All committers have signed the CLA.

Quest PR submitted!

@kartikay-bagla You are attempting to solve the issue and loot this Quest. Will you be successful?

Questions? Check out the docs.

Thanks @kartikay-bagla
Looking into your code now.
Running the tests is easy using nx (https://www.traceloop.com/docs/openllmetry/contributing/developing#basic-guide-for-using-nx) - nx run-many -t test

One comment about the test - I'd test it similarly to how we did it in Python (i.e. - instead of made-up vectors).

One comment about the test - I'd test it similarly to how we did it in Python (i.e. - instead of made-up vectors).

I thought since we're specifically testing the tracing with pinecone, it would be better to reduce dependency on other things.

One comment about the test - I'd test it similarly to how we did it in Python (i.e. - instead of made-up vectors).

I thought since we're specifically testing the tracing with pinecone, it would be better to reduce dependency on other things.

Yeah that also makes sense so I'm fine with that. There's always the tension between writing integration tests and unit tests. We chose the integration test path since they catch more bugs easily (specifically in the LLM domain where the library maintainers have a tendency to break APIs all the time), but they are harder to run locally and in CI.
So I'm also fine with us having more easy to run unit tests :)

BTW we have an effort on the Python SDK to instrument the vectors themselves so might worth doing it here as well.

BTW we have an effort on the Python SDK to instrument the vectors themselves so might worth doing it here as well.

I'm sorry but I'm not exactly clear on what "instrumenting the vectors" mean.

Also, I was able to get the unit tests working :)

BTW we have an effort on the Python SDK to instrument the vectors themselves so might worth doing it here as well.

I'm sorry but I'm not exactly clear on what "instrumenting the vectors" mean.

Also, I was able to get the unit tests working :)

Check out this PR. Basically it's sending as span events the vectors you get in response when calling Pinecone's query method.

@nirga I have added span events as well!.

Thanks @kartikay-bagla! Let me know when it's ready for review and I'll do a full check up :)

Thanks @kartikay-bagla! Let me know when it's ready for review and I'll do a full check up :)

@nirga apart from needing to call manuallyInstrument, everything else is good to go.

@kartikay-bagla I suspect that the auto-instrumentation is not working well since it can't find the module automatically. We encountered some similar issues in the langchain instrumentation, see the solution @Kartik1397 came up with there -

@kartikay-bagla I suspect that the auto-instrumentation is not working well since it can't find the module automatically. We encountered some similar issues in the langchain instrumentation, see the solution @Kartik1397 came up with there -

openllmetry-js/packages/instrumentation-langchain/src/instrumentation.ts

Line 38 in 109cf0e

protected init(): InstrumentationModuleDefinition<any>[] {

I think that fixes it! Sample app is now working as well!

@nirga you can do a full review now.

Looks good overall @kartikay-bagla! Can we just align the semantic conventions to the ones we have in Python, and add them to the ai-semantic-conventions package?

Also, I see the test is failing with 404... Does it also happen to you locally?

Also, I see the test is failing with 404... Does it also happen to you locally?

The 404 could happen due to a missing API key for pinecone. You need to add it to the environment as PINECONE_API_KEY

The API key was there, the problem was with the configuration we had in Pinecone :/
I've fixed that now but I see the build is failing as you're missing dependency on ai semantic conventions in the pinecone instrumentation. Also, when I ran the tests locally I get an error:

 1 failing
  1) Test Pinecone instrumentation
       should set attributes in span for DB query:
      AssertionError [ERR_ASSERTION]: Expected values to be strictly equal:
2 !== 8

      + expected - actual
      -2
      +8
      
      at Context.<anonymous> (test/instrumentation.test.ts:127:12)
      at processTicksAndRejections (node:internal/process/task_queues:95:5)

(BTW the build and test here fails automatically on your commits due to some Github limitation that doesn't allow outside contributors to use our secrets for running the tests and I'm triggering it manually everytime. So it should pass if the tests work and I trigger them).

Fixed package.json issue.

And I have attributed the test error to just not waiting long enough for pinecone to update. I tried running the tests locally about 100 times, it sometimes failed like yours did. Added a 30 second wait time before querying any freshly inserted datam and that seemed to fix the issue!

🧚 @kartikay-bagla congratulations for completing Quest #32

💰 A reward of $100 has been credited to you.

To claim your $100 reward follow the instructions here.

Questions? Check out the docs.