Skip to main content

Routing evidence

Every Lasso response tells you what served it and what it cost, and one query parameter adds which provider answered and why. Use this page to debug a slow or failed request, explain a bill, or prove a routing change worked.

Where evidence lives

Every header, metadata field and error code is listed in response signals and errors. The routing engine’s metadata fields are defined once, in request metadata.

Debug a request

  1. Capture x-lasso-request-id at your client.
  2. Check x-lasso-access and x-lasso-fallback: was it served where you expected, and if not, why?
  3. Read lasso_meta: which strategy applied, which providers were candidates, which one served it, and how many retries it took.
  4. For an error Lasso originates, read error.data.code and error.data.next_action; they name the fix. Provider errors keep their code and message.
  5. Check the key’s recent_errors and the dashboard’s provider health for a pattern across requests.
When something in Lasso is confusing or broken, report it, quoting the request ID:
See agent entry points for the fields.

Limits

  • Managed providers appear under public aliases such as premium_base_2; Custom profiles show your own provider names.
  • candidate_providers lists the order considered, not the attempts. Use selected_provider and retries.
  • An HTTP batch’s x-lasso-meta header describes one item. Use include_meta=body to see every item’s evidence.
  • Usage views summarize activity for insight; the account balance is the record of money.
  • The dashboard’s recent activity is a live stream, not a request log.

Next

Understand latency

lasso_meta timings measure work inside Lasso, including upstream calls. They exclude client network RTT, TCP/TLS setup, edge routing, and response delivery. Reuse connections with HTTP keep-alive or HTTP/2. In a production Base eth_blockNumber measurement on October 29, 2026, median client latency was 119 ms with a fresh connection versus 42 ms reused; metadata reported 13 ms. These are measurements from one client location, not a latency guarantee. The public tier uses free providers and has a wider latency tail than premium.