@zespan/sdk for TypeScript/Node and zespan for Python — built on the same core: auto-instrumentation for your LLM provider of choice, manual span tracing for anything else, and clients for prompts, datasets, and guardrails. Neither language is the “primary” one; both get the same features on the same release cadence.
Most feature pages below cover both languages on one page, with TypeScript and Python usage side by side — you only need the language-specific reference pages (below) for full installation, configuration, and initialization details.
Language references
TypeScript SDK
Install, initialize, and configure
@zespan/sdk — provider auto-patching, agent workflows, serverless flush handling.Python SDK
Install, initialize, and configure
zespan — provider auto-patching, agent workflows, serverless flush handling.By feature
Which page do I actually need?
- Setting up tracing for the first time? Start with the Quickstart, then the language reference above for your stack.
- Already tracing LLM calls, want to add something specific — agents, prompts, guardrails, a dataset run? Go straight to that feature’s page in the table above; each one has both languages in tabs, so you don’t need the full language reference for it.
- Using LangChain, Bedrock, Pinecone, or another named library? That’s Integrations, not this page — auto-instrumentation is organized by vendor there, not by Zespan feature.