Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions .env.example
Original file line number Diff line number Diff line change
Expand Up @@ -22,5 +22,10 @@ LANGSMITH_API_KEY=""
LANGSMITH_TRACING="false"
# Trace transport: how spans are shipped ("native" | "otel").
EVAL_TRACE_TRANSPORT="native"
# Optional Braintrust OTEL parent override (defaults to project_name:stagehand-dev,
# or project_name:stagehand in CI).
# BRAINTRUST_OTEL_PARENT="project_name:stagehand-dev"
# Optional Braintrust OTEL traces endpoint override.
# BRAINTRUST_OTEL_URL="https://api.braintrust.dev/otel/v1/traces"
# Primary trace sink when multiple are configured ("braintrust" | "langsmith").
EVAL_TRACE_PRIMARY="braintrust"
1 change: 1 addition & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,7 @@ ctrf/
packages/evals/playwright-mcp-screenshot-*.png
packages/evals/chrome-devtools-mcp-screenshot-*.png
.trajectories/
.rubric-cache/

# The browse CLI has a command source dir named "downloads" that must be tracked
# despite the broad downloads/ rule above.
Expand Down
149 changes: 149 additions & 0 deletions packages/evals/framework/otel.ts
Original file line number Diff line number Diff line change
@@ -0,0 +1,149 @@
import { ProxyTracerProvider, type Tracer } from "@opentelemetry/api";
import {
BatchSpanProcessor,
type SpanProcessor,
} from "@opentelemetry/sdk-trace-base";
import { NodeTracerProvider } from "@opentelemetry/sdk-trace-node";

import { hasLangSmithApiKey, resolveTraceTransport } from "./langsmith.js";

const TRACER_NAME = "stagehand-evals";
const SHUTDOWN_TIMEOUT_MS = 10_000;
const NOOP_TRACER = new ProxyTracerProvider().getTracer(TRACER_NAME);

// Provider state lives on globalThis, NOT in module scope: the CLI bundle
// (esbuild --bundle) inlines this module once per import site, so module-level
// variables would give each importer its own copy — the runner would register
// the provider in one copy while tracedSpan reads NOOP from another, silently
// dropping every span. A Symbol.for-keyed global is shared across all copies.
type TracingState = {
provider: NodeTracerProvider | null;
providerPromise: Promise<NodeTracerProvider | null> | null;
};
const STATE_KEY = Symbol.for("stagehand.evals.otel.state");
function state(): TracingState {
const g = globalThis as { [STATE_KEY]?: TracingState };
return (g[STATE_KEY] ??= { provider: null, providerPromise: null });
}

/** Test-only: clear shared provider state between vitest module resets. */
export function resetTracingStateForTests(): void {
Comment thread
miguelg719 marked this conversation as resolved.
const g = globalThis as { [STATE_KEY]?: TracingState };
delete g[STATE_KEY];
}

export async function buildTracerProvider(options?: {
braintrustParent?: string;
}): Promise<NodeTracerProvider | null> {
if (resolveTraceTransport() !== "otel") {
return null;
}

const s = state();
if (s.provider) {
return s.provider;
}

// The provider is initialized once per process, so the first call's options win.
const pendingProvider =
s.providerPromise ??
(s.providerPromise = initializeTracerProvider(options));

try {
return await pendingProvider;
} finally {
if (s.providerPromise === pendingProvider) {
s.providerPromise = null;
}
}
}

async function initializeTracerProvider(options?: {
braintrustParent?: string;
}): Promise<NodeTracerProvider | null> {
const spanProcessors: SpanProcessor[] = [];

const braintrustApiKey = process.env.BRAINTRUST_API_KEY;
if (braintrustApiKey) {
const { OTLPTraceExporter } = await import(
"@opentelemetry/exporter-trace-otlp-proto"
);
const braintrustProjectName =
process.env.CI === "true" ? "stagehand" : "stagehand-dev";
const parent =
options?.braintrustParent ??
process.env.BRAINTRUST_OTEL_PARENT ??
`project_name:${braintrustProjectName}`;
spanProcessors.push(
new BatchSpanProcessor(
new OTLPTraceExporter({
url:
process.env.BRAINTRUST_OTEL_URL ??
"https://api.braintrust.dev/otel/v1/traces",
headers: {
Authorization: `Bearer ${braintrustApiKey}`,
"x-bt-parent": parent,
},
}),
),
);
}

if (hasLangSmithApiKey() && process.env.LANGSMITH_TRACING === "true") {
const { LangSmithOTLPTraceExporter } = await import(
"langsmith/experimental/otel/exporter"
);
spanProcessors.push(
new BatchSpanProcessor(new LangSmithOTLPTraceExporter()),
);
}

if (spanProcessors.length === 0) {
return null;
}

const nextProvider = new NodeTracerProvider({ spanProcessors });
nextProvider.register();
state().provider = nextProvider;
return nextProvider;
}

export function getTracer(): Tracer {
return state().provider?.getTracer(TRACER_NAME) ?? NOOP_TRACER;
}

export async function shutdownTracing(): Promise<void> {
const s = state();
if (resolveTraceTransport() !== "otel" || !s.provider) {
return;
}

const activeProvider = s.provider;
s.provider = null;

let timeout: ReturnType<typeof setTimeout> | undefined;
const timeoutPromise = new Promise<never>((_, reject) => {
timeout = setTimeout(() => {
reject(
new Error(
`Timed out shutting down tracing after ${SHUTDOWN_TIMEOUT_MS}ms.`,
),
);
}, SHUTDOWN_TIMEOUT_MS);
timeout.unref?.();
});

try {
await Promise.race([
activeProvider
.forceFlush()
.catch(() => {})
.then(() => activeProvider.shutdown()),
timeoutPromise,
]);
} finally {
if (timeout) {
clearTimeout(timeout);
}
}
}
4 changes: 4 additions & 0 deletions packages/evals/package.json
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,10 @@
"@anthropic-ai/claude-agent-sdk": "^0.2.141",
"@browserbasehq/stagehand": "workspace:*",
"@openai/codex-sdk": "0.125.0",
"@opentelemetry/api": "^1.9.0",
"@opentelemetry/exporter-trace-otlp-proto": "^0.220.0",
"@opentelemetry/sdk-trace-base": "^2.0.0",
"@opentelemetry/sdk-trace-node": "^2.0.0",
"ai": "^5.0.133",
"dotenv": "^17.3.1",
"langsmith": "^0.5.16",
Expand Down
166 changes: 166 additions & 0 deletions packages/evals/tests/framework/otel.test.ts
Original file line number Diff line number Diff line change
@@ -0,0 +1,166 @@
import { afterAll, beforeEach, describe, expect, it, vi } from "vitest";

const mocks = vi.hoisted(() => {
const braintrustExporter = vi.fn(function () {
return {
export: vi.fn(),
shutdown: vi.fn().mockResolvedValue(undefined),
};
});
const langSmithExporter = vi.fn(function () {
return {
export: vi.fn(),
shutdown: vi.fn().mockResolvedValue(undefined),
};
});
const nodeTracerProvider = vi.fn(function (options: unknown) {
return {
forceFlush: vi.fn().mockResolvedValue(undefined),
getTracer: vi.fn(),
options,
register: vi.fn(),
shutdown: vi.fn().mockResolvedValue(undefined),
};
});

return {
braintrustExporter,
langSmithExporter,
nodeTracerProvider,
};
});

vi.mock("@opentelemetry/exporter-trace-otlp-proto", () => ({
OTLPTraceExporter: mocks.braintrustExporter,
}));

vi.mock("langsmith/experimental/otel/exporter", () => ({
LangSmithOTLPTraceExporter: mocks.langSmithExporter,
}));

vi.mock("@opentelemetry/sdk-trace-node", () => ({
NodeTracerProvider: mocks.nodeTracerProvider,
}));

const originalEnv = { ...process.env };

function constructedSpanProcessors(): unknown[] {
const options = mocks.nodeTracerProvider.mock.calls[0]?.[0] as
| { spanProcessors: unknown[] }
| undefined;
return options?.spanProcessors ?? [];
}

describe("buildTracerProvider", () => {
beforeEach(async () => {
process.env = { ...originalEnv };
delete process.env.BRAINTRUST_API_KEY;
delete process.env.BRAINTRUST_OTEL_PARENT;
delete process.env.BRAINTRUST_OTEL_URL;
delete process.env.CI;
delete process.env.EVAL_TRACE_TRANSPORT;
delete process.env.LANGSMITH_API_KEY;
delete process.env.LANGSMITH_TRACING;
vi.clearAllMocks();
vi.resetModules();
// Provider state is globalThis-backed (bundle-duplication proof), so a
// module reset alone no longer isolates tests — use the module's own
// test-reset helper to clear the shared slot.
const { resetTracingStateForTests } = await import(
"../../framework/otel.js"
);
resetTracingStateForTests();
});

afterAll(() => {
process.env = originalEnv;
});

it("does not construct a provider in native mode", async () => {
const { buildTracerProvider } = await import("../../framework/otel.js");

await expect(buildTracerProvider()).resolves.toBeNull();
expect(mocks.nodeTracerProvider).not.toHaveBeenCalled();
});

it("constructs a provider with both backend processors", async () => {
process.env.EVAL_TRACE_TRANSPORT = "otel";
process.env.BRAINTRUST_API_KEY = "braintrust-test-key";
process.env.LANGSMITH_API_KEY = "langsmith-test-key";
process.env.LANGSMITH_TRACING = "true";
const { buildTracerProvider } = await import("../../framework/otel.js");

await expect(buildTracerProvider()).resolves.not.toBeNull();
expect(mocks.braintrustExporter).toHaveBeenCalledWith({
url: "https://api.braintrust.dev/otel/v1/traces",
headers: {
Authorization: "Bearer braintrust-test-key",
"x-bt-parent": "project_name:stagehand-dev",
},
});
expect(mocks.langSmithExporter).toHaveBeenCalledOnce();
expect(constructedSpanProcessors()).toHaveLength(2);
});

it("constructs a provider with only the Braintrust processor", async () => {
process.env.EVAL_TRACE_TRANSPORT = "otel";
process.env.BRAINTRUST_API_KEY = "braintrust-test-key";
const { buildTracerProvider } = await import("../../framework/otel.js");

await expect(buildTracerProvider()).resolves.not.toBeNull();
expect(mocks.braintrustExporter).toHaveBeenCalledWith({
url: "https://api.braintrust.dev/otel/v1/traces",
headers: {
Authorization: "Bearer braintrust-test-key",
"x-bt-parent": "project_name:stagehand-dev",
},
});
expect(mocks.langSmithExporter).not.toHaveBeenCalled();
expect(constructedSpanProcessors()).toHaveLength(1);
});

it("uses the provided Braintrust parent over the environment default", async () => {
process.env.EVAL_TRACE_TRANSPORT = "otel";
process.env.BRAINTRUST_API_KEY = "braintrust-test-key";
process.env.BRAINTRUST_OTEL_PARENT = "project_name:from-env";
const { buildTracerProvider } = await import("../../framework/otel.js");

await expect(
buildTracerProvider({ braintrustParent: "project_name:custom" }),
).resolves.not.toBeNull();
expect(mocks.braintrustExporter).toHaveBeenCalledWith({
url: "https://api.braintrust.dev/otel/v1/traces",
headers: {
Authorization: "Bearer braintrust-test-key",
"x-bt-parent": "project_name:custom",
},
});
});

it("constructs a provider with only the LangSmith processor", async () => {
process.env.EVAL_TRACE_TRANSPORT = "otel";
process.env.LANGSMITH_API_KEY = "langsmith-test-key";
process.env.LANGSMITH_TRACING = "true";
const { buildTracerProvider } = await import("../../framework/otel.js");

await expect(buildTracerProvider()).resolves.not.toBeNull();
expect(mocks.braintrustExporter).not.toHaveBeenCalled();
expect(mocks.langSmithExporter).toHaveBeenCalledOnce();
expect(constructedSpanProcessors()).toHaveLength(1);
});

it("shuts down the provider when forceFlush rejects", async () => {
process.env.EVAL_TRACE_TRANSPORT = "otel";
process.env.BRAINTRUST_API_KEY = "braintrust-test-key";
const { buildTracerProvider, shutdownTracing } = await import(
"../../framework/otel.js"
);

await buildTracerProvider();
const activeProvider = mocks.nodeTracerProvider.mock.results[0]?.value;
activeProvider.forceFlush.mockRejectedValueOnce(new Error("export failed"));

await expect(shutdownTracing()).resolves.toBeUndefined();
expect(activeProvider.shutdown).toHaveBeenCalledOnce();
});
});
Loading
Loading