Cáceres
Account created on 9 October 2018, about 7 years ago
#

Merge Requests

More

Recent comments

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

xjm credited gxleano .

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

That's true, thanks Shibin!

Then I would like to improve the list to show the user the current status of the workflow and also provide the ability to enable/disable from operations.

What do you think?

🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

gxleano made their first commit to this issue’s fork.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Removing some redundant validation.

Thanks @ushma

It will be part of release 1.2.43

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

It makes totally sense, thanks Dieter!

It will be added into release 1.2.43

🇪🇸Spain gxleano Cáceres

gxleano made their first commit to this issue’s fork.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Thanks @hydra!

It looks good to me, so it will be added to release 1.2.43

🇪🇸Spain gxleano Cáceres

gxleano made their first commit to this issue’s fork.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Thank you very much @s_leu!!

It will be part of the next release 1.2.43

🇪🇸Spain gxleano Cáceres

gxleano made their first commit to this issue’s fork.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Thanks Pablo!!

It will be includes in next release 1.2.43

🇪🇸Spain gxleano Cáceres

gxleano made their first commit to this issue’s fork.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Thanks for the work done here!

It will be included to next release 1.2.43

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Testing instructions

After have configured a valid reranker provider from/admin/config/ai/settings using the new operation type coming from https://www.drupal.org/node/3488114 you would need to:

1. Assuming that you already have a Search API already configured using AI Search as backend, go to /admin/config/search/search-api/index/<your_index_name>/processors
2. Enable AI Search Reranker option
3. Configure it selecting the AI Provider and reranker model, select the maximum number of top results to rerank and add prompt instructions if it is needed, leave blank for default behavior.

🇪🇸Spain gxleano Cáceres

New processor

Reranker processor configuration

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

The AI Search module is included as a submodule within AI Core versions 1.1.x-dev and 1.2.x-dev, as well as any version below 2.0.x. This means AI Search will continue to work as long as you are using an AI Core version earlier than 2.0.x.

Starting with AI Core 2.0.x, the AI Search module has been moved out of AI Core. If you are using AI Core 2.0.x, you will need to install AI Search 1.0.x separately from the contrib space.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Seems like it is already covered on 2.0.x, so I will close this issue.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

I've been reviewing and testing the issue and everything works as expected.

Moving to the MR and using a test site with existing Chatbot block in place, after running the implemented hook post update the block is gone.

Thanks @harivansh.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Testing instructions

  1. Use AI Search module with MR https://www.drupal.org/project/ai_search/issues/3557901 Add Re-ranking Support to AI Search Active
  2. Install Cohere Provider module https://www.drupal.org/project/ai_provider_cohere and use it with MR https://www.drupal.org/project/ai_provider_cohere/issues/3557838 📌 Update Cohere provider to work with Re Rank operation type Needs review
  3. Configure the Search API server with AI Search backend to use Reranking by enabling reranking option and configure it
  4. Reindex content
  5. Do a query using AI Search Backend
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

I’ll review your comments and add some steps to test the changes.

In this case, we need to verify that the provider and model are working correctly with ReRank selection. Once that’s confirmed, we should debug whether the selected model behaves as expected. We have two options:

  1. Test it with custom code to validate this MR directly, or
  2. Wait for the related issues to be resolved and then test it through AI Search:
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

It looks good to me Scott!

But we would need to wait merging until the problem with AI Search package is resolved.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

gxleano created an issue. See original summary .

🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres

Yep, as mentioned by Valery in #11 Change namespace for Field Widget Actions module Active , it still point to the Drupal AI repo.

See:

composer show --available drupal/ai_search
name     : drupal/ai_search
descrip. : Offers an implementation of the Search API that uses a Vector DB server for indexing content.
keywords :
versions : 1.2.3, 1.2.2, 1.2.1, 1.2.0, 1.2.0-rc2, 1.2.0-rc1, 1.2.0-beta1, 1.2.0-alpha2, 1.2.0-alpha1, 1.1.6, 1.1.5, 1.1.4, 1.1.3, 1.1.2, 1.1.1, 1.1.0, 1.1.0-rc1, 1.1.0-beta1, 1.0.x-dev, 1.0.6, 1.0.5, 1.0.4, 1.0.3, 1.0.2, 1.0.1, 1.0.0, 1.0.0-beta7, 1.0.0-beta6, 1.0.0-beta5, 1.0.0-beta4, 1.0.0-beta3, 1.0.0-beta2, 1.0.0-beta1, 1.0.0-alpha8, 1.0.0-alpha7, 1.0.0-alpha6, 1.0.0-alpha5, dev-1.0.x
type     : metapackage
license  : GNU General Public License v2.0 or later (GPL-2.0-or-later) (OSI approved) https://spdx.org/licenses/GPL-2.0-or-later.html#licenseText
homepage : https://www.drupal.org/project/ai
source   : []
dist     : []
names    : drupal/ai_search

support
source : https://git.drupalcode.org/project/ai

requires
drupal/ai *
drupal/search_api *
drupal/core ^10.4 || ^11
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

I've been considering the possibility of bringing this back, so I’ll reopen the issue and try to dedicate some time to implement a working version of the re-ranking feature, which seems to have been neglected for a while.

After we have included this hook https://www.drupal.org/project/ai/issues/3556016 Add hook_ai_search_results_alter() to allow other modules to modify or re-rank AI search results Active on AI Search module, I think that could be nice to have the possibility to re-rank results.

So,re-ranking acts as a “second pass” that fine-tunes results to make them more relevant, personalized, or optimized for specific goals.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Facing same issue while I was using Ollama provider.

When going to /admin/config/ai/providers/ollama, I go the next error:

Symfony\Component\Routing\Exception\InvalidParameterException: Parameter "model_id" for route "ai.edit_model_settings_form" must match "[^/]++" ("thinkverse/towerinstruct:latest" given) to generate a corresponding URL. in Drupal\Core\Routing\UrlGenerator->doGenerate() (line 223 of core/lib/Drupal/Core/Routing/UrlGenerator.php).

After setting up the MR the error is gone, so the problem is solved.

Thanks Kevin and Stefan.

🇪🇸Spain gxleano Cáceres

gxleano made their first commit to this issue’s fork.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Thanks Scott to provide more details in description.

I've been testing carefully the fix, and it is working as expected.

The doSearchWithIteration() method now correctly:

  • Detects when access checks reduce results below the limit
  • Iterates to fetch more results from Milvus
  • Replaces filtered nodes with accessible alternatives
  • Returns the full requested number of results

Moving to RTBC, so feel free to merge it.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres

gxleano created an issue.

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Looks good to me Scott!

Thanks!

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

Look good to me Scott!

Thanks!

🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres
🇪🇸Spain gxleano Cáceres

@harivansh

Check the failing pipelines, when everything is ready just move the issue to Needs review, please.

🇪🇸Spain gxleano Cáceres

Hey Kirsten,

Not sure if this is what we are expecting here, but https://git.drupalcode.org/project/ai/-/merge_requests/1009/diffs#f091c4... chatConsumer documentation was already added inside AI Core docs.

Production build 0.71.5 2024