408 : 500,
2468
+ message: errorMsg
2469
+ }
2470
+ });
2471
+ }
2472
+ }
2473
+ return results;
2474
+ },
2475
+ async cancelBatch(providerBatchId) {
2476
+ await apiRequest(`/messages/batches/${providerBatchId}/cancel`, { method: "POST" });
2477
+ }
2478
+ };
2479
+ }
2480
+
1778
2481
  // src/client.ts
1779
2482
  var AnyModel = class {
1780
2483
  registry;
@@ -1790,6 +2493,9 @@ var AnyModel = class {
1790
2493
  constructor(config = {}) {
1791
2494
  this.config = resolveConfig(config);
1792
2495
  this.registry = new ProviderRegistry();
2496
+ if (this.config.io) {
2497
+ configureFsIO(this.config.io);
2498
+ }
1793
2499
  this.registerProviders();
1794
2500
  this.router = new Router(this.registry, this.config.aliases, this.config);
1795
2501
  this.chat = {
@@ -1839,8 +2545,10 @@ var AnyModel = class {
1839
2545
  };
1840
2546
  this.batchManager = new BatchManager(this.router, {
1841
2547
  dir: this.config.batch?.dir,
1842
- concurrency: this.config.batch?.concurrencyFallback
2548
+ concurrency: this.config.batch?.concurrencyFallback,
2549
+ pollInterval: this.config.batch?.pollInterval
1843
2550
  });
2551
+ this.registerBatchAdapters();
1844
2552
  this.batches = {
1845
2553
  create: (request) => this.batchManager.create(request),
1846
2554
  createAndPoll: (request, options) => this.batchManager.createAndPoll(request, options),
@@ -1892,6 +2600,17 @@ var AnyModel = class {
1892
2600
  }
1893
2601
  }
1894
2602
  }
2603
+ registerBatchAdapters() {
2604
+ const config = this.config;
2605
+ const openaiKey = config.openai?.apiKey || process.env.OPENAI_API_KEY;
2606
+ if (openaiKey) {
2607
+ this.batchManager.registerBatchAdapter("openai", createOpenAIBatchAdapter(openaiKey));
2608
+ }
2609
+ const anthropicKey = config.anthropic?.apiKey || process.env.ANTHROPIC_API_KEY;
2610
+ if (anthropicKey) {
2611
+ this.batchManager.registerBatchAdapter("anthropic", createAnthropicBatchAdapter(anthropicKey));
2612
+ }
2613
+ }
1895
2614
  applyDefaults(request) {
1896
2615
  const defaults = this.config.defaults;
1897
2616
  if (!defaults) return request;
@@ -1919,10 +2638,10 @@ var AnyModel = class {
1919
2638
 
1920
2639
  // src/server.ts
1921
2640
  function parseBody(req) {
1922
- return new Promise((resolve3, reject) => {
2641
+ return new Promise((resolve2, reject) => {
1923
2642
  const chunks = [];
1924
2643
  req.on("data", (chunk) => chunks.push(chunk));
1925
- req.on("end", () => resolve3(Buffer.concat(chunks).toString()));
2644
+ req.on("end", () => resolve2(Buffer.concat(chunks).toString()));
1926
2645
  req.on("error", reject);
1927
2646
  });
1928
2647
  }
@@ -1952,7 +2671,7 @@ function createAnyModelServer(options = {}) {
1952
2671
  const basePath = "/api/v1";
1953
2672
  const server = (0, import_node_http.createServer)(async (req, res) => {
1954
2673
  const url = new URL(req.url || "/", `http://${req.headers.host}`);
1955
- const path = url.pathname;
2674
+ const path2 = url.pathname;
1956
2675
  res.setHeader("Access-Control-Allow-Origin", "*");
1957
2676
  res.setHeader("Access-Control-Allow-Methods", "GET, POST, OPTIONS");
1958
2677
  res.setHeader("Access-Control-Allow-Headers", "Content-Type, Authorization");
@@ -1962,11 +2681,11 @@ function createAnyModelServer(options = {}) {
1962
2681
  return;
1963
2682
  }
1964
2683
  try {
1965
- if (path === "/health" && req.method === "GET") {
2684
+ if (path2 === "/health" && req.method === "GET") {
1966
2685
  sendJSON(res, 200, { status: "ok" });
1967
2686
  return;
1968
2687
  }
1969
- if (path === `${basePath}/chat/completions` && req.method === "POST") {
2688
+ if (path2 === `${basePath}/chat/completions` && req.method === "POST") {
1970
2689
  const body = JSON.parse(await parseBody(req));
1971
2690
  if (body.stream) {
1972
2691
  const stream = await client.chat.completions.create(body);
@@ -1977,14 +2696,14 @@ function createAnyModelServer(options = {}) {
1977
2696
  }
1978
2697
  return;
1979
2698
  }
1980
- if (path === `${basePath}/models` && req.method === "GET") {
2699
+ if (path2 === `${basePath}/models` && req.method === "GET") {
1981
2700
  const provider = url.searchParams.get("provider") || void 0;
1982
2701
  const models = await client.models.list({ provider });
1983
2702
  sendJSON(res, 200, { object: "list", data: models });
1984
2703
  return;
1985
2704
  }
1986
- if (path.startsWith(`${basePath}/generation/`) && req.method === "GET") {
1987
- const id = path.substring(`${basePath}/generation/`.length);
2705
+ if (path2.startsWith(`${basePath}/generation/`) && req.method === "GET") {
2706
+ const id = path2.substring(`${basePath}/generation/`.length);
1988
2707
  const stats = client.generation.get(id);
1989
2708
  if (!stats) {
1990
2709
  sendError(res, 404, `Generation ${id} not found`);
@@ -1993,26 +2712,26 @@ function createAnyModelServer(options = {}) {
1993
2712
  sendJSON(res, 200, stats);
1994
2713
  return;
1995
2714
  }
1996
- if (path === `${basePath}/batches` && req.method === "POST") {
2715
+ if (path2 === `${basePath}/batches` && req.method === "POST") {
1997
2716
  const body = JSON.parse(await parseBody(req));
1998
2717
  const batch = await client.batches.create(body);
1999
2718
  sendJSON(res, 201, batch);
2000
2719
  return;
2001
2720
  }
2002
- if (path === `${basePath}/batches` && req.method === "GET") {
2003
- const batches = client.batches.list();
2721
+ if (path2 === `${basePath}/batches` && req.method === "GET") {
2722
+ const batches = await client.batches.list();
2004
2723
  sendJSON(res, 200, { object: "list", data: batches });
2005
2724
  return;
2006
2725
  }
2007
- if (path.startsWith(`${basePath}/batches/`) && req.method === "GET") {
2008
- const parts = path.substring(`${basePath}/batches/`.length).split("/");
2726
+ if (path2.startsWith(`${basePath}/batches/`) && req.method === "GET") {
2727
+ const parts = path2.substring(`${basePath}/batches/`.length).split("/");
2009
2728
  const id = parts[0];
2010
2729
  if (parts[1] === "results") {
2011
- const results = client.batches.results(id);
2730
+ const results = await client.batches.results(id);
2012
2731
  sendJSON(res, 200, results);
2013
2732
  return;
2014
2733
  }
2015
- const batch = client.batches.get(id);
2734
+ const batch = await client.batches.get(id);
2016
2735
  if (!batch) {
2017
2736
  sendError(res, 404, `Batch ${id} not found`);
2018
2737
  return;
@@ -2020,16 +2739,16 @@ function createAnyModelServer(options = {}) {
2020
2739
  sendJSON(res, 200, batch);
2021
2740
  return;
2022
2741
  }
2023
- if (path.startsWith(`${basePath}/batches/`) && req.method === "POST") {
2024
- const parts = path.substring(`${basePath}/batches/`.length).split("/");
2742
+ if (path2.startsWith(`${basePath}/batches/`) && req.method === "POST") {
2743
+ const parts = path2.substring(`${basePath}/batches/`.length).split("/");
2025
2744
  const id = parts[0];
2026
2745
  if (parts[1] === "cancel") {
2027
- const batch = client.batches.cancel(id);
2746
+ const batch = await client.batches.cancel(id);
2028
2747
  sendJSON(res, 200, batch);
2029
2748
  return;
2030
2749
  }
2031
2750
  }
2032
- sendError(res, 404, `Not found: ${path}`);
2751
+ sendError(res, 404, `Not found: ${path2}`);
2033
2752
  } catch (err) {
2034
2753
  const code = err?.code || 500;
2035
2754
  const message = err?.message || "Internal server error";