@copilotkit/aimock 1.45.0 → 1.46.0

package/dist/ollama.js

@@ -1,20 +1,24 @@
-import { InvalidToolArgumentsError, estimatePromptTokens, estimateTokens, flattenHeaders, generateDeterministicEmbedding, getContext, getTestId, isCombinedFixtureResponse, isEmbeddingResponse, isErrorResponse, isJsonObject, isTextResponse, isToolCallResponse, normalizeEmbeddingInput, requireEmittedFunctionToolCalls, resolveFixtureBlockOutcome, resolveFixtureBlocks, resolveReasoningForModel, resolveResponse, resolveStrictMode, serializeErrorResponse, servedToolArgs, strictNoMatchLogLine, strictNoMatchMessage, strictOverrideField, toolArgsForWire, validateChatMessages, validateToolsField } from "./helpers.js";
+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 { 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 { render } from "./response/usage/codecs/ollama.js";
 
 //#region src/ollama.ts
 const DURATION_FIELDS = {
 	done_reason: "stop",
 	total_duration: 0,
 	load_duration: 0,
-	prompt_eval_count: 0,
+	...render({ kind: "ordinary-zero" }, "prompt"),
 	prompt_eval_duration: 0,
-	eval_count: 0,
+	...render({ kind: "ordinary-zero" }, "completion"),
 	eval_duration: 0
 };
 function ollamaToCompletionRequest(req) {
@@ -393,8 +397,11 @@ function prepareOllamaMisbehavior(plan, request, stream, chunkSize, effectiveStr
 	const terminal = chunks.length - 1;
 	chunks[terminal] = {
 		...chunks[terminal],
-		prompt_eval_count: estimatePromptTokens(request.messages),
-		eval_count: estimateTokens(completion)
+		...render({
+			kind: "applied-chat-estimate",
+			messages: request.messages,
+			completion
+		})
 	};
 	return {
 		chunks,
@@ -467,10 +474,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError("ollama", "validation", {
 			message: `Malformed JSON body: ${detail}`,
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	if (!isJsonObject(ollamaReq)) {
@@ -484,10 +491,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError("ollama", "validation", {
 			message: "Request body must be a JSON object",
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	let chatShapeError = validateChatMessages(ollamaReq.messages) ?? validateToolsField(ollamaReq.tools);
@@ -506,10 +513,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError("ollama", "validation", {
 			message: `Invalid request: ${chatShapeError}`,
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	const completionReq = ollamaToCompletionRequest(ollamaReq);
@@ -528,7 +535,7 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 		body: completionReq
 	}, fixture ? "fixture" : "proxy", defaults.registry, defaults.logger)) return;
 	if (!fixture) {
-		if (resolveStrictMode(defaults.strict, req.headers)) {
+		if (resolveRequestDecisions(defaults.strict, req.headers).fixtureMiss) {
 			const strictStatus = 503;
 			const strictMessage = strictNoMatchMessage(skippedBySequenceOrTurn);
 			logger.error(strictNoMatchLogLine(req.method ?? "POST", urlPath, skippedBySequenceOrTurn));
@@ -543,10 +550,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 					...strictOverrideField(defaults.strict, req.headers)
 				}
 			});
-			writeErrorResponse(res, strictStatus, JSON.stringify({ error: {
+			writeErrorResponse(res, strictStatus, renderError("ollama", "strict-miss", {
 				message: strictMessage,
 				type: "invalid_request_error"
-			} }));
+			}).body);
 			return;
 		}
 		if (defaults.record) {
@@ -578,10 +585,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 				...strictOverrideField(defaults.strict, req.headers)
 			}
 		});
-		writeErrorResponse(res, 404, JSON.stringify({ error: {
+		writeErrorResponse(res, 404, renderError("ollama", "fixture-miss", {
 			message: "No fixture matched",
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	const response = await resolveResponse(fixture, completionReq);
@@ -613,14 +620,17 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 		});
 		const message = `${misbehavior.code}: ${misbehavior.message}`;
 		if (!misbehavior.summary?.evaluations.length) defaults.logger.error(message);
-		writeErrorResponse(res, misbehavior.status, JSON.stringify({ error: message }));
+		writeErrorResponse(res, misbehavior.status, renderError("ollama", "misbehavior", {
+			message: misbehavior.message,
+			code: misbehavior.code
+		}).body);
 		return;
 	}
 	const latency = fixture.latency ?? defaults.latency;
 	const chunkSize = Math.max(1, fixture.chunkSize ?? defaults.chunkSize);
 	const streaming = ollamaReq.stream !== false;
 	if (misbehavior.kind === "applied") {
-		const prepared = prepareOllamaMisbehavior(misbehavior, completionReq, streaming, chunkSize, resolveStrictMode(defaults.strict, req.headers), logger);
+		const prepared = prepareOllamaMisbehavior(misbehavior, completionReq, streaming, chunkSize, resolveRequestDecisions(defaults.strict, req.headers).reasoningSuppression, logger);
 		const journalEntry = journal.add({
 			method: req.method ?? "POST",
 			path: urlPath,
@@ -675,7 +685,7 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 			defaults,
 			testId
 		});
-		writeErrorResponse(res, status, serializeErrorResponse(response), { retryAfter: response.retryAfter });
+		writeErrorResponse(res, status, renderError("openai", "fixture", response.error).body, { retryAfter: response.retryAfter });
 		return;
 	}
 	if (isCombinedFixtureResponse(response)) {
@@ -697,7 +707,7 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 			testId
 		});
 		const functionToolCalls = requireEmittedFunctionToolCalls(response, "Ollama");
-		const effectiveStrict = resolveStrictMode(defaults.strict, req.headers);
+		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);
@@ -740,7 +750,7 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 			defaults,
 			testId
 		});
-		const effectiveStrict = resolveStrictMode(defaults.strict, req.headers);
+		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);
@@ -784,7 +794,7 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 			testId
 		});
 		const functionToolCalls = requireEmittedFunctionToolCalls(response, "Ollama");
-		const effectiveStrict = resolveStrictMode(defaults.strict, req.headers);
+		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);
@@ -824,10 +834,10 @@ async function handleOllama(req, res, raw, fixtures, journal, defaults, setCorsH
 		defaults,
 		testId
 	});
-	writeErrorResponse(res, 500, JSON.stringify({ error: {
+	writeErrorResponse(res, 500, renderError("ollama", "unknown-response", {
 		message: "Fixture response did not match any known type",
 		type: "server_error"
-	} }));
+	}).body);
 }
 async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults, setCorsHeaders) {
 	setCorsHeaders(res);
@@ -847,10 +857,10 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError("ollama", "validation", {
 			message: `Malformed JSON body: ${detail}`,
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	if (!isJsonObject(generateReq)) {
@@ -864,10 +874,10 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError("ollama", "validation", {
 			message: "Request body must be a JSON object",
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	if (!generateReq.prompt || typeof generateReq.prompt !== "string") {
@@ -881,10 +891,10 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError("ollama", "validation", {
 			message: "Invalid request: prompt field is required",
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	const completionReq = ollamaGenerateToCompletionRequest(generateReq);
@@ -903,7 +913,7 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 		body: completionReq
 	}, fixture ? "fixture" : "proxy", defaults.registry, defaults.logger)) return;
 	if (!fixture) {
-		if (resolveStrictMode(defaults.strict, req.headers)) {
+		if (resolveRequestDecisions(defaults.strict, req.headers).fixtureMiss) {
 			const strictStatus = 503;
 			const strictMessage = strictNoMatchMessage(skippedBySequenceOrTurn);
 			defaults.logger.error(strictNoMatchLogLine(req.method ?? "POST", urlPath, skippedBySequenceOrTurn));
@@ -918,10 +928,10 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 					...strictOverrideField(defaults.strict, req.headers)
 				}
 			});
-			writeErrorResponse(res, strictStatus, JSON.stringify({ error: {
+			writeErrorResponse(res, strictStatus, renderError("ollama", "strict-miss", {
 				message: strictMessage,
 				type: "invalid_request_error"
-			} }));
+			}).body);
 			return;
 		}
 		if (defaults.record) {
@@ -953,10 +963,10 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 				...strictOverrideField(defaults.strict, req.headers)
 			}
 		});
-		writeErrorResponse(res, 404, JSON.stringify({ error: {
+		writeErrorResponse(res, 404, renderError("ollama", "fixture-miss", {
 			message: "No fixture matched",
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	const response = await resolveResponse(fixture, completionReq);
@@ -989,7 +999,10 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 		});
 		const message = `${misbehavior.code}: ${misbehavior.message}`;
 		if (!misbehavior.summary?.evaluations.length) defaults.logger.error(message);
-		writeErrorResponse(res, misbehavior.status, JSON.stringify({ error: message }));
+		writeErrorResponse(res, misbehavior.status, renderError("ollama", "misbehavior", {
+			message: misbehavior.message,
+			code: misbehavior.code
+		}).body);
 		return;
 	}
 	const latency = fixture.latency ?? defaults.latency;
@@ -1012,7 +1025,7 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 			defaults,
 			testId
 		});
-		writeErrorResponse(res, status, serializeErrorResponse(response), { retryAfter: response.retryAfter });
+		writeErrorResponse(res, status, renderError("openai", "fixture", response.error).body, { retryAfter: response.retryAfter });
 		return;
 	}
 	if (isTextResponse(response)) {
@@ -1032,7 +1045,7 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 			defaults,
 			testId
 		});
-		const effectiveStrict = resolveStrictMode(defaults.strict, req.headers);
+		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);
@@ -1073,10 +1086,10 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 			defaults,
 			testId
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError("ollama", "validation", {
 			message: "Tool call fixtures are not supported on /api/generate — use /api/chat instead",
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	recordMisbehaviorOutcome({
@@ -1094,15 +1107,17 @@ async function handleOllamaGenerate(req, res, raw, fixtures, journal, defaults,
 		defaults,
 		testId
 	});
-	writeErrorResponse(res, 500, JSON.stringify({ error: {
+	writeErrorResponse(res, 500, renderError("ollama", "unknown-response", {
 		message: "Fixture response did not match any known type",
 		type: "server_error"
-	} }));
+	}).body);
 }
 async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults, setCorsHeaders) {
 	const { logger } = defaults;
 	setCorsHeaders(res);
 	const urlPath = req.url ?? "/api/embeddings";
+	const currentEndpoint = urlPath.split("?")[0] === "/api/embed";
+	const errorWire = currentEndpoint ? "ollama-embed" : "ollama";
 	let embReq;
 	try {
 		embReq = JSON.parse(raw);
@@ -1118,10 +1133,10 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError(errorWire, "validation", {
 			message: `Malformed JSON body: ${detail}`,
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	if (!isJsonObject(embReq)) {
@@ -1135,10 +1150,10 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError(errorWire, "validation", {
 			message: "Request body must be a JSON object",
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	const promptShapeError = embReq.prompt !== void 0 && embReq.prompt !== null && typeof embReq.prompt !== "string" ? "prompt field must be a string" : embReq.input !== void 0 && embReq.input !== null && normalizeEmbeddingInput(embReq.input) === null ? "input field must be a string or an array of strings" : null;
@@ -1153,14 +1168,15 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError(errorWire, "validation", {
 			message: `Invalid request: ${promptShapeError}`,
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
-	const inputText = embReq.prompt ?? (typeof embReq.input === "string" ? embReq.input : void 0) ?? (Array.isArray(embReq.input) ? embReq.input.join(" ") : void 0);
-	if (!embReq.model) {
+	const inputs = normalizeEmbeddingInput(currentEndpoint ? embReq.input ?? embReq.prompt : embReq.prompt ?? embReq.input);
+	const inputText = currentEndpoint ? inputs?.join(" ") : embReq.prompt ?? (typeof embReq.input === "string" ? embReq.input : void 0) ?? (Array.isArray(embReq.input) ? embReq.input.join(" ") : void 0);
+	if (!embReq.model || currentEndpoint && typeof embReq.model !== "string") {
 		journal.add({
 			method: req.method ?? "POST",
 			path: urlPath,
@@ -1171,10 +1187,10 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError(errorWire, "validation", {
 			message: "Invalid request: model field is required",
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	if (!inputText) {
@@ -1188,10 +1204,10 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 				fixture: null
 			}
 		});
-		writeErrorResponse(res, 400, JSON.stringify({ error: {
+		writeErrorResponse(res, 400, renderError(errorWire, "validation", {
 			message: "Invalid request: prompt or input field is required",
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	const syntheticReq = {
@@ -1201,6 +1217,11 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 		_endpointType: "embedding",
 		_context: getContext(req)
 	};
+	setEmbeddingRequest(syntheticReq, {
+		operation: currentEndpoint ? "/api/embed" : "/api/embeddings",
+		input: currentEndpoint ? embReq.input ?? embReq.prompt ?? "" : embReq.prompt ?? embReq.input ?? "",
+		...embReq.dimensions !== void 0 && { dimensions: embReq.dimensions }
+	});
 	const testId = getTestId(req);
 	const { fixture, skippedBySequenceOrTurn } = matchFixtureDiagnostic(fixtures, syntheticReq, journal.getFixtureMatchCountsForTest(testId), defaults.requestTransform, recordMatchOptions(!!defaults.record, defaults.logger));
 	if (fixture) {
@@ -1227,7 +1248,23 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 					fixture
 				}
 			});
-			writeErrorResponse(res, status, serializeErrorResponse(response), { retryAfter: response.retryAfter });
+			writeErrorResponse(res, status, renderError(currentEndpoint ? "ollama-embed" : "openai", "fixture", response.error).body, { retryAfter: response.retryAfter });
+			return;
+		}
+		if (currentEndpoint && isJSONResponse(response)) {
+			const status = response.status ?? 200;
+			journal.add({
+				method: req.method ?? "POST",
+				path: urlPath,
+				headers: flattenHeaders(req.headers),
+				body: syntheticReq,
+				response: {
+					status,
+					fixture
+				}
+			});
+			res.writeHead(status, { "Content-Type": "application/json" });
+			res.end(JSON.stringify(response.json));
 			return;
 		}
 		if (isEmbeddingResponse(response)) {
@@ -1241,7 +1278,10 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 					fixture
 				}
 			});
-			const body = {
+			const body = currentEndpoint ? {
+				model: embReq.model,
+				embeddings: (inputs ?? []).map(() => [...response.embedding])
+			} : {
 				model: embReq.model,
 				embedding: [...response.embedding]
 			};
@@ -1259,13 +1299,13 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 				fixture
 			}
 		});
-		writeErrorResponse(res, 500, JSON.stringify({ error: {
+		writeErrorResponse(res, 500, renderError(errorWire, "unknown-response", {
 			message: "Fixture response did not match any known embedding type (must have embedding or error)",
 			type: "server_error"
-		} }));
+		}).body);
 		return;
 	}
-	if (resolveStrictMode(defaults.strict, req.headers)) {
+	if (resolveRequestDecisions(defaults.strict, req.headers).fixtureMiss) {
 		const strictMessage = strictNoMatchMessage(skippedBySequenceOrTurn);
 		logger.error(strictNoMatchLogLine(req.method ?? "POST", urlPath, skippedBySequenceOrTurn));
 		journal.add({
@@ -1279,10 +1319,10 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 				...strictOverrideField(defaults.strict, req.headers)
 			}
 		});
-		writeErrorResponse(res, 503, JSON.stringify({ error: {
+		writeErrorResponse(res, 503, renderError(errorWire, "strict-miss", {
 			message: strictMessage,
 			type: "invalid_request_error"
-		} }));
+		}).body);
 		return;
 	}
 	if (defaults.record) {
@@ -1303,8 +1343,25 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 			return;
 		}
 	}
+	const dimensions = currentEndpoint ? validateEmbeddingDimensions(embReq.dimensions) : 1536;
+	if (dimensions === null) {
+		journal.add({
+			method: req.method ?? "POST",
+			path: urlPath,
+			headers: flattenHeaders(req.headers),
+			body: syntheticReq,
+			response: {
+				status: 400,
+				fixture: null
+			}
+		});
+		writeErrorResponse(res, 400, renderError(errorWire, "validation", {
+			message: `Invalid request: dimensions must be an integer between 1 and ${MAX_EMBEDDING_DIMENSIONS}`,
+			type: "invalid_request_error"
+		}).body);
+		return;
+	}
 	logger.warn(`No embedding fixture matched for "${inputText.slice(0, 80)}" — returning deterministic fallback`);
-	const embedding = generateDeterministicEmbedding(inputText);
 	journal.add({
 		method: req.method ?? "POST",
 		path: urlPath,
@@ -1315,9 +1372,12 @@ async function handleOllamaEmbeddings(req, res, raw, fixtures, journal, defaults
 			fixture: null
 		}
 	});
-	const body = {
+	const body = currentEndpoint ? {
+		model: embReq.model,
+		embeddings: (inputs ?? []).map((input) => generateDeterministicEmbedding(input, dimensions))
+	} : {
 		model: embReq.model,
-		embedding
+		embedding: generateDeterministicEmbedding(inputText)
 	};
 	res.writeHead(200, { "Content-Type": "application/json" });
 	res.end(JSON.stringify(body));