[sannin]

3 months, 3 weeks ago #33548

Hello,

I see that vespa integration is being pushed back in releases. I understand that it is quite complex, but do you have an estimation on when it will be released?

[wpsolr]

3 months, 3 weeks ago #33549

The integration is 95% done, but a few issues remain that need to be fixed by the Vespa team.

You can check the showstoppers on the roadmap: https://github.com/vespa-engine/vespa/issues/26694

What would be your use case that would require Vespa over Weaviate for instance?

[sannin]

3 months, 2 weeks ago #34192

Hey, sorry for the late reply. Vespa support ai ranking models such as xgboost.
Unfortunately the showstopper issues seems that it will take a while to be completed…

[wpsolr]

3 months, 2 weeks ago #34195

Dynamic summary https://github.com/vespa-engine/vespa/issues/26693 is fixed already.

Arrays sorting https://github.com/vespa-engine/vespa/issues/26681 is the main showstopper now.

But as I understand, you do not need sorting, as you wish to sort by a custom xgboost ranking.

If you’re interested, we could discuss the opportunity of a pre-release.

[sannin]

3 months, 2 weeks ago #34200

Yes, i would like to try a pre-release version.

[wpsolr]

3 months, 2 weeks ago #34205

I just sent you a chat invite via Gmail.

[sannin]

3 months, 2 weeks ago #34213

I have not received anything…

[wpsolr]

3 months, 2 weeks ago #34218

Nice talk.

Summary: WooCommerce with 200K products. The POC will be built with a Vespa docker instance, with a goal to validate the scalability during indexing on a 30 CPUs hardware. If all goes well, a cluster deployment will be investigated. The POC will test the deployment of xgboost ranking, and eventually Vespa will be used also for recommendations.

Here is the beta release including Vespa:
https://www.dropbox.com/s/1uva1f4x0btl7ln/wpsolr-pro-vespa-beta-1.zip?dl=1

Here is a sample docker:

services:
  node1:
    hostname: node1.vespanet
    image: ${WPSOLR_IMAGE_NAME}:vespa
    command: configserver,services
    healthcheck:
      test: curl https://localhost:19071/state/v1/health
      timeout: 10s
      retries: 3
      start_period: 40s
    ports:
      - 8080:8080
      - 19071:19071
    environment:
      VESPA_CONFIGSERVERS: node1.vespanet
      VESPA_CONFIGSERVER_JVMARGS: "-Xms32M -Xmx128M"
      VESPA_CONFIGPROXY_JVMARGS: "-Xms32M -Xmx32M"
    networks:
      - wpsolr-external-network-created-with-docker-network-create

In WPSOLR, you can setup the urls as:
– Admin cluster: https://node1.vespanet:19071
– Stateless cluster: https://node1.vespanet:8080

[wpsolr]

3 months ago #34899

Could you make some progress with WPSOLR’s Vespa pre-release?

[sannin]

1 month, 2 weeks ago #35323

Hello,

Sorry for the late reply :/ I must admit that i failed miserably with vespa intergration. 😛 I did not have enough patience to dig in deeper and i encountered some bugs in the beta version of wpsolr that prevented the normal operation. I am not giving it up, but i will wait the official release to try it out again.

But nevertheless thanks to one of your blog posts i discovered metarank, which i managed to implement, but i will discuss it more at another thread.

[wpsolr]

1 month, 2 weeks ago #35324

Metarank looks great indeed. We are still planning to integrate it in WPSOLR. I’d be more than happy to discuss about it with you.

We also built new Vespa.ai demos that you can try now.

If you want help for Vespa.ai, let me know. You can pick a meeting slot in the chatbot calendar link.