Trouble with Next OpenAI Telemetry App Demo Connection Issues

·Oct 08, 2024 07:58 PM

Hi all, I've been having a fair amount of trouble getting the demo [next-openi-telemetry-app](https://github.com/Arize-ai/openinference/tree/main/js/examples/next-openai-telemetry-app) demo working. Have attempted every permutation of PHOENIX_COLLECTOR_ENDPOINT and PHOENIX_API_KEY I can think of. Have tried connecting to a live Phoenix Space at https://app.phoenix.arize.com/v1/traces and to local docker instance. Nothing seems to work here. I guess first of all I wanted to verify that this should be working?

30 comments

· Sorted by Oldest

RunLLM
APP
·
Yes, the demo should work with both a live Phoenix Space and a local Docker instance. Here are some key points to verify: 1. Environment Variables: Ensure that PHOENIX_COLLECTOR_ENDPOINT and PHOENIX_API_KEY are correctly set. For a live Phoenix Space, PHOENIX_COLLECTOR_ENDPOINT should be [https://app.phoenix.arize.com/v1/traces](https://app.phoenix.arize.com/v1/traces). For a local Docker instance, it should be [http://localhost:6006/v1/traces](http://localhost:6006/v1/traces).2. Network Accessibility: Ensure that the Phoenix server is accessible from your application. For local Docker, verify that the Docker container is running and accessible on the specified port.3. Instrumentation Initialization: Ensure that the instrumentation code is correctly initialized before any application code is executed. This is crucial for the traces to be captured.Would you like more detailed troubleshooting steps or a code example?Sources: - https://arize-ai.slack.com/archives/C04R3GXC8HK/p1722778288.923159 - https://arize-ai.slack.com/archives/C04R3GXC8HK/p1720418482.719169
💡 Hint: Mention RunLLM in the thread for followups.
Michael S.
·
I've been through most steps. This actually helped me spot the issue I was having locally — trying to use the grpc endpoint rather than the http endpoint. The basic issue remains though that attempting to use the exact same, working setup to send traces to a hosted Phoenix just give me a 401 error. API Key has been checked and double checked.
Mikyo
·
Hey Michael S. - def. can help. Which setup would you like to work on local or the managed phoenix instance?
Michael S.
·
Seems like I have the local instance settled. That was just for debugging purposes. The big mystery all along has been my inability to connect to the managed instance.

Michael S.

Swapping in PHOENIX_API_KEY and PHOENIX_COLLECTOR_ENDPOINT='https://app.phoenix.arize.com/v1/traces' for the working local vals (no api key for local, obviously) I get the following error logged:

{"stack":"OTLPExporterError: Unauthorized\n    at new OTLPExporterError (webpack-internal:///(instrument)/./node_modules/@opentelemetry/otlp-exporter-base/build/esm/types.js:41:28)\n    at IncomingMessage.eval (webpack-internal:///(instrument)/./node_modules/@opentelemetry/otlp-exporter-base/build/esm/platform/node/util.js:134:37)\n    at IncomingMessage.emit (node:events:529:35)\n    at endReadableNT (node:internal/streams/readable:1400:12)\n    at process.processTicksAndRejections (node:internal/process/task_queues:82:21)","message":"Unauthorized","name":"OTLPExporterError","code":"401"}

Michael S.
·
I've also attempted to set a space_id header, and tried a few different endpoints I've seen in other examples, but no luck.

Michael S.

So just to set a baseline for repro:

1. Cloned the repo above, went to folder, `npm install`
2. Copied .env.local.example -> .env.local
3. Set env values:
OPENAI_API_KEY='sk-pw-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX'
PHOENIX_API_KEY='bffXXXXXXXXXXXXXX'
PHOENIX_COLLECTOR_ENDPOINT='http://localhost:6006/v1/traces'
4. Started local Phoenix instance
5. Started npm run dev, verified it logged
6. Changed env value:
PHOENIX_COLLECTOR_ENDPOINT='https://app.phoenix.arize.com/v1/traces'
7. Restarted npm run dev
8. Sent a message. Saw the 401 error logged.

Michael S.
·
Have attempted with API keys from two different accts, one which has some old traces in it, and one which is new.
Michael S.
·
Experimenting, I set up auth on the local Phoenix, and noticed that the example is not correct for that — I had to change the headers to authorization: Bearer MY_TOKEN. I've tried something similar against the hosted Phoenix, but no luck.
Mikyo
·
hey Michael S. sorry got pulled into another meeting. It's actually a bit different in managed phoenix than local phoenix (we are aware that this is confusing - cc Kiko C. Eric X.) So managed phoenix uses keys a bit differently. It has it's own headers:
os.environ["PHOENIX_CLIENT_HEADERS"] = "api_key=...:..."
Mikyo
·
so the key you see in the UI in managed phoenix need to be set as spelled out below 👇
Mikyo
·
can you give this part a try?
Mikyo
·
https://docs.arize.com/phoenix/hosted-phoenix#using-the-client-sdk-downloading-data-and-uploading-datasets
Michael S.
·
I see — though in this case I'm using openinference and OTEL, not the client, so wouldn't those env params be roughly equivalent?

Michael S.

exporter: new OTLPTraceExporter({
   headers: {
     api_key: process.env["PHOENIX_API_KEY"]
   },
   url: process.env["PHOENIX_COLLECTOR_ENDPOINT"] ||            "https://app.phoenix.arize.com/v1/traces"
}),

RunLLM
APP
·
Yes, the demo should work with both a live Phoenix Space and a local Docker instance. Here are some key points to verify: 1. Environment Variables: Ensure that PHOENIX_COLLECTOR_ENDPOINT and PHOENIX_API_KEY are correctly set. For a live Phoenix Space, PHOENIX_COLLECTOR_ENDPOINT should be [https://app.phoenix.arize.com/v1/traces](https://app.phoenix.arize.com/v1/traces). For a local Docker instance, it should be [http://localhost:6006/v1/traces](http://localhost:6006/v1/traces).2. Network Accessibility: Ensure that the Phoenix server is accessible from your application. For local Docker, verify that the Docker container is running and accessible on the specified port.3. Instrumentation Initialization: Ensure that the instrumentation code is correctly initialized before any application code is executed. This is crucial for the traces to be captured.Would you like more detailed troubleshooting steps or a code example?Sources: - https://arize-ai.slack.com/archives/C04R3GXC8HK/p1722778288.923159 - https://arize-ai.slack.com/archives/C04R3GXC8HK/p1720418482.719169
💡 Hint: Mention RunLLM in the thread for followups.
Michael S.
·
I've been through most steps. This actually helped me spot the issue I was having locally — trying to use the grpc endpoint rather than the http endpoint. The basic issue remains though that attempting to use the exact same, working setup to send traces to a hosted Phoenix just give me a 401 error. API Key has been checked and double checked.
Mikyo
·
Hey Michael S. - def. can help. Which setup would you like to work on local or the managed phoenix instance?
Michael S.
·
Seems like I have the local instance settled. That was just for debugging purposes. The big mystery all along has been my inability to connect to the managed instance.

Michael S.

Swapping in PHOENIX_API_KEY and PHOENIX_COLLECTOR_ENDPOINT='https://app.phoenix.arize.com/v1/traces' for the working local vals (no api key for local, obviously) I get the following error logged:

{"stack":"OTLPExporterError: Unauthorized\n    at new OTLPExporterError (webpack-internal:///(instrument)/./node_modules/@opentelemetry/otlp-exporter-base/build/esm/types.js:41:28)\n    at IncomingMessage.eval (webpack-internal:///(instrument)/./node_modules/@opentelemetry/otlp-exporter-base/build/esm/platform/node/util.js:134:37)\n    at IncomingMessage.emit (node:events:529:35)\n    at endReadableNT (node:internal/streams/readable:1400:12)\n    at process.processTicksAndRejections (node:internal/process/task_queues:82:21)","message":"Unauthorized","name":"OTLPExporterError","code":"401"}

Michael S.
·
I've also attempted to set a space_id header, and tried a few different endpoints I've seen in other examples, but no luck.

Michael S.

So just to set a baseline for repro:

1. Cloned the repo above, went to folder, `npm install`
2. Copied .env.local.example -> .env.local
3. Set env values:
OPENAI_API_KEY='sk-pw-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX'
PHOENIX_API_KEY='bffXXXXXXXXXXXXXX'
PHOENIX_COLLECTOR_ENDPOINT='http://localhost:6006/v1/traces'
4. Started local Phoenix instance
5. Started npm run dev, verified it logged
6. Changed env value:
PHOENIX_COLLECTOR_ENDPOINT='https://app.phoenix.arize.com/v1/traces'
7. Restarted npm run dev
8. Sent a message. Saw the 401 error logged.

Michael S.
·
Have attempted with API keys from two different accts, one which has some old traces in it, and one which is new.
Michael S.
·
Experimenting, I set up auth on the local Phoenix, and noticed that the example is not correct for that — I had to change the headers to authorization: Bearer MY_TOKEN. I've tried something similar against the hosted Phoenix, but no luck.
Mikyo
·
hey Michael S. sorry got pulled into another meeting. It's actually a bit different in managed phoenix than local phoenix (we are aware that this is confusing - cc Kiko C. Eric X.) So managed phoenix uses keys a bit differently. It has it's own headers:
os.environ["PHOENIX_CLIENT_HEADERS"] = "api_key=...:..."
Mikyo
·
so the key you see in the UI in managed phoenix need to be set as spelled out below 👇
Mikyo
·
can you give this part a try?
Mikyo
·
https://docs.arize.com/phoenix/hosted-phoenix#using-the-client-sdk-downloading-data-and-uploading-datasets
Michael S.
·
I see — though in this case I'm using openinference and OTEL, not the client, so wouldn't those env params be roughly equivalent?

Michael S.

exporter: new OTLPTraceExporter({
   headers: {
     api_key: process.env["PHOENIX_API_KEY"]
   },
   url: process.env["PHOENIX_COLLECTOR_ENDPOINT"] ||            "https://app.phoenix.arize.com/v1/traces"
}),