GitHub - SKE-Labs/lightrace: A lightweight tracing platform for LLM development

# npm (recommended)
npm install -g @lightrace/cli

# Homebrew
brew install SKE-Labs/tap/lightrace

# Go
go install github.com/SKE-Labs/lightrace-cli@latest

lightrace init      # generate config with secrets
lightrace start     # pull images & start all services

lightrace status           # service health + URLs
lightrace status -o env    # export SDK connection vars

from lightrace import Lightrace, trace

lt = Lightrace(
    public_key="pk-lt-demo",
    secret_key="sk-lt-demo",
    host="http://localhost:3000",
)

@trace()
def run_agent(query: str):
    return search(query)

@trace(type="tool")  # remotely invocable from the UI
def search(query: str) -> list:
    return ["result"]

run_agent("hello")
lt.flush()

import { LightRace, trace } from "lightrace";

const lt = new LightRace({
  publicKey: "pk-lt-demo",
  secretKey: "sk-lt-demo",
  host: "http://localhost:3000",
});

const search = trace("search", { type: "tool" }, async (query: string) => {
  return ["result"];
});

const runAgent = trace("run-agent", async (query: string) => {
  return search(query);
});

await runAgent("hello");
lt.flush();

from langchain_core.messages import HumanMessage
from langchain_openai import ChatOpenAI
from lightrace import Lightrace
from lightrace.integrations.langchain import LightraceCallbackHandler

lt = Lightrace(public_key="pk-lt-demo", secret_key="sk-lt-demo")

handler = LightraceCallbackHandler(client=lt)
model = ChatOpenAI(model="gpt-4o-mini", max_tokens=256)

response = model.invoke(
    [HumanMessage(content="What is the speed of light?")],
    config={"callbacks": [handler]},
)

lt.flush()
lt.shutdown()

import anyio
from claude_agent_sdk import AssistantMessage, ResultMessage, TextBlock
from lightrace import Lightrace
from lightrace.integrations.claude_agent_sdk import traced_query

lt = Lightrace(public_key="pk-lt-demo", secret_key="sk-lt-demo")

async def main():
    async for message in traced_query(
        prompt="Read the files in the current directory and summarize them.",
        options={"max_turns": 5},
        client=lt,
        trace_name="file-summarizer",
    ):
        if isinstance(message, AssistantMessage):
            for block in message.content:
                if isinstance(block, TextBlock):
                    print(block.text)
        elif isinstance(message, ResultMessage):
            print(f"Cost: ${message.total_cost_usd:.4f}")

    lt.flush()
    lt.shutdown()

anyio.run(main)

import { Lightrace } from "lightrace";
import { tracedQuery } from "lightrace/integrations/claude-agent-sdk";

const lt = new Lightrace({ publicKey: "pk-lt-demo", secretKey: "sk-lt-demo" });

for await (const message of tracedQuery({
  prompt: "Read the files in the current directory and summarize them.",
  options: { maxTurns: 5 },
  client: lt,
  traceName: "file-summarizer",
})) {
  if (message.type === "result") {
    const r = message as Record<string, unknown>;
    console.log(r.result);
  }
}

lt.flush();
await lt.shutdown();

export LANGFUSE_PUBLIC_KEY=pk-lt-demo
export LANGFUSE_SECRET_KEY=sk-lt-demo
export LANGFUSE_HOST=http://localhost:3000

packages/
  shared/    — Prisma schema, DB + Redis clients, Zod validation
  backend/   — Hono server (tRPC + REST ingestion + OTel + tool registry)
  frontend/  — Next.js 15 dashboard + Auth.js