Elastic Observability

​

Step 1: Get your Elastic endpoint and API key

​

Find your OTLP endpoint

1. In Kibana, go to Add data > Application > OpenTelemetry.
2. Copy the OTLP endpoint URL that Elastic shows for your deployment or project.

​

Create an API key

1. In Kibana, go to Stack Management > API keys, or create the key from the OpenTelemetry onboarding page.
2. Create an API key that can write trace data.
3. Copy the encoded API key.

​

Step 2: Enable Broadcast in OpenRouter

​

Step 3: Configure Elastic Observability

• Endpoint: Your Elastic OTLP endpoint. OpenRouter appends /v1/traces for you, so enter the base URL. An endpoint that already ends in /v1 or /v1/traces also works.
• API Key: Your encoded Elastic API key. OpenRouter sends it as Authorization: ApiKey <key>.
• Headers (optional): Extra HTTP headers as JSON, for example {"X-Custom-Header": "value"}. Custom headers can’t override Authorization or Content-Type.

​

Step 4: Test and save

​

Step 5: Send a test trace

​

Viewing your traces

• Applications: In Kibana, open Observability > Applications > Service inventory and select the openrouter service.
• Discover: Open Discover, pick a data view that covers traces-*, and search for a generation ID. For example:

attributes.gen_ai.response.id : "gen-1234567890-abcdefghijklmnop"

​

Trace attributes

​

Resource attributes

• service.name: Always openrouter

​

Span attributes

• openrouter.trace.id: The OpenRouter trace ID
• gen_ai.operation.name: The operation type (for example, chat)
• gen_ai.provider.name: The provider that served the request
• gen_ai.request.model: The requested model
• gen_ai.response.model: The model that served the request
• gen_ai.response.id: The OpenRouter generation ID
• gen_ai.response.finish_reasons: Why the generation ended, as a string array (for example, ["stop"])
• gen_ai.usage.input_tokens: Number of input tokens
• gen_ai.usage.output_tokens: Number of output tokens
• gen_ai.usage.cache_read.input_tokens: Input tokens served from the provider’s prompt cache
• gen_ai.usage.audio.input_tokens: Audio input tokens, when the provider reports them
• gen_ai.usage.reasoning.output_tokens: Reasoning output tokens, when the provider reports them
• gen_ai.input.messages: The request messages as a JSON string, in the OpenTelemetry role and parts format
• gen_ai.output.messages: The model’s response as a JSON string, with reasoning, text, and tool call parts
• gen_ai.tool.definitions: The tools offered to the model, as a JSON string

​

Custom Metadata

​

Supported Metadata Keys

​

Example

{
  "model": "openai/gpt-4o",
  "messages": [{ "role": "user", "content": "Summarize this incident..." }],
  "user": "user_12345",
  "session_id": "session_abc",
  "trace": {
    "trace_name": "Incident Summary",
    "generation_name": "Summarize",
    "environment": "production"
  }
}

​

Additional Context

• The user field maps to user.id in span attributes.
• The session_id field maps to session.id in span attributes.

​

Troubleshooting

​

Test Connection fails with 401

​

Traces don’t appear

1. Check the time range: Widen the Kibana time picker to include the request time.
2. Check the endpoint: Use the OTLP endpoint from the OpenTelemetry onboarding page, not your Kibana or Elasticsearch URL.

​

Additional resources

• Elastic Managed OTLP endpoint
• Elastic LLM observability
• Elastic API keys

​

Privacy Mode