🪢 Open source agent evals & observability: Trace, evaluate, and improve LLM applications with one open platform.
TypeScript analyticsautogenevaluationlangchainlarge-language-modelsllama-indexllmllm-evaluationllm-observabilityllmopsmonitoringobservabilityopen-sourceopenaiplaygroundprompt-engineeringprompt-managementself-hostedycombinator
import { describe, expect, it } from "vitest";
import {
mapEventEvalFilterColumnIdToField,
type ObservationForEval,
zipObservationToolCalls,
} from "./observationForEval";
describe("zipObservationToolCalls", () => {
it("zips ClickHouse parallel arrays into named tool calls with parsed arguments", () => {
const result = zipObservationToolCalls({
tool_calls: [
'{"id":"call_1","arguments":"{\\"query\\":\\"weather\\"}","type":"function","index":0}',
'{"id":"toolu_2","arguments":"{\\"city\\":\\"Berlin\\"}","type":"tool_use","index":1}',
],
tool_call_names: ["search", "get_weather"],
});
expect(result).toEqual([
{
id: "call_1",
name: "search",
arguments: { query: "weather" },
type: "function",
index: 0,
},
{
id: "toolu_2",
name: "get_weather",
arguments: { city: "Berlin" },
type: "tool_use",
index: 1,
},
]);
});
it("returns an empty array for observations without tool calls", () => {
expect(
zipObservationToolCalls({ tool_calls: [], tool_call_names: [] }),
).toEqual([]);
});
it("keeps unparsable arguments as the raw string", () => {
const result = zipObservationToolCalls({
tool_calls: [
'{"id":"call_1","arguments":"not json","type":"","index":0}',
],
tool_call_names: ["search"],
});
expect(result[0]?.arguments).toBe("not json");
});
it("falls back to defaults when an entry is missing or malformed", () => {
const result = zipObservationToolCalls({
tool_calls: ["not json at all"],
tool_call_names: ["search", "unmatched"],
});
expect(result).toEqual([
{ id: "", name: "search", arguments: {}, type: "", index: 0 },
{ id: "", name: "unmatched", arguments: {}, type: "", index: 0 },
]);
});
it("accepts entries that are already objects", () => {
const result = zipObservationToolCalls({
tool_calls: [
{ id: "call_1", arguments: { a: 1 }, type: "function", index: 2 },
],
tool_call_names: ["search"],
});
expect(result).toEqual([
{
id: "call_1",
name: "search",
arguments: { a: 1 },
type: "function",
index: 2,
},
]);
});
});
describe("mapEventEvalFilterColumnIdToField", () => {
it("maps filters backed directly by the evaluation payload", () => {
const observation = {
provided_model_name: "gpt-4o",
prompt_name: "support-agent",
prompt_version: 2,
release: "2026-08",
status_message: "rate limit exceeded",
experiment_name: "checkout-eval",
} as ObservationForEval;
expect(
mapEventEvalFilterColumnIdToField(observation, "providedModelName"),
).toBe("gpt-4o");
expect(mapEventEvalFilterColumnIdToField(observation, "promptName")).toBe(
"support-agent",
);
expect(
mapEventEvalFilterColumnIdToField(observation, "promptVersion"),
).toBe(2);
expect(mapEventEvalFilterColumnIdToField(observation, "release")).toBe(
"2026-08",
);
expect(
mapEventEvalFilterColumnIdToField(observation, "statusMessage"),
).toBe("rate limit exceeded");
expect(
mapEventEvalFilterColumnIdToField(observation, "experimentName"),
).toBe("checkout-eval");
});
it("maps the experiment root filter to a boolean", () => {
expect(
mapEventEvalFilterColumnIdToField(
{
span_id: "root-span",
experiment_item_root_span_id: "root-span",
} as ObservationForEval,
"isExperimentItemRootSpan",
),
).toBe(true);
expect(
mapEventEvalFilterColumnIdToField(
{
span_id: "child-span",
experiment_item_root_span_id: "root-span",
} as ObservationForEval,
"isExperimentItemRootSpan",
),
).toBe(false);
});
});