Same agent Smaller bill. The tools and infrastructure to get more work from your AI. Fewer tokens. Less compute. Lower cost.
TypeScript agent-harnessagent-harness-engineeringai-engineeringai-observabilityamd-gpuclaude-codeclickhousecodexcoding-agentcoding-agentscursorgenaigpu-monitoringllm-observabilityllmopsnvidia-smiobservabilityopen-sourceopenaiopentelemetry
#!/usr/bin/env python3
"""Send synthetic GenAI spans to a local OTLP/HTTP receiver.
Pairs with scripts/dev-realtime.sh to exercise the realtime path:
receiver -> NATS -> openlit-engine -> /realtime findings.
Usage:
OPENLIT_API_KEY=... scripts/dev-realtime-send-spans.py --count 20 --error-rate 1
"""
import argparse
import json
import os
import random
import secrets
import sys
import time
import urllib.error
import urllib.request
def attr(key, value):
if isinstance(value, bool):
return {"key": key, "value": {"boolValue": value}}
if isinstance(value, int):
return {"key": key, "value": {"intValue": str(value)}}
if isinstance(value, float):
return {"key": key, "value": {"doubleValue": value}}
return {"key": key, "value": {"stringValue": str(value)}}
def build(args):
now = time.time_ns()
spans = []
for i in range(args.count):
failed = random.random() < args.error_rate
duration_ns = int(args.latency_ms * 1_000_000)
start = now - duration_ns - i * 1_000_000
spans.append(
{
"traceId": secrets.token_hex(16),
"spanId": secrets.token_hex(8),
"name": f"chat {args.model}",
"kind": 3,
"startTimeUnixNano": str(start),
"endTimeUnixNano": str(start + duration_ns),
"status": {"code": 2, "message": "provider error"} if failed else {"code": 1},
"attributes": [
attr("gen_ai.operation.name", "chat"),
attr("gen_ai.provider.name", args.provider),
attr("gen_ai.request.model", args.model),
attr("gen_ai.usage.input_tokens", args.input_tokens),
attr("gen_ai.usage.output_tokens", args.output_tokens),
attr("gen_ai.usage.cost", args.cost),
]
+ ([attr("http.response.status_code", args.http_status)] if args.http_status else []),
}
)
return {
"resourceSpans": [
{
"resource": {
"attributes": [
attr("service.name", args.service),
attr("deployment.environment", args.environment),
]
},
"scopeSpans": [{"scope": {"name": "openlit.dev"}, "spans": spans}],
}
]
}
def main():
p = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
p.add_argument("--endpoint", default=os.environ.get("OTLP_ENDPOINT", "http://127.0.0.1:4318"))
p.add_argument("--api-key", default=os.environ.get("OPENLIT_API_KEY", ""))
p.add_argument("--count", type=int, default=20)
p.add_argument("--error-rate", type=float, default=0.0)
p.add_argument("--latency-ms", type=float, default=800)
p.add_argument("--service", default="realtime-demo")
p.add_argument("--environment", default="production")
p.add_argument("--provider", default="openai")
p.add_argument("--model", default="gpt-4o-mini")
p.add_argument("--input-tokens", type=int, default=400)
p.add_argument("--output-tokens", type=int, default=200)
p.add_argument("--cost", type=float, default=0.002)
p.add_argument("--http-status", type=int, default=0, help="Set http.response.status_code, e.g. 401")
args = p.parse_args()
body = json.dumps(build(args)).encode()
headers = {"Content-Type": "application/json"}
if args.api_key:
headers["Authorization"] = f"Bearer {args.api_key}"
req = urllib.request.Request(f"{args.endpoint.rstrip('/')}/v1/traces", data=body, headers=headers)
try:
with urllib.request.urlopen(req, timeout=15) as resp:
print(f"sent {args.count} spans: HTTP {resp.status}")
except urllib.error.HTTPError as e:
print(f"receiver rejected spans: HTTP {e.code} {e.read().decode(errors='replace')}", file=sys.stderr)
sys.exit(1)
if __name__ == "__main__":
main()