Odoo Changelog — 2025-05-22

Thursday, May 22, 2025

2 changes · saas-18.3

Bug Fixes(2)

Resolved issues and error corrections

[FIX] account_accountant: hide redundant secondary buttons

Enterprisesaas-18.3account_accountant

Bank statement lines no longer show the same action in both the main and secondary button areas. This reduces visual clutter and helps users choose the right reconciliation action more easily.

Original PR description

Before this **PR**:
The same button could appear in both the primary and secondary button groups on a bank statement line, leading to unnecessary duplication.

After this **PR**:
If a button is already shown as a primary action, it is excluded from the list of secondary buttons to avoid redundancy.

**task**-4780841

#bank reconciliation#user interface#buttons

[FIX] ai: fix module behaviour when `vector` extension is missing

PR #85180

Enterprisesaas-18.3aiMEDIUM

The AI module now handles databases where the PostgreSQL vector extension is unavailable without blocking installation or crashing related screens. It also improves checks before using AI embedding features and provides clearer guidance in logs for administrators.

Original PR description

There were multiple issues when the PostgreSQL extension `vector` isn't available in the database: - The module couldn't be installed because, despite wrapping the extension creation in a `try ...…

There were multiple issues when the PostgreSQL extension `vector` isn't available in the database:
- The module couldn't be installed because, despite wrapping the extension creation in a `try ... except`, the SQL transaction was in a failed state in case of an error, any subsequent query causes an traceback. A savepoint is necessary to be able to recover from it.
- The same problem also occurred in `_get_similar_chunks` and `_cron_generate_embedding` when performing queries on the `embedding_vector`. Instead of using a savepoint (which could be costly and cause other problems), a new `_is_rag_enabled()` function has been introduced to first check if the column is present. This approach can't be used in the first case as it is expected that the column doesn't exist at that point (that's where we create the column in the first place).
- There was a missing check of the presence of the `embedding_vector` column in `_compute_attachment_processing_percentage`, causing a crash when opening any `ai.agent`.
- The log appearing when the extension isn't available didn't include a link to the `pgvector` project.

#ai#installation#database#stability

Before this **PR**: The same button could appear in both the primary and secondary button groups on a bank statement line, leading to unnecessary duplication. After this **PR**: If a button is already shown as a primary action, it is excluded from the list of secondary buttons to avoid redundancy. **task**-4780841

There were multiple issues when the PostgreSQL extension `vector` isn't available in the database: - The module couldn't be installed because, despite wrapping the extension creation in a `try ... except`, the SQL transaction was in a failed state in case of an error, any subsequent query causes an traceback. A savepoint is necessary to be able to recover from it. - The same problem also occurred in `_get_similar_chunks` and `_cron_generate_embedding` when performing queries on the `embedding_vector`. Instead of using a savepoint (which could be costly and cause other problems), a new `_is_rag_enabled()` function has been introduced to first check if the column is present. This approach can't be used in the first case as it is expected that the column doesn't exist at that point (that's where we create the column in the first place). - There was a missing check of the presence of the `embedding_vector` column in `_compute_attachment_processing_percentage`, causing a crash when opening any `ai.agent`. - The log appearing when the extension isn't available didn't include a link to the `pgvector` project.