@copilotkit/aimock 1.46.0 → 1.47.0

package/dist/ollama.js

@@ -1,14 +1,17 @@
 import { renderError } from "./response/errors.js";
 import { resolveRequestDecisions, strictOverrideField } from "./response/policy.js";
-import { InvalidToolArgumentsError, MAX_EMBEDDING_DIMENSIONS, flattenHeaders, generateDeterministicEmbedding, getContext, getTestId, isCombinedFixtureResponse, isEmbeddingResponse, isErrorResponse, isJSONResponse, isJsonObject, isTextResponse, isToolCallResponse, normalizeEmbeddingInput, requireEmittedFunctionToolCalls, resolveFixtureBlockOutcome, resolveFixtureBlocks, resolveReasoningForModel, resolveResponse, servedToolArgs, strictNoMatchLogLine, strictNoMatchMessage, toolArgsForWire, validateChatMessages, validateEmbeddingDimensions, validateToolsField } from "./helpers.js";
+import { InvalidToolArgumentsError, MAX_EMBEDDING_DIMENSIONS, flattenHeaders, generateDeterministicEmbedding, getContext, getTestId, isCombinedFixtureResponse, isEmbeddingResponse, isErrorResponse, isJSONResponse, isJsonObject, isRawResponse, isTextResponse, isToolCallResponse, normalizeEmbeddingInput, requireEmittedFunctionToolCalls, resolveFixtureBlockOutcome, resolveFixtureBlocks, resolveReasoningForModel, resolveResponse, servedToolArgs, strictNoMatchLogLine, strictNoMatchMessage, toolArgsForWire, validateChatMessages, validateEmbeddingDimensions, validateToolsField } from "./helpers.js";
 import { writeErrorResponse } from "./sse-writer.js";
 import { applyChaosAsync } from "./chaos.js";
 import { planMisbehavior, recordMisbehaviorOutcome } from "./misbehavior.js";
 import { setEmbeddingRequest } from "./embedding-request.js";
 import { matchFixtureDiagnostic, recordMatchOptions } from "./router.js";
 import { proxyAndRecord } from "./recorder.js";
-import { createInterruptionSignal } from "./interruption.js";
-import { writeNDJSONStream } from "./ndjson-writer.js";
+import { misbehaviorPayloadLength, nonStreamingTimingEnabled, servedPayloadLength } from "./response/timing.js";
+import { deliverJson, deliverStream } from "./response/deliver.js";
+import { proxyInterruption, withHangUpDetection } from "./response/proxy.js";
+import { ndjsonCodec } from "./ndjson-writer.js";
+import { serveRawResponse } from "./response/raw.js";
 import { render } from "./response/usage/codecs/ollama.js";
 
 //#region src/ollama.ts
@@ -330,8 +333,10 @@ function prepareOllamaMisbehavior(plan, request, stream, chunkSize, effectiveStr
 			for (const block of outcome.ordered) if (block.type === "text") priorContent += block.text;
 			else if (index++ === plan.target?.index) break;
 		}
+		const chunks = stream ? [...buildOllamaChatTextChunks(priorContent, model, chunkSize).slice(0, -1), { error }] : [{ error }];
 		return {
-			chunks: stream ? [...buildOllamaChatTextChunks(priorContent, model, chunkSize).slice(0, -1), { error }] : [{ error }],
+			build: () => chunks,
+			reasoning: void 0,
 			status: stream ? 200 : 500,
 			summary: {
 				...plan.summary,
@@ -347,64 +352,64 @@ function prepareOllamaMisbehavior(plan, request, stream, chunkSize, effectiveStr
 			arguments: parsed.text
 		};
 	});
-	let chunks;
-	let reasoning = "";
-	if (plan.summary.fault === "reasoning-only") {
-		reasoning = plan.reasoning ?? "";
-		const base = {
-			model,
-			created_at: (/* @__PURE__ */ new Date()).toISOString()
-		};
-		if (stream) {
-			chunks = [];
-			for (let i = 0; i < reasoning.length; i += chunkSize) chunks.push({
+	const reasoning = plan.summary.fault === "reasoning-only" ? plan.reasoning ?? "" : resolveReasoningForModel("reasoning" in response ? response.reasoning : void 0, model, effectiveStrict, logger) ?? "";
+	const completion = content + reasoning + servedToolCalls.map((call) => call.name + call.arguments).join("");
+	const build = () => {
+		let chunks;
+		if (plan.summary.fault === "reasoning-only") {
+			const base = {
+				model,
+				created_at: (/* @__PURE__ */ new Date()).toISOString()
+			};
+			if (stream) {
+				chunks = [];
+				for (let i = 0; i < reasoning.length; i += chunkSize) chunks.push({
+					...base,
+					message: {
+						role: "assistant",
+						content: "",
+						thinking: reasoning.slice(i, i + chunkSize)
+					},
+					done: false
+				});
+				chunks.push({
+					...base,
+					message: {
+						role: "assistant",
+						content: ""
+					},
+					done: true,
+					...DURATION_FIELDS,
+					done_reason: "length"
+				});
+			} else chunks = [{
 				...base,
 				message: {
 					role: "assistant",
 					content: "",
-					thinking: reasoning.slice(i, i + chunkSize)
-				},
-				done: false
-			});
-			chunks.push({
-				...base,
-				message: {
-					role: "assistant",
-					content: ""
+					thinking: reasoning
 				},
 				done: true,
 				...DURATION_FIELDS,
 				done_reason: "length"
-			});
-		} else chunks = [{
-			...base,
-			message: {
-				role: "assistant",
-				content: "",
-				thinking: reasoning
-			},
-			done: true,
-			...DURATION_FIELDS,
-			done_reason: "length"
-		}];
-	} else {
-		reasoning = resolveReasoningForModel("reasoning" in response ? response.reasoning : void 0, model, effectiveStrict, logger) ?? "";
-		if (combined) chunks = stream ? buildOllamaChatContentWithToolCallsChunks(content, calls, model, chunkSize, reasoning, response.blocks) : [buildOllamaChatContentWithToolCallsResponse(content, calls, model, reasoning, response.blocks)];
+			}];
+		} else if (combined) chunks = stream ? buildOllamaChatContentWithToolCallsChunks(content, calls, model, chunkSize, reasoning, response.blocks) : [buildOllamaChatContentWithToolCallsResponse(content, calls, model, reasoning, response.blocks)];
 		else if (isToolCallResponse(response)) chunks = stream ? buildOllamaChatToolCallChunks(calls, model, chunkSize, reasoning) : [buildOllamaChatToolCallResponse(calls, model, reasoning)];
 		else chunks = stream ? buildOllamaChatTextChunks(content, model, chunkSize, reasoning) : [buildOllamaChatTextResponse(content, model, reasoning)];
-	}
-	const completion = content + reasoning + servedToolCalls.map((call) => call.name + call.arguments).join("");
-	const terminal = chunks.length - 1;
-	chunks[terminal] = {
-		...chunks[terminal],
-		...render({
-			kind: "applied-chat-estimate",
-			messages: request.messages,
-			completion
-		})
+		const terminal = chunks.length - 1;
+		chunks[terminal] = {
+			...chunks[terminal],
+			...render({
+				kind: "applied-chat-estimate",
+				messages: request.messages,
+				completion
+			})
+		};
+		return chunks;
 	};
 	return {
-		chunks,
+		build,
+		reasoning,
 		status: 200,
 		summary: {
 			...plan.summary,
@@ -523,7 +528,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 	completionReq._endpointType = "chat";
 	completionReq._context = getContext(req);
 	const testId = getTestId(req);
-	const { fixture, skippedBySequenceOrTurn } = matchFixtureDiagnostic(fixtures, completionReq, journal.getFixtureMatchCountsForTest(testId), defaults.requestTransform, recordMatchOptions(!!defaults.record, defaults.logger));
+	const { fixture, skippedBySequenceOrTurn } = matchFixtureDiagnostic(fixtures, completionReq, journal.getFixtureMatchCountsForTest(testId), defaults.requestTransform, {
+		...recordMatchOptions(!!defaults.record, defaults.logger),
+		wire: "ollama"
+	});
 	if (fixture) {
 		journal.incrementFixtureMatchCount(fixture, fixtures, testId);
 		logger.debug(`Fixture matched: ${JSON.stringify(fixture.match).slice(0, 120)}`);
@@ -557,7 +565,7 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 			return;
 		}
 		if (defaults.record) {
-			const outcome = await proxyAndRecord(req, res, completionReq, "ollama", urlPath, fixtures, defaults, raw);
+			const { outcome, clientHungUp } = await withHangUpDetection(req, res, () => proxyAndRecord(req, res, completionReq, "ollama", urlPath, fixtures, defaults, raw));
 			if (outcome === "handled_by_hook") return;
 			if (outcome !== "not_configured") {
 				journal.add({
@@ -568,7 +576,8 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 					response: {
 						status: res.statusCode ?? 200,
 						fixture: null,
-						source: "proxy"
+						source: "proxy",
+						...proxyInterruption(res, clientHungUp)
 					}
 				});
 				return;
@@ -626,11 +635,44 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 		}).body);
 		return;
 	}
+	if (isRawResponse(response)) {
+		await serveRawResponse({
+			wire: "ollama",
+			stream: ollamaReq.stream !== false,
+			req,
+			res,
+			fixture,
+			response,
+			body: completionReq,
+			journal,
+			defaults,
+			testId,
+			misbehaviorSummary: misbehavior.summary,
+			errorWire: "ollama"
+		});
+		return;
+	}
 	const latency = fixture.latency ?? defaults.latency;
 	const chunkSize = Math.max(1, fixture.chunkSize ?? defaults.chunkSize);
+	const replaySpeed = fixture.replaySpeed ?? defaults.replaySpeed;
+	const streamTiming = {
+		latency,
+		streamingProfile: fixture.streamingProfile,
+		recordedTimings: fixture.recordedTimings,
+		replaySpeed
+	};
+	const jsonOpts = (payloadLength, duplicateId = false) => ({
+		surface: "other",
+		payloadLength,
+		chunkSize,
+		recordedTimings: fixture.recordedTimings,
+		duplicateId,
+		replaySpeed
+	});
 	const streaming = ollamaReq.stream !== false;
 	if (misbehavior.kind === "applied") {
 		const prepared = prepareOllamaMisbehavior(misbehavior, completionReq, streaming, chunkSize, resolveRequestDecisions(defaults.strict, req.headers).reasoningSuppression, logger);
+		const chunks = !streaming && prepared.status === 200 && nonStreamingTimingEnabled(fixture, "other") ? void 0 : prepared.build();
 		const journalEntry = journal.add({
 			method: req.method ?? "POST",
 			path: urlPath,
@@ -647,24 +689,16 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 			defaults,
 			testId
 		});
-		if (!streaming) {
+		if (!streaming && prepared.status !== 200) {
 			res.writeHead(prepared.status, { "Content-Type": "application/json" });
-			res.end(JSON.stringify(prepared.chunks[0]));
-		} else {
-			const interruption = createInterruptionSignal(fixture);
-			if (!await writeNDJSONStream(res, prepared.chunks, {
-				latency,
-				streamingProfile: fixture.streamingProfile,
-				recordedTimings: fixture.recordedTimings,
-				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed,
-				signal: interruption?.signal,
-				onChunkSent: interruption?.tick
-			})) {
-				if (!res.writableEnded) res.destroy();
-				journalEntry.response.interrupted = true;
-				journalEntry.response.interruptReason = interruption?.reason();
-			}
-			interruption?.cleanup();
+			res.end(JSON.stringify((chunks ?? prepared.build())[0]));
+		} else if (!streaming) await deliverJson(res, journalEntry, fixture, () => (chunks ?? prepared.build())[0], {
+			...jsonOpts(misbehaviorPayloadLength(misbehavior, misbehavior.reasoning === void 0 ? prepared.reasoning : void 0), !!misbehavior.duplicateId),
+			status: prepared.status
+		});
+		else {
+			const frames = chunks ?? prepared.build();
+			await deliverStream(res, journalEntry, fixture, ndjsonCodec(), frames, streamTiming);
 		}
 		return;
 	}
@@ -709,26 +743,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 		const functionToolCalls = requireEmittedFunctionToolCalls(response, "Ollama");
 		const effectiveStrict = resolveRequestDecisions(defaults.strict, req.headers).reasoningSuppression;
 		const effReasoning = resolveReasoningForModel(response.reasoning, completionReq.model, effectiveStrict, logger);
-		if (!streaming) {
-			const body = buildOllamaChatContentWithToolCallsResponse(response.content ?? "", functionToolCalls, completionReq.model, effReasoning, response.blocks);
-			res.writeHead(200, { "Content-Type": "application/json" });
-			res.end(JSON.stringify(body));
-		} else {
+		if (!streaming) await deliverJson(res, journalEntry, fixture, () => buildOllamaChatContentWithToolCallsResponse(response.content ?? "", functionToolCalls, completionReq.model, effReasoning, response.blocks), jsonOpts(servedPayloadLength(response, effReasoning)));
+		else {
 			const chunks = buildOllamaChatContentWithToolCallsChunks(response.content ?? "", functionToolCalls, completionReq.model, chunkSize, effReasoning, response.blocks);
-			const interruption = createInterruptionSignal(fixture);
-			if (!await writeNDJSONStream(res, chunks, {
-				latency,
-				streamingProfile: fixture.streamingProfile,
-				recordedTimings: fixture.recordedTimings,
-				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed,
-				signal: interruption?.signal,
-				onChunkSent: interruption?.tick
-			})) {
-				if (!res.writableEnded) res.destroy();
-				journalEntry.response.interrupted = true;
-				journalEntry.response.interruptReason = interruption?.reason();
-			}
-			interruption?.cleanup();
+			await deliverStream(res, journalEntry, fixture, ndjsonCodec(), chunks, streamTiming);
 		}
 		return;
 	}
@@ -752,26 +770,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 		});
 		const effectiveStrict = resolveRequestDecisions(defaults.strict, req.headers).reasoningSuppression;
 		const effReasoning = resolveReasoningForModel(response.reasoning, completionReq.model, effectiveStrict, logger);
-		if (!streaming) {
-			const body = buildOllamaChatTextResponse(response.content, completionReq.model, effReasoning);
-			res.writeHead(200, { "Content-Type": "application/json" });
-			res.end(JSON.stringify(body));
-		} else {
+		if (!streaming) await deliverJson(res, journalEntry, fixture, () => buildOllamaChatTextResponse(response.content, completionReq.model, effReasoning), jsonOpts(servedPayloadLength(response, effReasoning)));
+		else {
 			const chunks = buildOllamaChatTextChunks(response.content, completionReq.model, chunkSize, effReasoning);
-			const interruption = createInterruptionSignal(fixture);
-			if (!await writeNDJSONStream(res, chunks, {
-				latency,
-				streamingProfile: fixture.streamingProfile,
-				recordedTimings: fixture.recordedTimings,
-				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed,
-				signal: interruption?.signal,
-				onChunkSent: interruption?.tick
-			})) {
-				if (!res.writableEnded) res.destroy();
-				journalEntry.response.interrupted = true;
-				journalEntry.response.interruptReason = interruption?.reason();
-			}
-			interruption?.cleanup();
+			await deliverStream(res, journalEntry, fixture, ndjsonCodec(), chunks, streamTiming);
 		}
 		return;
 	}
@@ -796,26 +798,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 		const functionToolCalls = requireEmittedFunctionToolCalls(response, "Ollama");
 		const effectiveStrict = resolveRequestDecisions(defaults.strict, req.headers).reasoningSuppression;
 		const effReasoning = resolveReasoningForModel(response.reasoning, completionReq.model, effectiveStrict, logger);
-		if (!streaming) {
-			const body = buildOllamaChatToolCallResponse(functionToolCalls, completionReq.model, effReasoning);
-			res.writeHead(200, { "Content-Type": "application/json" });
-			res.end(JSON.stringify(body));
-		} else {
+		if (!streaming) await deliverJson(res, journalEntry, fixture, () => buildOllamaChatToolCallResponse(functionToolCalls, completionReq.model, effReasoning), jsonOpts(servedPayloadLength(response, effReasoning)));
+		else {
 			const chunks = buildOllamaChatToolCallChunks(functionToolCalls, completionReq.model, chunkSize, effReasoning);
-			const interruption = createInterruptionSignal(fixture);
-			if (!await writeNDJSONStream(res, chunks, {
-				latency,
-				streamingProfile: fixture.streamingProfile,
-				recordedTimings: fixture.recordedTimings,
-				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed,
-				signal: interruption?.signal,
-				onChunkSent: interruption?.tick
-			})) {
-				if (!res.writableEnded) res.destroy();
-				journalEntry.response.interrupted = true;
-				journalEntry.response.interruptReason = interruption?.reason();
-			}
-			interruption?.cleanup();
+			await deliverStream(res, journalEntry, fixture, ndjsonCodec(), chunks, streamTiming);
 		}
 		return;
 	}
@@ -901,7 +887,10 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 	completionReq._endpointType = "chat";
 	completionReq._context = getContext(req);
 	const testId = getTestId(req);
-	const { fixture, skippedBySequenceOrTurn } = matchFixtureDiagnostic(fixtures, completionReq, journal.getFixtureMatchCountsForTest(testId), defaults.requestTransform, recordMatchOptions(!!defaults.record, defaults.logger));
+	const { fixture, skippedBySequenceOrTurn } = matchFixtureDiagnostic(fixtures, completionReq, journal.getFixtureMatchCountsForTest(testId), defaults.requestTransform, {
+		...recordMatchOptions(!!defaults.record, defaults.logger),
+		wire: "ollama-generate"
+	});
 	if (fixture) {
 		journal.incrementFixtureMatchCount(fixture, fixtures, testId);
 		defaults.logger.debug(`Fixture matched: ${JSON.stringify(fixture.match).slice(0, 120)}`);
@@ -935,7 +924,7 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 			return;
 		}
 		if (defaults.record) {
-			const outcome = await proxyAndRecord(req, res, completionReq, "ollama", urlPath, fixtures, defaults, raw);
+			const { outcome, clientHungUp } = await withHangUpDetection(req, res, () => proxyAndRecord(req, res, completionReq, "ollama", urlPath, fixtures, defaults, raw));
 			if (outcome === "handled_by_hook") return;
 			if (outcome !== "not_configured") {
 				journal.add({
@@ -946,7 +935,8 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 					response: {
 						status: res.statusCode ?? 200,
 						fixture: null,
-						source: "proxy"
+						source: "proxy",
+						...proxyInterruption(res, clientHungUp)
 					}
 				});
 				return;
@@ -1005,8 +995,40 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 		}).body);
 		return;
 	}
+	if (isRawResponse(response)) {
+		await serveRawResponse({
+			wire: "ollama-generate",
+			stream: generateReq.stream !== false,
+			req,
+			res,
+			fixture,
+			response,
+			body: completionReq,
+			journal,
+			defaults,
+			testId,
+			misbehaviorSummary: misbehavior.summary,
+			errorWire: "ollama"
+		});
+		return;
+	}
 	const latency = fixture.latency ?? defaults.latency;
 	const chunkSize = Math.max(1, fixture.chunkSize ?? defaults.chunkSize);
+	const replaySpeed = fixture.replaySpeed ?? defaults.replaySpeed;
+	const streamTiming = {
+		latency,
+		streamingProfile: fixture.streamingProfile,
+		recordedTimings: fixture.recordedTimings,
+		replaySpeed
+	};
+	const jsonOpts = (payloadLength, duplicateId = false) => ({
+		surface: "other",
+		payloadLength,
+		chunkSize,
+		recordedTimings: fixture.recordedTimings,
+		duplicateId,
+		replaySpeed
+	});
 	const streaming = generateReq.stream !== false;
 	if (isErrorResponse(response)) {
 		const status = response.status ?? 500;
@@ -1047,26 +1069,10 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 		});
 		const effectiveStrict = resolveRequestDecisions(defaults.strict, req.headers).reasoningSuppression;
 		const effReasoning = resolveReasoningForModel(response.reasoning, completionReq.model, effectiveStrict, defaults.logger);
-		if (!streaming) {
-			const body = buildOllamaGenerateTextResponse(response.content, completionReq.model, effReasoning);
-			res.writeHead(200, { "Content-Type": "application/json" });
-			res.end(JSON.stringify(body));
-		} else {
+		if (!streaming) await deliverJson(res, journalEntry, fixture, () => buildOllamaGenerateTextResponse(response.content, completionReq.model, effReasoning), jsonOpts(servedPayloadLength({ content: response.content }, effReasoning)));
+		else {
 			const chunks = buildOllamaGenerateTextChunks(response.content, completionReq.model, chunkSize, effReasoning);
-			const interruption = createInterruptionSignal(fixture);
-			if (!await writeNDJSONStream(res, chunks, {
-				latency,
-				streamingProfile: fixture.streamingProfile,
-				recordedTimings: fixture.recordedTimings,
-				replaySpeed: fixture.replaySpeed ?? defaults.replaySpeed,
-				signal: interruption?.signal,
-				onChunkSent: interruption?.tick
-			})) {
-				if (!res.writableEnded) res.destroy();
-				journalEntry.response.interrupted = true;
-				journalEntry.response.interruptReason = interruption?.reason();
-			}
-			interruption?.cleanup();
+			await deliverStream(res, journalEntry, fixture, ndjsonCodec(), chunks, streamTiming);
 		}
 		return;
 	}