@copilotkit/aimock 1.37.4 → 1.39.0

package/dist/server.d.cts

@@ -20,6 +20,12 @@ interface ServerInstance {
   veoVideoJobs: VeoVideoJobMap;
   grokVideoJobs: GrokVideoJobMap;
 }
+/**
+ * The per-server state a full reset clears. `ServerInstance` structurally
+ * satisfies this, so `LLMock.reset()` and the control-API full-reset route
+ * share a single definition of "everything" instead of two lists that drift.
+ */
+
 interface ServiceFixtures {
   search: SearchFixture[];
   rerank: RerankFixture[];
@@ -29,8 +35,7 @@ declare function createServer(fixtures: Fixture[], options?: MockServerOptions,
   path: string;
   handler: Mountable;
 }>, serviceFixtures?: ServiceFixtures): Promise<ServerInstance>;
-//# sourceMappingURL=server.d.ts.map
-
+/** @internal Config startup uses this to avoid parsing a selected auth source twice. */
 //#endregion
 export { ServerInstance, ServiceFixtures, createServer };
 //# sourceMappingURL=server.d.cts.map

package/dist/server.d.cts.map

@@ -1 +1 @@
-{"version":3,"file":"server.d.cts","names":[],"sources":["../src/server.ts"],"sourcesContent":[],"mappings":";;;;;;;;;;;;UA+FiB,cAAA;UACP,MAAA,CAAK;EADE,OAAA,EAEN,OAFM;EAAc,GAAA,EAAA,MAAA;UACrB,EAGE,eAHG;aACJ,EAGI,aAHJ;qBAEC,EAEW,qBAFX;cACG,EAEC,cAFD;eACQ,EAEN,eAFM;;AAEN,UAq+BA,eAAA,CAr+BA;EAAe,MAAA,EAs+BtB,aAt+BsB,EAAA;EAq+Bf,MAAA,EAEP,aAFsB,EAAA;EAAA,UAAA,EAGlB,iBAHkB,EAAA;;AAEtB,iBAOY,YAAA,CAPZ,QAAA,EAQE,OARF,EAAA,EAAA,OAAA,CAAA,EASE,iBATF,EAAA,MAOV,CAPU,EAUC,KAVD,CAAA;MACI,EAAA,MAAA;EAAiB,OAAA,EASW,SATX;AAM/B,CAAA,CAAA,EAAsB,eAAY,CAAA,EAId,eAJc,CAAA,EAK/B,OAL+B,CAKvB,cALuB,CAAA"}
+{"version":3,"file":"server.d.cts","names":[],"sources":["../src/server.ts"],"sourcesContent":[],"mappings":";;;;;;;;;;;;UAyHiB,cAAA;EAAA,MAAA,EACP,MAAA,CAAK,MADgB;EAAA,OAAA,EAEpB,OAFoB;KACrB,EAAA,MAAK;UACJ,EAEC,eAFD;aAEC,EACG,aADH;qBACG,EACQ,qBADR;cACQ,EACP,cADO;eACP,EACC,eADD;;;AAorChB;;;;;UAAiB,eAAA;UACP;UACA;cACI;;iBAMQ,YAAA,WACV,qBACA,4BACD;;WAA+B;sBACtB,kBACjB,QAAQ"}

package/dist/server.d.ts

@@ -20,6 +20,12 @@ interface ServerInstance {
   veoVideoJobs: VeoVideoJobMap;
   grokVideoJobs: GrokVideoJobMap;
 }
+/**
+ * The per-server state a full reset clears. `ServerInstance` structurally
+ * satisfies this, so `LLMock.reset()` and the control-API full-reset route
+ * share a single definition of "everything" instead of two lists that drift.
+ */
+
 interface ServiceFixtures {
   search: SearchFixture[];
   rerank: RerankFixture[];
@@ -29,8 +35,7 @@ declare function createServer(fixtures: Fixture[], options?: MockServerOptions,
   path: string;
   handler: Mountable;
 }>, serviceFixtures?: ServiceFixtures): Promise<ServerInstance>;
-//# sourceMappingURL=server.d.ts.map
-
+/** @internal Config startup uses this to avoid parsing a selected auth source twice. */
 //#endregion
 export { ServerInstance, ServiceFixtures, createServer };
 //# sourceMappingURL=server.d.ts.map

package/dist/server.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"server.d.ts","names":[],"sources":["../src/server.ts"],"sourcesContent":[],"mappings":";;;;;;;;;;;;UA+FiB,cAAA;UACP,MAAA,CAAK;EADE,OAAA,EAEN,OAFM;EAAc,GAAA,EAAA,MAAA;UACrB,EAGE,eAHG;aACJ,EAGI,aAHJ;qBAEC,EAEW,qBAFX;cACG,EAEC,cAFD;eACQ,EAEN,eAFM;;AAEN,UAq+BA,eAAA,CAr+BA;EAAe,MAAA,EAs+BtB,aAt+BsB,EAAA;EAq+Bf,MAAA,EAEP,aAFsB,EAAA;EAAA,UAAA,EAGlB,iBAHkB,EAAA;;AAEtB,iBAOY,YAAA,CAPZ,QAAA,EAQE,OARF,EAAA,EAAA,OAAA,CAAA,EASE,iBATF,EAAA,MAOV,CAPU,EAUC,KAVD,CAAA;MACI,EAAA,MAAA;EAAiB,OAAA,EASW,SATX;AAM/B,CAAA,CAAA,EAAsB,eAAY,CAAA,EAId,eAJc,CAAA,EAK/B,OAL+B,CAKvB,cALuB,CAAA"}
+{"version":3,"file":"server.d.ts","names":[],"sources":["../src/server.ts"],"sourcesContent":[],"mappings":";;;;;;;;;;;;UAyHiB,cAAA;EAAA,MAAA,EACP,MAAA,CAAK,MADgB;EAAA,OAAA,EAEpB,OAFoB;KACrB,EAAA,MAAK;UACJ,EAEC,eAFD;aAEC,EACG,aADH;qBACG,EACQ,qBADR;cACQ,EACP,cADO;eACP,EACC,eADD;;;AAorChB;;;;;UAAiB,eAAA;UACP;UACA;cACI;;iBAMQ,YAAA,WACV,qBACA,4BACD;;WAA+B;sBACtB,kBACjB,QAAQ"}

package/dist/server.js

@@ -1,10 +1,12 @@
-import { buildContentWithToolCallsChunks, buildContentWithToolCallsCompletion, buildTextChunks, buildTextCompletion, buildToolCallChunks, buildToolCallCompletion, buildUsageChunk, estimatePromptTokens, estimateTokens, extractOverrides, flattenHeaders, getContext, getTestId, isAudioResponse, isContentWithToolCallsResponse, isErrorResponse, isTextResponse, isToolCallResponse, readBody, resolveReasoningForModel, resolveResponse, resolveStrictMode, serializeErrorResponse, strictNoMatchLogLine, strictNoMatchMessage, strictOverrideField } from "./helpers.js";
+import { markAuthenticatedRequest, resolveInboundAuth, validateRequestApiKey, writeApiKeyHttpRejection, writeApiKeyUpgradeRejection } from "./api-key-auth.js";
+import { buildContentWithToolCallsChunks, buildContentWithToolCallsCompletion, buildTextChunks, buildTextCompletion, buildToolCallChunks, buildToolCallCompletion, buildUsageChunk, extractOverrides, flattenHeaders, getContext, getTestId, isAudioResponse, isContentWithToolCallsResponse, isErrorResponse, isTextResponse, isToolCallResponse, readBody, resolveFixtureBlocks, resolveReasoningForModel, resolveResponse, resolveStrictMode, resolveUsage, serializeErrorResponse, strictNoMatchLogLine, strictNoMatchMessage, strictOverrideField } from "./helpers.js";
 import { Logger } from "./logger.js";
 import { Journal } from "./journal.js";
 import { matchFixtureDiagnostic, recordMatchOptions } from "./router.js";
 import { entryToFixture, validateFixtures } from "./fixture-loader.js";
 import { writeErrorResponse, writeSSEStream } from "./sse-writer.js";
 import { createInterruptionSignal } from "./interruption.js";
+import { buildOpenRouterCandidates, handleOpenRouterCredits, handleOpenRouterKey, handleOpenRouterModels, isOpenRouterPath, resolveOpenRouterShaping, serializeOpenRouterError, shapeOpenRouterChunks, shapeOpenRouterCompletion } from "./openrouter-chat.js";
 import { applyChaosAction, evaluateChaos } from "./chaos.js";
 import { proxyAndRecord } from "./recorder.js";
 import { handleResponses } from "./responses.js";
@@ -141,23 +143,31 @@ function handleNotFound(res, message) {
 }
 const CONTROL_PREFIX = "/__aimock";
 /**
-* Perform a full fixtures reset: clear the fixtures array, journal, video/fal
-* generation state, and the interaction/event-id counters, then zero the
-* `aimock_fixtures_loaded` gauge. Shared by `/reset/fixtures` and the
-* deprecated `/reset` alias.
+* Perform a full reset: clear the fixtures array, the journal (entries *and*
+* per-test fixture match-counts, i.e. sequence position), the video and fal.ai
+* job/queue state, and the Gemini interaction/event-id counters, then re-zero
+* the `aimock_fixtures_loaded` gauge.
+*
+* `targets` is `null` when no server is running (an in-process `reset()` before
+* `start()`). The process-global generation state is reset either way, since it
+* is not owned by any one server instance.
+*
+* Shared by `POST /__aimock/reset` (canonical), its deprecated
+* `POST /__aimock/reset/fixtures` alias, and `LLMock.reset()`.
 */
-function performFixturesReset(fixtures, journal, videoStates, openRouterVideoJobs, veoVideoJobs, grokVideoJobs, defaults) {
+function performFullReset(fixtures, targets) {
 	fixtures.length = 0;
-	journal.clear();
-	videoStates.clear();
-	openRouterVideoJobs.clear();
-	veoVideoJobs.clear();
-	grokVideoJobs.clear();
 	falJobs.clear();
 	falQueueStates.clear();
 	resetInteractionCounter();
 	resetEventIdCounter();
-	if (defaults.registry) defaults.registry.setGauge("aimock_fixtures_loaded", {}, fixtures.length);
+	if (!targets) return;
+	targets.journal.clear();
+	targets.videoStates.clear();
+	targets.openRouterVideoJobs.clear();
+	targets.veoVideoJobs.clear();
+	targets.grokVideoJobs.clear();
+	if (targets.defaults.registry) targets.defaults.registry.setGauge("aimock_fixtures_loaded", {}, fixtures.length);
 }
 /**
 * Handle requests under `/__aimock/`. Returns `true` if the request was
@@ -226,8 +236,16 @@ async function handleControlAPI(req, res, pathname, fixtures, journal, videoStat
 		res.end(JSON.stringify({ cleared: true }));
 		return true;
 	}
-	if (subPath === "/reset/fixtures" && req.method === "POST") {
-		performFixturesReset(fixtures, journal, videoStates, openRouterVideoJobs, veoVideoJobs, grokVideoJobs, defaults);
+	const resetTargets = () => ({
+		journal,
+		videoStates,
+		openRouterVideoJobs,
+		veoVideoJobs,
+		grokVideoJobs,
+		defaults
+	});
+	if (subPath === "/reset" && req.method === "POST") {
+		performFullReset(fixtures, resetTargets());
 		res.writeHead(200, { "Content-Type": "application/json" });
 		res.end(JSON.stringify({ reset: true }));
 		return true;
@@ -238,10 +256,10 @@ async function handleControlAPI(req, res, pathname, fixtures, journal, videoStat
 		res.end(JSON.stringify({ reset: true }));
 		return true;
 	}
-	if (subPath === "/reset" && req.method === "POST") {
-		performFixturesReset(fixtures, journal, videoStates, openRouterVideoJobs, veoVideoJobs, grokVideoJobs, defaults);
-		const deprecation = "POST /__aimock/reset is deprecated; use POST /__aimock/reset/fixtures (full reset) or POST /__aimock/reset/journal (journal only)";
-		defaults.logger.warn("POST /__aimock/reset is deprecated; use /__aimock/reset/fixtures or /__aimock/reset/journal");
+	if (subPath === "/reset/fixtures" && req.method === "POST") {
+		performFullReset(fixtures, resetTargets());
+		const deprecation = "POST /__aimock/reset/fixtures is deprecated; use POST /__aimock/reset (full reset) or POST /__aimock/reset/journal (journal only)";
+		defaults.logger.warn("POST /__aimock/reset/fixtures is deprecated; use /__aimock/reset or /__aimock/reset/journal");
 		res.writeHead(200, {
 			"Content-Type": "application/json",
 			Deprecation: "true"
@@ -309,7 +327,7 @@ async function handleControlAPI(req, res, pathname, fixtures, journal, videoStat
 	handleNotFound(res, `Unknown control endpoint: ${pathname}`);
 	return true;
 }
-async function handleCompletions(req, res, fixtures, journal, defaults, modelFallback, providerKey) {
+async function handleCompletions(req, res, fixtures, journal, defaults, modelFallback, providerKey, openRouter = false) {
 	setCorsHeaders(res);
 	let raw;
 	try {
@@ -326,7 +344,7 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 500, JSON.stringify({ error: {
+		writeErrorResponse(res, 500, openRouter ? serializeOpenRouterError(500, `Request body read failed: ${msg}`) : JSON.stringify({ error: {
 			message: `Request body read failed: ${msg}`,
 			type: "server_error"
 		} }));
@@ -348,7 +366,7 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, openRouter ? serializeOpenRouterError(400, `Malformed JSON: ${detail}`) : JSON.stringify({ error: {
 			message: `Malformed JSON: ${detail}`,
 			type: "invalid_request_error",
 			param: null,
@@ -367,7 +385,7 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, openRouter ? serializeOpenRouterError(400, "Missing required parameter: 'messages'") : JSON.stringify({ error: {
 			message: "Missing required parameter: 'messages'",
 			type: "invalid_request_error",
 			param: null,
@@ -381,9 +399,66 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 	body._endpointType = "chat";
 	body._context = getContext(req);
 	const testId = getTestId(req);
-	const { fixture, skippedBySequenceOrTurn } = matchFixtureDiagnostic(fixtures, body, journal.getFixtureMatchCountsForTest(testId), defaults.requestTransform, recordMatchOptions(!!(defaults.record && providerKey), defaults.logger));
+	const matchOptions = recordMatchOptions(!!(defaults.record && providerKey), defaults.logger);
+	let fixture = null;
+	let skippedBySequenceOrTurn = 0;
+	let preResolvedResponse = null;
+	let responseModel = body.model;
+	let fixtureCountIncremented = false;
+	if (openRouter && Array.isArray(body.models) && body.models.length > 0) {
+		const candidates = buildOpenRouterCandidates(body);
+		let matchCounts = journal.getFixtureMatchCountsForTest(testId);
+		const allowFallbacks = body.provider?.allow_fallbacks !== false;
+		let lastErrorFixture = null;
+		let lastErrorResponse = null;
+		const resolveCandidate = async (candidate) => {
+			const probe = {
+				...body,
+				model: candidate
+			};
+			const attempt = matchFixtureDiagnostic(fixtures, probe, matchCounts, defaults.requestTransform, matchOptions);
+			skippedBySequenceOrTurn = Math.max(skippedBySequenceOrTurn, attempt.skippedBySequenceOrTurn);
+			if (!attempt.fixture) return null;
+			const response = await resolveResponse(attempt.fixture, probe);
+			return {
+				fixture: attempt.fixture,
+				response,
+				isError: isErrorResponse(response)
+			};
+		};
+		for (const candidate of candidates) {
+			const outcome = await resolveCandidate(candidate);
+			if (!outcome) {
+				if (!allowFallbacks) break;
+				continue;
+			}
+			journal.incrementFixtureMatchCount(outcome.fixture, fixtures, testId);
+			fixtureCountIncremented = true;
+			matchCounts = journal.getFixtureMatchCountsForTest(testId);
+			responseModel = candidate;
+			if (outcome.isError) {
+				const errorFallsThrough = outcome.response.fallthrough !== false;
+				if (allowFallbacks && errorFallsThrough) {
+					lastErrorFixture = outcome.fixture;
+					lastErrorResponse = outcome.response;
+					continue;
+				}
+			}
+			fixture = outcome.fixture;
+			preResolvedResponse = outcome.response;
+			break;
+		}
+		if (!fixture && lastErrorFixture) {
+			fixture = lastErrorFixture;
+			preResolvedResponse = lastErrorResponse;
+		}
+	} else {
+		const single = matchFixtureDiagnostic(fixtures, body, journal.getFixtureMatchCountsForTest(testId), defaults.requestTransform, matchOptions);
+		fixture = single.fixture;
+		skippedBySequenceOrTurn = single.skippedBySequenceOrTurn;
+	}
 	if (fixture) {
-		journal.incrementFixtureMatchCount(fixture, fixtures, testId);
+		if (!fixtureCountIncremented) journal.incrementFixtureMatchCount(fixture, fixtures, testId);
 		defaults.logger.debug(`Fixture matched: ${JSON.stringify(fixture.match).slice(0, 120)}`);
 	} else {
 		const lastUserMsg = body.messages.filter((m) => m.role === "user").pop();
@@ -421,7 +496,7 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 					...strictOverrideField(defaults.strict, req.headers)
 				}
 			});
-			writeErrorResponse(res, strictStatus, JSON.stringify({ error: {
+			writeErrorResponse(res, strictStatus, openRouter ? serializeOpenRouterError(strictStatus, strictMessage) : JSON.stringify({ error: {
 				message: strictMessage,
 				type: "invalid_request_error",
 				param: null,
@@ -472,7 +547,7 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				...strictOverrideField(defaults.strict, req.headers)
 			}
 		});
-		writeErrorResponse(res, 404, JSON.stringify({ error: {
+		writeErrorResponse(res, 404, openRouter ? serializeOpenRouterError(404, "No fixture matched") : JSON.stringify({ error: {
 			message: "No fixture matched",
 			type: "invalid_request_error",
 			param: null,
@@ -480,10 +555,23 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 		} }));
 		return;
 	}
-	const response = await resolveResponse(fixture, body);
+	const response = preResolvedResponse ?? await resolveResponse(fixture, body);
 	const latency = fixture.latency ?? defaults.latency;
 	const chunkSize = Math.max(1, fixture.chunkSize ?? defaults.chunkSize);
-	const includeUsage = body.stream === true && body.stream_options?.include_usage === true;
+	const emitStreamingUsage = body.stream === true && body.stream_options?.include_usage === true || openRouter;
+	const streamingPromptText = body.messages.map((m) => typeof m.content === "string" ? m.content : Array.isArray(m.content) ? m.content.map((p) => p.text ?? "").join("") : "").join("");
+	const resolveStreamingUsageTokens = (overrides, completionText) => resolveUsage(overrides, streamingPromptText, completionText);
+	const shapeORCompletion = (completion, overrides) => {
+		if (!openRouter) return completion;
+		return shapeOpenRouterCompletion(completion, resolveOpenRouterShaping(overrides, overrides?.model ?? responseModel), overrides?.id !== void 0);
+	};
+	const shapeORChunks = (chunks, usageChunk, overrides) => {
+		if (!openRouter) return;
+		const shaping = resolveOpenRouterShaping(overrides, overrides?.model ?? responseModel);
+		shapeOpenRouterChunks(chunks, shaping, overrides?.id !== void 0);
+		if (usageChunk) shapeOpenRouterChunks([usageChunk], shaping, overrides?.id !== void 0);
+	};
+	const openRouterProcessing = !!(openRouter && fixture.openRouterProcessing);
 	if (isErrorResponse(response)) {
 		const status = response.status ?? 500;
 		journal.add({
@@ -496,7 +584,7 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				fixture
 			}
 		});
-		writeErrorResponse(res, status, serializeErrorResponse(response), { retryAfter: response.retryAfter });
+		writeErrorResponse(res, status, openRouter ? serializeOpenRouterError(status, response.error.message, response.error.metadata) : serializeErrorResponse(response), { retryAfter: response.retryAfter });
 		return;
 	}
 	if (isAudioResponse(response)) {
@@ -510,7 +598,7 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				fixture
 			}
 		});
-		writeErrorResponse(res, 422, JSON.stringify({ error: {
+		writeErrorResponse(res, 422, openRouter ? serializeOpenRouterError(422, "Audio responses are not supported on the chat completions endpoint. Use Gemini generateContent or a dedicated audio endpoint.") : JSON.stringify({ error: {
 			message: "Audio responses are not supported on the chat completions endpoint. Use Gemini generateContent or a dedicated audio endpoint.",
 			type: "invalid_request_error"
 		} }));
@@ -520,7 +608,9 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 		if (response.webSearches?.length) defaults.logger.warn("webSearches in fixture response are not supported for Chat Completions API — ignoring");
 		const overrides = extractOverrides(response);
 		const effectiveStrict = resolveStrictMode(defaults.strict, req.headers);
-		const effReasoning = resolveReasoningForModel(response.reasoning, body.model, effectiveStrict, defaults.logger);
+		const effReasoning = resolveReasoningForModel(response.reasoning, responseModel, effectiveStrict, defaults.logger);
+		const streaming = body.stream === true;
+		const streamingBlocks = streaming && response.blocks && response.blocks.length > 0 ? resolveFixtureBlocks(response.blocks) : void 0;
 		const journalEntry = journal.add({
 			method: req.method ?? "POST",
 			path: req.url ?? COMPLETIONS_PATH,
@@ -531,22 +621,15 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				fixture
 			}
 		});
-		if (body.stream !== true) {
-			const completion = buildContentWithToolCallsCompletion(response.content ?? "", response.toolCalls ?? [], body.model, effReasoning, overrides, body.messages);
+		if (!streaming) {
+			const completion = buildContentWithToolCallsCompletion(response.content ?? "", response.toolCalls ?? [], responseModel, effReasoning, overrides, body.messages);
 			res.writeHead(200, { "Content-Type": "application/json" });
-			res.end(JSON.stringify(completion));
+			res.end(JSON.stringify(shapeORCompletion(completion, overrides)));
 		} else {
-			const chunks = buildContentWithToolCallsChunks(response.content ?? "", response.toolCalls ?? [], body.model, chunkSize, effReasoning, overrides, response.blocks);
-			const completionText = (response.content ?? "") + (response.toolCalls ?? []).map((tc) => tc.name + tc.arguments).join("");
-			const usageChunk = includeUsage ? buildUsageChunk(chunks[0]?.id ?? "chatcmpl-unknown", overrides?.model ?? body.model, chunks[0]?.created ?? Math.floor(Date.now() / 1e3), overrides?.usage ? {
-				prompt_tokens: overrides.usage.prompt_tokens ?? 0,
-				completion_tokens: overrides.usage.completion_tokens ?? 0,
-				total_tokens: overrides.usage.total_tokens ?? (overrides.usage.prompt_tokens ?? 0) + (overrides.usage.completion_tokens ?? 0)
-			} : {
-				prompt_tokens: estimatePromptTokens(body.messages),
-				completion_tokens: estimateTokens(completionText),
-				total_tokens: estimatePromptTokens(body.messages) + estimateTokens(completionText)
-			}, overrides?.systemFingerprint) : void 0;
+			const chunks = buildContentWithToolCallsChunks(response.content ?? "", response.toolCalls ?? [], responseModel, chunkSize, effReasoning, overrides, streamingBlocks);
+			const completionText = streamingBlocks ? streamingBlocks.map((b) => b.type === "text" ? b.text : b.name + b.arguments).join("") : (response.content ?? "") + (response.toolCalls ?? []).map((tc) => tc.name + tc.arguments).join("");
+			const usageChunk = emitStreamingUsage ? buildUsageChunk(chunks[0]?.id ?? "chatcmpl-unknown", overrides?.model ?? responseModel, chunks[0]?.created ?? Math.floor(Date.now() / 1e3), resolveStreamingUsageTokens(overrides, completionText), overrides?.systemFingerprint) : void 0;
+			shapeORChunks(chunks, usageChunk, overrides);
 			const interruption = createInterruptionSignal(fixture);
 			if (!await writeSSEStream(res, chunks, {
 				latency,
@@ -555,7 +638,8 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				onChunkSent: interruption?.tick,
 				usageChunk,
 				recordedTimings: fixture.recordedTimings,
-				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed
+				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed,
+				openRouterProcessing
 			})) {
 				if (!res.writableEnded) res.destroy();
 				journalEntry.response.interrupted = true;
@@ -569,7 +653,7 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 		if (response.webSearches?.length) defaults.logger.warn("webSearches in fixture response are not supported for Chat Completions API — ignoring");
 		const overrides = extractOverrides(response);
 		const effectiveStrict = resolveStrictMode(defaults.strict, req.headers);
-		const effReasoning = resolveReasoningForModel(response.reasoning, body.model, effectiveStrict, defaults.logger);
+		const effReasoning = resolveReasoningForModel(response.reasoning, responseModel, effectiveStrict, defaults.logger);
 		const journalEntry = journal.add({
 			method: req.method ?? "POST",
 			path: req.url ?? COMPLETIONS_PATH,
@@ -581,20 +665,13 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 			}
 		});
 		if (body.stream !== true) {
-			const completion = buildTextCompletion(response.content, body.model, effReasoning, overrides, body.messages);
+			const completion = buildTextCompletion(response.content, responseModel, effReasoning, overrides, body.messages);
 			res.writeHead(200, { "Content-Type": "application/json" });
-			res.end(JSON.stringify(completion));
+			res.end(JSON.stringify(shapeORCompletion(completion, overrides)));
 		} else {
-			const chunks = buildTextChunks(response.content, body.model, chunkSize, effReasoning, overrides);
-			const usageChunk = includeUsage ? buildUsageChunk(chunks[0]?.id ?? "chatcmpl-unknown", overrides?.model ?? body.model, chunks[0]?.created ?? Math.floor(Date.now() / 1e3), overrides?.usage ? {
-				prompt_tokens: overrides.usage.prompt_tokens ?? 0,
-				completion_tokens: overrides.usage.completion_tokens ?? 0,
-				total_tokens: overrides.usage.total_tokens ?? (overrides.usage.prompt_tokens ?? 0) + (overrides.usage.completion_tokens ?? 0)
-			} : {
-				prompt_tokens: estimatePromptTokens(body.messages),
-				completion_tokens: estimateTokens(response.content),
-				total_tokens: estimatePromptTokens(body.messages) + estimateTokens(response.content)
-			}, overrides?.systemFingerprint) : void 0;
+			const chunks = buildTextChunks(response.content, responseModel, chunkSize, effReasoning, overrides);
+			const usageChunk = emitStreamingUsage ? buildUsageChunk(chunks[0]?.id ?? "chatcmpl-unknown", overrides?.model ?? responseModel, chunks[0]?.created ?? Math.floor(Date.now() / 1e3), resolveStreamingUsageTokens(overrides, response.content), overrides?.systemFingerprint) : void 0;
+			shapeORChunks(chunks, usageChunk, overrides);
 			const interruption = createInterruptionSignal(fixture);
 			if (!await writeSSEStream(res, chunks, {
 				latency,
@@ -603,7 +680,8 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				onChunkSent: interruption?.tick,
 				usageChunk,
 				recordedTimings: fixture.recordedTimings,
-				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed
+				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed,
+				openRouterProcessing
 			})) {
 				if (!res.writableEnded) res.destroy();
 				journalEntry.response.interrupted = true;
@@ -617,7 +695,7 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 		if (response.webSearches?.length) defaults.logger.warn("webSearches in fixture response are not supported for Chat Completions API — ignoring");
 		const overrides = extractOverrides(response);
 		const effectiveStrict = resolveStrictMode(defaults.strict, req.headers);
-		const effReasoning = resolveReasoningForModel(response.reasoning, body.model, effectiveStrict, defaults.logger);
+		const effReasoning = resolveReasoningForModel(response.reasoning, responseModel, effectiveStrict, defaults.logger);
 		const journalEntry = journal.add({
 			method: req.method ?? "POST",
 			path: req.url ?? COMPLETIONS_PATH,
@@ -629,21 +707,14 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 			}
 		});
 		if (body.stream !== true) {
-			const completion = buildToolCallCompletion(response.toolCalls, body.model, effReasoning, overrides, body.messages);
+			const completion = buildToolCallCompletion(response.toolCalls, responseModel, effReasoning, overrides, body.messages);
 			res.writeHead(200, { "Content-Type": "application/json" });
-			res.end(JSON.stringify(completion));
+			res.end(JSON.stringify(shapeORCompletion(completion, overrides)));
 		} else {
-			const chunks = buildToolCallChunks(response.toolCalls, body.model, chunkSize, effReasoning, overrides);
+			const chunks = buildToolCallChunks(response.toolCalls, responseModel, chunkSize, effReasoning, overrides);
 			const completionText = response.toolCalls.map((tc) => tc.name + tc.arguments).join("");
-			const usageChunk = includeUsage ? buildUsageChunk(chunks[0]?.id ?? "chatcmpl-unknown", overrides?.model ?? body.model, chunks[0]?.created ?? Math.floor(Date.now() / 1e3), overrides?.usage ? {
-				prompt_tokens: overrides.usage.prompt_tokens ?? 0,
-				completion_tokens: overrides.usage.completion_tokens ?? 0,
-				total_tokens: overrides.usage.total_tokens ?? (overrides.usage.prompt_tokens ?? 0) + (overrides.usage.completion_tokens ?? 0)
-			} : {
-				prompt_tokens: estimatePromptTokens(body.messages),
-				completion_tokens: estimateTokens(completionText),
-				total_tokens: estimatePromptTokens(body.messages) + estimateTokens(completionText)
-			}, overrides?.systemFingerprint) : void 0;
+			const usageChunk = emitStreamingUsage ? buildUsageChunk(chunks[0]?.id ?? "chatcmpl-unknown", overrides?.model ?? responseModel, chunks[0]?.created ?? Math.floor(Date.now() / 1e3), resolveStreamingUsageTokens(overrides, completionText), overrides?.systemFingerprint) : void 0;
+			shapeORChunks(chunks, usageChunk, overrides);
 			const interruption = createInterruptionSignal(fixture);
 			if (!await writeSSEStream(res, chunks, {
 				latency,
@@ -652,7 +723,8 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 				onChunkSent: interruption?.tick,
 				usageChunk,
 				recordedTimings: fixture.recordedTimings,
-				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed
+				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed,
+				openRouterProcessing
 			})) {
 				if (!res.writableEnded) res.destroy();
 				journalEntry.response.interrupted = true;
@@ -672,12 +744,19 @@ async function handleCompletions(req, res, fixtures, journal, defaults, modelFal
 			fixture
 		}
 	});
-	writeErrorResponse(res, 500, JSON.stringify({ error: {
+	writeErrorResponse(res, 500, openRouter ? serializeOpenRouterError(500, "Fixture response did not match any known type") : JSON.stringify({ error: {
 		message: "Fixture response did not match any known type",
 		type: "server_error"
 	} }));
 }
 async function createServer(fixtures, options, mounts, serviceFixtures) {
+	return createServerWithResolvedAuth(fixtures, options, resolveInboundAuth({
+		value: options?.auth,
+		label: "options.auth"
+	}), mounts, serviceFixtures);
+}
+/** @internal Config startup uses this to avoid parsing a selected auth source twice. */
+async function createServerWithResolvedAuth(fixtures, options, resolvedAuth, mounts, serviceFixtures) {
 	const host = options?.host ?? "127.0.0.1";
 	const port = options?.port ?? 0;
 	const logger = new Logger(options?.logLevel ?? "silent");
@@ -791,13 +870,10 @@ async function createServer(fixtures, options, mounts, serviceFixtures) {
 		});
 	});
 	async function handleHttpRequest(req, res) {
-		if (req.method === "OPTIONS") {
-			handleOptions(res);
-			return;
-		}
 		const startTime = registry ? process.hrtime.bigint() : 0n;
 		const parsedUrl = new URL(req.url ?? "/", `http://${req.headers.host ?? "localhost"}`);
 		let pathname = parsedUrl.pathname;
+		const isOpenRouter = isOpenRouterPath(pathname);
 		if (registry) res.on("finish", () => {
 			try {
 				const normalizedPath = normalizePathLabel(pathname);
@@ -817,6 +893,17 @@ async function createServer(fixtures, options, mounts, serviceFixtures) {
 				defaults.logger.warn("metrics instrumentation error", err);
 			}
 		});
+		if (req.method === "OPTIONS" && (!resolvedAuth.policy.enabled || req.headers.origin !== void 0 && req.headers["access-control-request-method"] !== void 0)) {
+			handleOptions(res);
+			return;
+		}
+		if (!(req.method === "GET" && (pathname === HEALTH_PATH || pathname === READY_PATH || pathname === "/metrics"))) {
+			if (!validateRequestApiKey(req, resolvedAuth.policy).ok) {
+				writeApiKeyHttpRejection(res, setCorsHeaders);
+				return;
+			}
+			markAuthenticatedRequest(req, resolvedAuth.policy);
+		}
 		if (pathname.startsWith(CONTROL_PREFIX)) {
 			await handleControlAPI(req, res, pathname, fixtures, journal, videoStates, openRouterVideoJobs, veoVideoJobs, grokVideoJobs, defaults);
 			return;
@@ -1001,6 +1088,18 @@ async function createServer(fixtures, options, mounts, serviceFixtures) {
 			}
 			return;
 		}
+		if (pathname === "/api/v1/models" && req.method === "GET") {
+			handleOpenRouterModels(req, res, fixtures, journal, defaults, setCorsHeaders);
+			return;
+		}
+		if (pathname === "/api/v1/key" && req.method === "GET") {
+			handleOpenRouterKey(req, res, journal, setCorsHeaders);
+			return;
+		}
+		if (pathname === "/api/v1/credits" && req.method === "GET") {
+			handleOpenRouterCredits(req, res, journal, setCorsHeaders);
+			return;
+		}
 		let azureDeploymentId;
 		const azureMatch = pathname.match(AZURE_DEPLOYMENT_RE);
 		if (azureMatch && req.method === "POST") {
@@ -1726,17 +1825,17 @@ async function createServer(fixtures, options, mounts, serviceFixtures) {
 			handleNotFound(res, "Not found");
 			return;
 		}
-		const completionsProvider = azureDeploymentId ? "azure" : "openai";
+		const completionsProvider = azureDeploymentId ? "azure" : isOpenRouter ? "openrouter" : "openai";
 		try {
-			await handleCompletions(req, res, fixtures, journal, defaults, azureDeploymentId, completionsProvider);
+			await handleCompletions(req, res, fixtures, journal, defaults, azureDeploymentId, completionsProvider, isOpenRouter);
 		} catch (err) {
 			const msg = err instanceof Error ? err.message : "Internal error";
-			if (!res.headersSent) writeErrorResponse(res, 500, JSON.stringify({ error: {
+			if (!res.headersSent) writeErrorResponse(res, 500, isOpenRouter ? serializeOpenRouterError(500, msg) : JSON.stringify({ error: {
 				message: msg,
 				type: "server_error"
 			} }));
 			else if (!res.writableEnded) try {
-				res.write(`data: ${JSON.stringify({ error: {
+				res.write(`data: ${isOpenRouter ? serializeOpenRouterError(500, msg) : JSON.stringify({ error: {
 					message: msg,
 					type: "server_error"
 				} })}\n\n`);
@@ -1757,6 +1856,10 @@ async function createServer(fixtures, options, mounts, serviceFixtures) {
 	async function handleUpgradeRequest(req, socket, head) {
 		const parsedUrl = new URL(req.url ?? "/", `http://${req.headers.host ?? "localhost"}`);
 		let pathname = parsedUrl.pathname;
+		if (!validateRequestApiKey(req, resolvedAuth.policy).ok) {
+			writeApiKeyUpgradeRejection(socket);
+			return;
+		}
 		if (mounts) {
 			for (const { path: mountPath, handler } of mounts) if ((pathname === mountPath || pathname.startsWith(mountPath + "/")) && handler.handleUpgrade) {
 				const subPath = pathname.slice(mountPath.length) || "/";
@@ -1795,10 +1898,12 @@ async function createServer(fixtures, options, mounts, serviceFixtures) {
 			upgradeHeaders: req.headers
 		});
 		else if (pathname === REALTIME_PATH) {
-			const model = parsedUrl.searchParams.get("model") ?? "gpt-realtime-2";
+			const transcriptionIntent = parsedUrl.searchParams.get("intent") === "transcription";
+			const model = transcriptionIntent ? "gpt-transcribe" : parsedUrl.searchParams.get("model") ?? "gpt-realtime-2";
 			handleWebSocketRealtime(ws, fixtures, journal, {
 				...defaults,
 				model,
+				transcriptionIntent,
 				testId: wsTestId,
 				upgradeHeaders: req.headers
 			});
@@ -1847,5 +1952,5 @@ async function createServer(fixtures, options, mounts, serviceFixtures) {
 }
 
 //#endregion
-export { createServer };
+export { createServer, createServerWithResolvedAuth, performFullReset };
 //# sourceMappingURL=server.js.map