phantomllm 0.3.0 → 1.0.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/dist/index.cjs CHANGED
@@ -1,40 +1,13 @@
1
1
  'use strict';
2
2
 
3
- var testcontainers = require('testcontainers');
3
+ var Fastify = require('fastify');
4
+ var fp = require('fastify-plugin');
5
+ var crypto = require('crypto');
4
6
 
5
- // src/driver/container.config.ts
6
- var DEFAULT_IMAGE = "phantomllm-server:latest";
7
- var DEFAULT_PORT = 8080;
8
- function resolveContainerConfig(options) {
9
- return {
10
- image: options?.image ?? process.env["PHANTOMLLM_IMAGE"] ?? DEFAULT_IMAGE,
11
- containerPort: options?.containerPort ?? DEFAULT_PORT,
12
- reuse: options?.reuse ?? true,
13
- startupTimeout: options?.startupTimeout ?? 3e4
14
- };
15
- }
16
- var ContainerManager = class {
17
- constructor(config) {
18
- this.config = config;
19
- }
20
- container = null;
21
- async start() {
22
- const builder = new testcontainers.GenericContainer(this.config.image).withExposedPorts(this.config.containerPort).withWaitStrategy(
23
- testcontainers.Wait.forHttp("/_admin/health", this.config.containerPort).forStatusCode(200)
24
- ).withStartupTimeout(this.config.startupTimeout).withLabels({ "com.phantomllm": "true" });
25
- this.container = await builder.start();
26
- return {
27
- host: this.container.getHost(),
28
- port: this.container.getMappedPort(this.config.containerPort)
29
- };
30
- }
31
- async stop() {
32
- if (this.container) {
33
- await this.container.stop();
34
- this.container = null;
35
- }
36
- }
37
- };
7
+ function _interopDefault (e) { return e && e.__esModule ? e : { default: e }; }
8
+
9
+ var Fastify__default = /*#__PURE__*/_interopDefault(Fastify);
10
+ var fp__default = /*#__PURE__*/_interopDefault(fp);
38
11
 
39
12
  // src/errors/base.ts
40
13
  var MockLLMError = class extends Error {
@@ -274,19 +247,483 @@ var ContainerNotStartedError = class extends MockLLMError {
274
247
  }
275
248
  };
276
249
 
250
+ // src/server/stubs/stub.matcher.ts
251
+ function matchModel(matcher, model) {
252
+ if (matcher.model === void 0) return true;
253
+ return matcher.model === model;
254
+ }
255
+ function matchContent(matcher, messages) {
256
+ if (matcher.content === void 0) return true;
257
+ const needle = matcher.content.toLowerCase();
258
+ return messages.some(
259
+ (m) => m.role === "user" && m.content !== null && m.content.toLowerCase().includes(needle)
260
+ );
261
+ }
262
+ function matchInput(matcher, input) {
263
+ if (matcher.input === void 0) return true;
264
+ const needle = matcher.input.toLowerCase();
265
+ const items = Array.isArray(input) ? input : [input];
266
+ return items.some((item) => item.toLowerCase().includes(needle));
267
+ }
268
+ function stubMatches(entry, endpoint, model, messages, input) {
269
+ const { matcher } = entry;
270
+ if (matcher.endpoint !== void 0 && matcher.endpoint !== endpoint) {
271
+ return false;
272
+ }
273
+ if (!matchModel(matcher, model)) return false;
274
+ if (messages !== void 0 && !matchContent(matcher, messages)) return false;
275
+ if (input !== void 0 && !matchInput(matcher, input)) return false;
276
+ return true;
277
+ }
278
+ function specificity(matcher) {
279
+ let count = 0;
280
+ if (matcher.model !== void 0) count++;
281
+ if (matcher.content !== void 0) count++;
282
+ if (matcher.input !== void 0) count++;
283
+ return count;
284
+ }
285
+
286
+ // src/server/stubs/stub.registry.ts
287
+ var StubRegistry = class {
288
+ stubs = [];
289
+ requests = [];
290
+ register(matcher, response, delay2) {
291
+ const entry = {
292
+ id: crypto.randomUUID(),
293
+ createdAt: Date.now(),
294
+ matcher,
295
+ response,
296
+ delay: delay2 ?? 0,
297
+ callCount: 0
298
+ };
299
+ this.stubs.push(entry);
300
+ return entry;
301
+ }
302
+ findMatch(endpoint, model, messages, input) {
303
+ const matches = this.stubs.filter(
304
+ (s) => stubMatches(s, endpoint, model, messages, input)
305
+ );
306
+ if (matches.length === 0) return void 0;
307
+ matches.sort((a, b) => {
308
+ const specDiff = specificity(b.matcher) - specificity(a.matcher);
309
+ if (specDiff !== 0) return specDiff;
310
+ return a.createdAt - b.createdAt;
311
+ });
312
+ const match = matches[0];
313
+ match.callCount++;
314
+ return match;
315
+ }
316
+ clear() {
317
+ const count = this.stubs.length;
318
+ this.stubs = [];
319
+ return count;
320
+ }
321
+ getAll() {
322
+ return this.stubs;
323
+ }
324
+ recordRequest(request) {
325
+ this.requests.push(request);
326
+ }
327
+ getRequests() {
328
+ return this.requests;
329
+ }
330
+ clearRequests() {
331
+ this.requests = [];
332
+ }
333
+ };
334
+
335
+ // src/server/plugins/registry.plugin.ts
336
+ var registryPlugin = async (fastify) => {
337
+ const registry = new StubRegistry();
338
+ fastify.decorate("stubRegistry", registry);
339
+ };
340
+ var stubRegistryPlugin = fp__default.default(registryPlugin, {
341
+ name: "stub-registry"
342
+ });
343
+ var authPlugin = async (fastify) => {
344
+ const config = {
345
+ apiKey: process.env["PHANTOMLLM_API_KEY"] || void 0
346
+ };
347
+ fastify.decorate("authConfig", config);
348
+ fastify.addHook("onRequest", async (request, reply) => {
349
+ if (request.url.startsWith("/_admin")) return;
350
+ if (!fastify.authConfig.apiKey) return;
351
+ const header = request.headers.authorization;
352
+ if (!header) {
353
+ return reply.status(401).send({
354
+ error: {
355
+ message: "Missing Authorization header. Expected: Bearer <api-key>",
356
+ type: "authentication_error",
357
+ param: null,
358
+ code: "missing_api_key"
359
+ }
360
+ });
361
+ }
362
+ const token = header.replace(/^Bearer\s+/i, "");
363
+ if (token !== fastify.authConfig.apiKey) {
364
+ return reply.status(401).send({
365
+ error: {
366
+ message: "Invalid API key provided.",
367
+ type: "authentication_error",
368
+ param: null,
369
+ code: "invalid_api_key"
370
+ }
371
+ });
372
+ }
373
+ });
374
+ };
375
+ var apiKeyAuthPlugin = fp__default.default(authPlugin, { name: "api-key-auth" });
376
+
377
+ // src/server/admin/admin.routes.ts
378
+ var adminRoutes = async (fastify) => {
379
+ fastify.post(
380
+ "/_admin/stubs",
381
+ async (request, reply) => {
382
+ const { matcher, response, delay: delay2 } = request.body;
383
+ const stub = fastify.stubRegistry.register(matcher, response, delay2);
384
+ return reply.status(201).send({ id: stub.id, stub });
385
+ }
386
+ );
387
+ fastify.post(
388
+ "/_admin/stubs/batch",
389
+ async (request, reply) => {
390
+ const { stubs: definitions } = request.body;
391
+ const created = definitions.map(
392
+ (def) => fastify.stubRegistry.register(def.matcher, def.response, def.delay)
393
+ );
394
+ return reply.status(201).send(created);
395
+ }
396
+ );
397
+ fastify.delete("/_admin/stubs", async (_request, reply) => {
398
+ const count = fastify.stubRegistry.clear();
399
+ const body = { cleared: count };
400
+ return reply.send(body);
401
+ });
402
+ fastify.get("/_admin/health", async (_request, reply) => {
403
+ const body = {
404
+ status: "ok",
405
+ stubCount: fastify.stubRegistry.getAll().length,
406
+ uptime: process.uptime()
407
+ };
408
+ return reply.send(body);
409
+ });
410
+ fastify.get("/_admin/requests", async (_request, reply) => {
411
+ return reply.send({ requests: fastify.stubRegistry.getRequests() });
412
+ });
413
+ fastify.delete("/_admin/requests", async (_request, reply) => {
414
+ fastify.stubRegistry.clearRequests();
415
+ return reply.status(204).send();
416
+ });
417
+ fastify.post(
418
+ "/_admin/config",
419
+ async (request, reply) => {
420
+ const { apiKey } = request.body;
421
+ if (apiKey !== void 0) {
422
+ fastify.authConfig.apiKey = apiKey ?? void 0;
423
+ }
424
+ return reply.send({ ok: true });
425
+ }
426
+ );
427
+ };
428
+ function generateChatCompletionId() {
429
+ return "chatcmpl-" + crypto.randomBytes(18).toString("base64url").slice(0, 24);
430
+ }
431
+
432
+ // src/server/utils/token.counter.ts
433
+ function estimateTokens(text) {
434
+ return Math.max(1, Math.ceil(text.length / 4));
435
+ }
436
+ function estimatePromptTokens(messages) {
437
+ let tokens = 2;
438
+ for (const msg of messages) {
439
+ tokens += 4;
440
+ tokens += estimateTokens(msg.content ?? "");
441
+ }
442
+ return tokens;
443
+ }
444
+
445
+ // src/server/responses/chat.response.ts
446
+ function buildChatCompletion(content, model, messages) {
447
+ const completionTokens = estimateTokens(content);
448
+ const promptTokens = messages ? estimatePromptTokens(messages) : 0;
449
+ return {
450
+ id: generateChatCompletionId(),
451
+ object: "chat.completion",
452
+ created: Math.floor(Date.now() / 1e3),
453
+ model,
454
+ system_fingerprint: "fp_mock",
455
+ choices: [
456
+ {
457
+ index: 0,
458
+ message: { role: "assistant", content },
459
+ finish_reason: "stop",
460
+ logprobs: null
461
+ }
462
+ ],
463
+ usage: {
464
+ prompt_tokens: promptTokens,
465
+ completion_tokens: completionTokens,
466
+ total_tokens: promptTokens + completionTokens
467
+ }
468
+ };
469
+ }
470
+
471
+ // src/server/responses/error.response.ts
472
+ function buildErrorResponse(_status, message, type) {
473
+ return {
474
+ error: {
475
+ message,
476
+ type: type ?? "invalid_request_error",
477
+ param: null,
478
+ code: null
479
+ }
480
+ };
481
+ }
482
+ function buildNoStubMatchResponse(method, path, model, messages) {
483
+ const msgSummary = messages !== void 0 ? ` messages=${JSON.stringify(messages)}` : "";
484
+ return {
485
+ status: 418,
486
+ body: buildErrorResponse(
487
+ 418,
488
+ `No stub matched request: ${method} ${path} model=${model}${msgSummary}`,
489
+ "stub_not_found"
490
+ )
491
+ };
492
+ }
493
+
494
+ // src/server/chat/chat.streaming.ts
495
+ function buildChunk(id, model, content, finishReason, usage) {
496
+ return {
497
+ id,
498
+ object: "chat.completion.chunk",
499
+ created: Math.floor(Date.now() / 1e3),
500
+ model,
501
+ system_fingerprint: "fp_mock",
502
+ choices: [
503
+ {
504
+ index: 0,
505
+ delta: finishReason === null ? { role: "assistant", content } : {},
506
+ finish_reason: finishReason,
507
+ logprobs: null
508
+ }
509
+ ],
510
+ usage: usage ?? null
511
+ };
512
+ }
513
+ function writeSSE(reply, chunk) {
514
+ reply.raw.write(`data: ${JSON.stringify(chunk)}
515
+
516
+ `);
517
+ }
518
+ function streamChunks(reply, model, chunks, messages, includeUsage) {
519
+ reply.raw.writeHead(200, {
520
+ "content-type": "text/event-stream",
521
+ "cache-control": "no-cache",
522
+ connection: "keep-alive"
523
+ });
524
+ reply.hijack();
525
+ const id = generateChatCompletionId();
526
+ for (const word of chunks) {
527
+ writeSSE(reply, buildChunk(id, model, word, null));
528
+ }
529
+ const fullContent = chunks.join("");
530
+ const completionTokens = estimateTokens(fullContent);
531
+ const promptTokens = estimatePromptTokens(messages);
532
+ const usage = includeUsage ? {
533
+ prompt_tokens: promptTokens,
534
+ completion_tokens: completionTokens,
535
+ total_tokens: promptTokens + completionTokens
536
+ } : void 0;
537
+ writeSSE(reply, buildChunk(id, model, "", "stop", usage));
538
+ reply.raw.write("data: [DONE]\n\n");
539
+ reply.raw.end();
540
+ }
541
+
542
+ // src/server/chat/chat.handler.ts
543
+ function delay(ms) {
544
+ return new Promise((resolve) => {
545
+ setTimeout(resolve, ms);
546
+ });
547
+ }
548
+ async function handleChatCompletion(request, reply) {
549
+ const { model, messages, stream, stream_options } = request.body;
550
+ const registry = request.server.stubRegistry;
551
+ const recorded = {
552
+ timestamp: Date.now(),
553
+ method: request.method,
554
+ path: request.url,
555
+ headers: request.headers,
556
+ body: request.body
557
+ };
558
+ registry.recordRequest(recorded);
559
+ const stub = registry.findMatch("chat", model, messages);
560
+ if (!stub) {
561
+ const noMatch = buildNoStubMatchResponse(
562
+ request.method,
563
+ request.url,
564
+ model,
565
+ messages
566
+ );
567
+ return reply.status(noMatch.status).send(noMatch.body);
568
+ }
569
+ if (stub.response.type === "error") {
570
+ return reply.status(stub.response.status).send({
571
+ error: {
572
+ message: stub.response.error.message,
573
+ type: stub.response.error.type,
574
+ param: null,
575
+ code: stub.response.error.code
576
+ }
577
+ });
578
+ }
579
+ if (stub.delay > 0) {
580
+ await delay(stub.delay);
581
+ }
582
+ if (stream === true) {
583
+ const includeUsage = stream_options?.include_usage === true;
584
+ const chunks = stub.response.type === "streaming-chat" ? stub.response.chunks : stub.response.type === "chat" ? stub.response.body.split(/(\s+)/).filter((s) => s.length > 0) : [];
585
+ streamChunks(reply, model, chunks, messages, includeUsage);
586
+ return;
587
+ }
588
+ if (stub.response.type === "chat") {
589
+ return reply.send(buildChatCompletion(stub.response.body, model, messages));
590
+ }
591
+ if (stub.response.type === "streaming-chat") {
592
+ const fullContent = stub.response.chunks.join("");
593
+ return reply.send(buildChatCompletion(fullContent, model, messages));
594
+ }
595
+ }
596
+
597
+ // src/server/chat/chat.routes.ts
598
+ var chatRoutes = async (fastify) => {
599
+ fastify.post("/chat/completions", handleChatCompletion);
600
+ };
601
+
602
+ // src/server/responses/embedding.response.ts
603
+ function buildEmbeddingResponse(vectors, model, input) {
604
+ const inputs = Array.isArray(input) ? input : [input];
605
+ const promptTokens = inputs.reduce(
606
+ (sum, text) => sum + estimateTokens(text),
607
+ 0
608
+ );
609
+ return {
610
+ object: "list",
611
+ data: vectors.map((embedding, index) => ({
612
+ object: "embedding",
613
+ index,
614
+ embedding
615
+ })),
616
+ model,
617
+ usage: {
618
+ prompt_tokens: promptTokens,
619
+ total_tokens: promptTokens
620
+ }
621
+ };
622
+ }
623
+
624
+ // src/server/embeddings/embeddings.handler.ts
625
+ var DEFAULT_DIMENSION = 1536;
626
+ function zeroVector(dim) {
627
+ return new Array(dim).fill(0);
628
+ }
629
+ async function handleEmbeddings(request, reply) {
630
+ const { model, input } = request.body;
631
+ const registry = request.server.stubRegistry;
632
+ const recorded = {
633
+ timestamp: Date.now(),
634
+ method: request.method,
635
+ path: request.url,
636
+ headers: request.headers,
637
+ body: request.body
638
+ };
639
+ registry.recordRequest(recorded);
640
+ const stub = registry.findMatch("embeddings", model, void 0, input);
641
+ if (stub && stub.response.type === "error") {
642
+ return reply.status(stub.response.status).send({
643
+ error: {
644
+ message: stub.response.error.message,
645
+ type: stub.response.error.type,
646
+ param: null,
647
+ code: stub.response.error.code
648
+ }
649
+ });
650
+ }
651
+ if (stub && stub.response.type === "embedding") {
652
+ return reply.send(
653
+ buildEmbeddingResponse(stub.response.vectors, model, input)
654
+ );
655
+ }
656
+ const inputs = Array.isArray(input) ? input : [input];
657
+ const vectors = inputs.map(() => zeroVector(DEFAULT_DIMENSION));
658
+ return reply.send(buildEmbeddingResponse(vectors, model, input));
659
+ }
660
+
661
+ // src/server/embeddings/embeddings.routes.ts
662
+ var embeddingsRoutes = async (fastify) => {
663
+ fastify.post("/embeddings", handleEmbeddings);
664
+ };
665
+
666
+ // src/server/models/models.routes.ts
667
+ var DEFAULT_MODELS = [
668
+ { id: "gpt-4", object: "model", created: 1687882411, owned_by: "openai" },
669
+ { id: "gpt-4o", object: "model", created: 1715367049, owned_by: "openai" },
670
+ {
671
+ id: "gpt-3.5-turbo",
672
+ object: "model",
673
+ created: 1677610602,
674
+ owned_by: "openai"
675
+ },
676
+ {
677
+ id: "text-embedding-3-small",
678
+ object: "model",
679
+ created: 1705948997,
680
+ owned_by: "openai"
681
+ }
682
+ ];
683
+ var modelsRoutes = async (fastify) => {
684
+ fastify.get("/models", async (_request, reply) => {
685
+ const registry = fastify.stubRegistry;
686
+ const allStubs = registry.getAll();
687
+ const modelsStub = allStubs.find((s) => s.response.type === "models");
688
+ if (modelsStub) {
689
+ modelsStub.callCount++;
690
+ const config = modelsStub.response;
691
+ if (config.type !== "models") return;
692
+ const data = config.models.map((m) => ({
693
+ id: m.id,
694
+ object: "model",
695
+ created: Math.floor(Date.now() / 1e3),
696
+ owned_by: m.ownedBy ?? "phantomllm"
697
+ }));
698
+ const body2 = { object: "list", data };
699
+ return reply.send(body2);
700
+ }
701
+ const body = { object: "list", data: DEFAULT_MODELS };
702
+ return reply.send(body);
703
+ });
704
+ };
705
+
706
+ // src/server/app.ts
707
+ async function buildApp(opts = {}) {
708
+ const app = Fastify__default.default(opts);
709
+ await app.register(stubRegistryPlugin);
710
+ await app.register(apiKeyAuthPlugin);
711
+ await app.register(adminRoutes);
712
+ await app.register(chatRoutes, { prefix: "/v1" });
713
+ await app.register(embeddingsRoutes, { prefix: "/v1" });
714
+ await app.register(modelsRoutes, { prefix: "/v1" });
715
+ return app;
716
+ }
717
+
277
718
  // src/driver/mock-llm.ts
278
719
  var MockLLM = class {
279
720
  state = "idle";
280
721
  startPromise = null;
281
- containerManager;
722
+ app = null;
282
723
  adminClient = null;
283
724
  _given = null;
284
725
  _expect = null;
285
726
  _baseUrl = null;
286
- constructor(options) {
287
- const config = resolveContainerConfig(options);
288
- this.containerManager = new ContainerManager(config);
289
- }
290
727
  async start() {
291
728
  if (this.state === "running") return;
292
729
  if (this.state === "starting" && this.startPromise) return this.startPromise;
@@ -295,8 +732,11 @@ var MockLLM = class {
295
732
  return this.startPromise;
296
733
  }
297
734
  async doStart() {
298
- const { host, port } = await this.containerManager.start();
299
- this._baseUrl = `http://${host}:${port}`;
735
+ this.app = await buildApp({ logger: false });
736
+ await this.app.listen({ port: 0, host: "127.0.0.1" });
737
+ const address = this.app.server.address();
738
+ const port = typeof address === "object" && address ? address.port : 0;
739
+ this._baseUrl = `http://127.0.0.1:${port}`;
300
740
  this.adminClient = new AdminClient(this._baseUrl);
301
741
  this._given = new GivenStubs(this.adminClient);
302
742
  this._expect = new ExpectConditions(this.adminClient);
@@ -306,9 +746,12 @@ var MockLLM = class {
306
746
  if (this.state === "stopped" || this.state === "idle") return;
307
747
  this.state = "stopping";
308
748
  try {
309
- await this.containerManager.stop();
749
+ if (this.app) {
750
+ await this.app.close();
751
+ }
310
752
  } finally {
311
753
  this.state = "stopped";
754
+ this.app = null;
312
755
  this.adminClient = null;
313
756
  this._given = null;
314
757
  this._expect = null;