feat: Add search template mutation

[Dschoordsch]

Description

Fixes #9774, #9775
Add a template search query to the User object.

Demo

https://www.loom.com/share/114efbf0a1864ba09816c0bf87a28620?sid=86867080-17c0-405a-bf6a-828a591bdfad

Testing scenarios

• search for templates in the activity library
• create custom templates, wait for the embedding (ca. 1min, you can check EmbeddingsJobQueue is empty) and search for it

Final checklist

• I checked the code review guidelines
• I have added Metrics Representative as reviewer(s) if my PR invovles metrics/data/analytics related changes
• I have performed a self-review of my code, the same way I'd do it for any other team member
• I have tested all cases I listed in the testing scenarios and I haven't found any issues or regressions
• Whenever I took a non-obvious choice I added a comment explaining why I did it this way
• I added the label Skip Maintainer Review Indicating the PR only requires reviewer review and can be merged right after it's approved if the PR introduces only minor changes, does not contain any architectural changes or does not introduce any new patterns and I think one review is sufficient'
• PR title is human readable and could be used in changelog

[Dschoordsch]

@tianrunhe I changed the debounced query logic. Because this is used for analytics, let me know if that's ok.

[nickoferrall]

Hey @Dschoordsch, I'm getting a timeout error when using the semantic search: https://www.loom.com/share/11e27dce1ba747829b205b87f090a932

Is there any set-up required to test this one?

 rejection:
17:13:43 1|Socket Server     | Error: TIMEOUT
17:13:43 1|Socket Server     |     at Timeout._onTimeout (/Users/nickoferrall/parabol/dev/web.js:80818:16)
17:13:43 1|Socket Server     |     at listOnTimeout (node:internal/timers:573:17)
17:13:43 1|Socket Server     |     at process.processTimers (node:internal/timers:514:7)
17:14:25 2|GraphQL Executor  | TypeError: fetch failed
17:14:25 2|GraphQL Executor  |     at fetch (/Users/nickoferrall/parabol/node_modules/undici/index.js:112:15)
17:14:25 2|GraphQL Executor  |     at async Object.templateSearch (/Users/nickoferrall/parabol/dev/gqlExecutor.js:36867:22)
17:14:25 2|GraphQL Executor  | Trace: {
17:14:25 2|GraphQL Executor  |   error: '{"data":null,"errors":[{"message":"fetch failed","locations":[{"line":1,"column":575}],"path":["viewer","templateSearch"]}]}'
17:14:25 2|GraphQL Executor  | }

[tianrunhe]

LGTM metrics side

[Dschoordsch]

I'm getting a timeout error when using the semantic search

I tried to reproduce this, but had no success yet:

• when no embedder is running and the embedding table is empty, I don't get a result but no error either
• when I stop the text-embeddings-inference container, I get an error in fetch but a different one
• I tried querying right after startup when the text-embeddings-inference container could be busy, but did not get any error

@nickoferrall Are you processing lots of data atm? Can you check your EmbeddingsJobQueue table how many entries you have?

[nickoferrall]

My EmbeddingsJobQueue had over 100 entries, and my Embeddings_ember_1 table had no entries. I removed all of the entries in the job queue, but got the same result.

Are these env vars the same as .env.example for you?

[nickoferrall]

@Dschoordsch after our call, I removed everything from Docker, restarted my laptop, cleared the job queue and metadata a couple of times, but no luck, unfortunately. The embeddings service is still slow to respond so I get a timeout.

We don't know whether it shouldn't be working for you, or it should be working for me, so perhaps someone else could test this PR so we can isolate the problem?

Could @tianrunhe test it as you're here please?

[tianrunhe]

My EmbeddingsJobQueue had over 100 entries, and my Embeddings_ember_1 table had no entries. I removed all of the entries in the job queue, but got the same result.

I encountered the same behavior as Nick. The stateMessage field for the rows are unable to get tokens:

[Dschoordsch]

Could one of you try deleting the dev folder

rm -r dev

and the embedding data

delete from "Embeddings_ember_1";
delete from "EmbeddingsMetadata";
delete from "EmbeddingsJobQueue";

and then try again?

[Dschoordsch]

@nickoferrall ☝️ could you try one more time? I noticed that locally I have some issues with the embedding service not always rebuilding.

[nickoferrall]

I tried the step above, but no luck.

I'm working on getting the text-embeddings-inference server install on bare metal here following this now

[nickoferrall]

I got the embedder up and running on my m1, but it’s still not working 😞

The job queue is empty as is the Embeddings_ember_1 table

[Dschoordsch]

I will go ahead and merge the change. If it fails, it will just yield no results, which is safe. As embeddings work in production, I'm positive the search will work there.

[nickoferrall]

-1 this should be a env variable instead. I'm now using port 3041 for the text-embeddings-inference server because I'm running on my mac rather than Docker

[rafaelromcar-parabol]

Also, no localhost or port should be hardcoded into our code as each component can run in any port and location in other environments.