🐛 Bug Report: Do i need a KEY for local use?

[DSgUY]

Which component is this bug for?

Traceloop SDK

📜 Description

I'm using Grafana localy. Im trying to create a simple test script but I do not want to create a TRACELOOP KEY. I cant use OPENLLMETRY?

👟 Reproduction steps

NC

👍 Expected behavior

NC

👎 Actual Behavior with Screenshots

NC

🤖 Python Version

3.12

📃 Provide any additional context for the Bug.

No response

👀 Have you spent some time to check if this bug has been raised before?

• I checked and didn't find similar issue

Are you willing to submit PR?

None

[dosubot[bot]]

To use the Traceloop SDK locally without a TRACELOOP KEY, you can configure it to work with OpenTelemetry directly. Here are the steps you can follow:

1. Set the Base URL: Define the OpenTelemetry endpoint to connect to. For local use, set this to a local endpoint or a different observability platform.

   TRACELOOP_BASE_URL=<Your Local OpenTelemetry Endpoint>

2. Disable Batch Processing: Disable batch processing to see traces immediately.

   Traceloop.init(disable_batch=True)

3. Use a Custom Traces Exporter: If you cannot use the default OTLP/HTTP or OTLP/GRPC exporter, set a custom exporter, such as Zipkin, for local testing.

   Traceloop.init(exporter=ZipkinExporter(endpoint="http://localhost:9411/api/v2/spans"))

4. Disable Tracing of Prompt Content: For privacy or to reduce trace size, disable logging of prompt content.

   TRACELOOP_TRACE_CONTENT=false

5. Control Logging Level: Set the logging level to debug for more detailed logs during local development.

   Traceloop.initialize({ logLevel: "debug" })

These configurations allow you to use the Traceloop SDK locally without needing a TRACELOOP KEY ^[1].

To continue talking to Dosu, mention @dosu.

[dosubot[bot]]

To use the Traceloop SDK locally without a TRACELOOP KEY, you can configure it to work with OpenTelemetry directly. Here are the steps you can follow:

1. Set the Base URL: Define the OpenTelemetry endpoint to connect to. For local use, set this to a local endpoint or a different observability platform.

   TRACELOOP_BASE_URL=<Your Local OpenTelemetry Endpoint>

2. Disable Batch Processing: Disable batch processing to see traces immediately.

   Traceloop.init(disable_batch=True)

3. Use a Custom Traces Exporter: If you cannot use the default OTLP/HTTP or OTLP/GRPC exporter, set a custom exporter, such as Zipkin, for local testing.

   Traceloop.init(exporter=ZipkinExporter(endpoint="http://localhost:9411/api/v2/spans"))

4. Disable Tracing of Prompt Content: For privacy or to reduce trace size, disable logging of prompt content.

   TRACELOOP_TRACE_CONTENT=false

5. Control Logging Level: Set the logging level to debug for more detailed logs during local development.

   Traceloop.initialize({ logLevel: "debug" })

These configurations allow you to use the Traceloop SDK locally without needing a TRACELOOP KEY ^[1].

To continue talking to Dosu, mention @dosu.

[nirga]

@DSgUY hey! have you looked at https://www.traceloop.com/docs/openllmetry/integrations/grafana?

[DSgUY]

Yes, I saw that but still cant make it work. I have everthing set in docker: tempo, grafana, etc.

[nirga]

Interesting, what was missing for you?
I have on my todo list to revisit our docs as we had someone recently facing some issues. He posted his comments there -
https://github.com/traceloop/openllmetry/issues/1715#issuecomment-2261367959

Does this help? I didn't have time to spin off a local grafana to test it myself that's why haven't changed the docs - but if you can verify that this helps (and even help me fix the docs?) - that would be tremendously helpful!

[nirga]

In any chance, if you set TRACELOOP_BASE_URL it means the SDK is sending data to whatever you defined there - so the API key is no longer needed nor requird.