Skip to main content
Broadcast automatically sends traces from AIHubMix API requests to a configured external observability platform. Monitor, debug, and analyze LLM usage in LangWatch without adding instrumentation to your application code. AIHubMix currently supports LangWatch as a Broadcast destination. After setup, LangWatch shows the model, token usage, actual billed cost, latency, session, and trace information for forwarded calls.

​
Prerequisites

Before you begin, prepare:
  • An active AIHubMix API key
  • A LangWatch account and target Project
  • Access to the AIHubMix console
This integration requires the existing Project API Key for the target Project. Its format is sk-lw-*. Organization-scoped or team-scoped keys and the pkey_* credential for a Personal OTLP Endpoint are not supported by this configuration. AIHubMix currently uses the official LangWatch Cloud endpoint and does not support self-hosted instances.

​
  1. Get the LangWatch Project API Key

  1. Sign in to LangWatch and select the Project that will receive AIHubMix traces.
  2. Open Settings → API Keys.
  3. Find Project API Key. Confirm that its Scope is the target Project, then select the copy button.
You do not need to create a new secret key. AIHubMix uses the Project API Key provided with the Project.
Project API Key in LangWatch Settings

​
  1. Add a LangWatch destination

  1. Open AIHubMix Observability.
  2. Under Add destination, select LangWatch.
LangWatch destination on the AIHubMix Observability page
  1. Complete the settings:
The x-session-id request header and the request user field are passed through with the record and can be used to correlate sessions or users in LangWatch.
LangWatch destination settings in AIHubMix

​
Test and save

  1. Select Test. After a successful connection, the button displays Connected.
  2. Select Add destination to save a new destination.
  3. Confirm that the destination is enabled and its status is OK.
For later configuration changes, select Save changes.
LangWatch destination with OK status in AIHubMix

​
  1. Send a verification request

Confirm that AIHUBMIX_API_KEY is already set in the current shell, then send a normal model request:
cURL
Records are delivered asynchronously. A trace can take several seconds to appear in LangWatch after the request completes.

​
  1. Verify the trace in LangWatch

  1. Open Traces in the target Project.
  2. Find the model used by the request, or locate the record through the x-session-id or user value.
  3. Confirm that the record includes latency, cost, token, and model information.
AIHubMix API call in LangWatch Traces
When Privacy is enabled, content columns such as Prompt remain empty while latency, cost, token, and model information remain available.
Empty content columns in LangWatch with Privacy enabled

​
Troubleshooting

​
Test returns 401

Confirm that you copied the Project API Key from the target Project. Do not use a pkey_* credential from a Personal OTLP Endpoint or an organization-scoped or team-scoped key.

​
The destination is saved but no trace appears

Check each item:
  • The destination is enabled and its status is OK.
  • Sampling rate is greater than 0%.
  • If Only forward these API keys is configured, the request uses one of the selected AIHubMix API keys.
  • Wait several seconds after the request completes, then refresh LangWatch Traces.

​
Does Privacy remove metrics?

Privacy stops prompt and completion text from being sent. Token usage, actual billed cost, latency, and model information are still sent.
Last updated: 2026-08-25