feat(diagnostics-otel): add genai usage span attrs

This commit is contained in:
Vincent Koc
2026-04-25 11:55:26 -07:00
parent bb2b68b34e
commit cd7a8f870b
3 changed files with 78 additions and 0 deletions

View File

@@ -35,6 +35,7 @@ Docs: https://docs.openclaw.ai
to the state-managed `plugins/installs.json` ledger, with legacy config reads
kept as a deprecated compatibility fallback. Thanks @vincentkoc.
- Diagnostics/OTEL: add the GenAI `gen_ai.client.operation.duration` histogram for model-call latency in seconds with bounded provider/model/API and error attributes. Thanks @vincentkoc.
- Diagnostics/OTEL: add GenAI usage token attributes to model-usage spans, including cache read/write input token counts without session identifiers or prompt/response content. Thanks @vincentkoc.
- Diagnostics/OTEL: add bounded outbound message delivery lifecycle diagnostics and export them as low-cardinality delivery spans/metrics without message body, recipient, room, or media-path data. (#71471) Thanks @vincentkoc and @jlapenna.
- Diagnostics/OTEL: emit bounded exec-process diagnostics and export them as `openclaw.exec` spans without exposing command text, working directories, or container identifiers. (#71451) Thanks @vincentkoc and @jlapenna.
- Diagnostics/OTEL: support `OPENCLAW_OTEL_PRELOADED=1` so the plugin can reuse an already-registered OpenTelemetry SDK while keeping OpenClaw diagnostic listeners wired. (#71450) Thanks @vincentkoc and @jlapenna.

View File

@@ -740,6 +740,53 @@ describe("diagnostics-otel service", () => {
await service.stop?.(ctx);
});
test("exports GenAI usage attributes on model usage spans without diagnostic identifiers", async () => {
const service = createDiagnosticsOtelService();
const ctx = createOtelContext(OTEL_TEST_ENDPOINT, { traces: true });
await service.start(ctx);
emitDiagnosticEvent({
type: "model.usage",
sessionKey: "session-key",
sessionId: "session-id",
provider: "anthropic",
model: "claude-sonnet-4.6",
usage: {
input: 100,
output: 40,
cacheRead: 30,
cacheWrite: 20,
promptTokens: 150,
total: 190,
},
durationMs: 25,
});
await flushDiagnosticEvents();
const modelUsageCall = telemetryState.tracer.startSpan.mock.calls.find(
(call) => call[0] === "openclaw.model.usage",
);
expect(modelUsageCall?.[1]).toMatchObject({
attributes: {
"gen_ai.usage.input_tokens": 150,
"gen_ai.usage.output_tokens": 40,
"gen_ai.usage.cache_read.input_tokens": 30,
"gen_ai.usage.cache_creation.input_tokens": 20,
},
});
expect(modelUsageCall?.[1]).toEqual({
attributes: expect.not.objectContaining({
"openclaw.sessionKey": expect.anything(),
"openclaw.sessionId": expect.anything(),
"gen_ai.input.messages": expect.anything(),
"gen_ai.output.messages": expect.anything(),
}),
startTime: expect.any(Number),
});
expect(JSON.stringify(modelUsageCall)).not.toContain("session-key");
await service.stop?.(ctx);
});
test("exports GenAI client operation duration histogram without diagnostic identifiers", async () => {
const service = createDiagnosticsOtelService();
const ctx = createOtelContext(OTEL_TEST_ENDPOINT, { metrics: true });

View File

@@ -177,6 +177,21 @@ function genAiOperationName(
return "chat";
}
function positiveFiniteNumber(value: number | undefined): number | undefined {
return typeof value === "number" && Number.isFinite(value) && value > 0 ? value : undefined;
}
function assignPositiveNumberAttr(
attrs: Record<string, string | number>,
key: string,
value: number | undefined,
): void {
const normalized = positiveFiniteNumber(value);
if (normalized !== undefined) {
attrs[key] = normalized;
}
}
function assignGenAiModelCallAttrs(
attrs: Record<string, string | number | boolean>,
evt: ModelCallLifecycleDiagnosticEvent,
@@ -933,6 +948,9 @@ export function createDiagnosticsOtelService(): OpenClawPluginService {
if (!tracesEnabled) {
return;
}
const genAiInputTokens =
usage.promptTokens ??
(usage.input ?? 0) + (usage.cacheRead ?? 0) + (usage.cacheWrite ?? 0);
const spanAttrs: Record<string, string | number> = {
...attrs,
"openclaw.tokens.input": usage.input ?? 0,
@@ -941,6 +959,18 @@ export function createDiagnosticsOtelService(): OpenClawPluginService {
"openclaw.tokens.cache_write": usage.cacheWrite ?? 0,
"openclaw.tokens.total": usage.total ?? 0,
};
assignPositiveNumberAttr(spanAttrs, "gen_ai.usage.input_tokens", genAiInputTokens);
assignPositiveNumberAttr(spanAttrs, "gen_ai.usage.output_tokens", usage.output);
assignPositiveNumberAttr(
spanAttrs,
"gen_ai.usage.cache_read.input_tokens",
usage.cacheRead,
);
assignPositiveNumberAttr(
spanAttrs,
"gen_ai.usage.cache_creation.input_tokens",
usage.cacheWrite,
);
const span = spanWithDuration("openclaw.model.usage", spanAttrs, evt.durationMs, {
parentContext: contextForTrustedDiagnosticSpanParent(evt, metadata),