Quickstart
Install the Spanloom Python SDK, create a project, and see your first trace in under five minutes.
1. Install the SDK
Install the spanloom package from PyPI:
$ pip install spanloom
The SDK requires Python 3.9 or later.
2. Get your API key
Sign up at spanloom.com, create a project, and copy the API key from Settings. You will use it in the next step.
3. Instrument your first LLM call
Add three lines to your existing OpenAI code:
import openai
import spanloom
spanloom.init(api_key="spl_your_key_here")
client = openai.OpenAI()
response = client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user", "content": "Summarise the RAG pipeline."}]
)
print(response.choices[0].message.content)
That is it. Spanloom patches the OpenAI client automatically on init(). Every call is traced.
4. View your trace
Run your script. Within a few seconds, the trace appears in the Spanloom dashboard under the project you created. You will see:
- The span name, model, and prompt hash
- Input and output token counts
- Latency (time-to-first-token and total)
- Estimated cost based on the model's published token rate
Advanced: wrapping a multi-step pipeline
For pipelines with multiple LLM calls, retrieval steps, and tool invocations, use the context manager to group spans into a single named trace:
import spanloom
with spanloom.trace("rag_answer", user_id="u_123") as trace:
docs = retrieve_documents(query)
with trace.span("retrieve_docs"):
docs = retrieve_documents(query)
with trace.span("llm_answer"):
answer = call_llm(query, docs)
return answer
Each trace.span() block creates a child span in the waterfall. Nesting is unlimited.
TypeScript / Node.js
The TypeScript SDK is available in open beta:
npm install @spanloom/sdk
import { init } from "@spanloom/sdk";
init({ apiKey: "spl_your_key_here" });
Usage mirrors the Python SDK. The TypeScript SDK auto-instruments openai, LangChain.js, and the Vercel AI SDK on init.