Upgrading to 1.7

RubyLLM 1.7 introduces a DB-backed model registry that replaces string model fields with ActiveRecord associations, adding richer model metadata, provider routing, and usage tracking. The release preserves backward compatibility while offering a recommended migration path for Rails apps.

RubyLLM 1.7 replaces simple string model fields with a DB-backed model registry implemented as ActiveRecord associations to improve data integrity and surface richer model metadata. The release notes emphasize that core functionality continues to work in legacy mode, but new and upgraded deployments can use the registry for structured model information. Fresh installs receive the model registry automatically, while existing 1.6 apps continue to function with string fields and will show a deprecation warning on Rails boot.

The model registry exposes richer metadata on model records. Examples in the documentation show fields and delegated methods such as name, context_window, supports_vision, and input_token_cost. Provider routing is supported by specifying a provider at creation time. The registry enables model associations and SQL-level queries such as joining chats to models and filtering by provider, and it supports delegated predicate methods for model capabilities. Alias resolution is documented so that a call supplying an aliased model id and a provider resolves to the canonical provider/model combination. Usage tracking is possible by joining models to chats and grouping and ordering by chat count for reporting.

The recommended upgrade path from 1.6 is to migrate to the model registry. The documented steps are to run rails generate ruby_llm:install, rails generate ruby_llm:migrate_model_fields, and rails db:migrate, then enable the registry in the initializer by setting config.model_registry_class = ´Model´. The migration creates the models table, loads models from models.json automatically, migrates existing string-based data to use foreign keys, and preserves existing columns by renaming the old model_id column to model_id_string. New applications should run the install generator and database migration to get the registry by default.

58

Impact Score

Google and other chatbots surface real phone numbers

Generative Artificial Intelligence chatbots are surfacing real phone numbers and other personal details, sometimes by pulling from obscure public sources and sometimes by inventing plausible but wrong contact information. Privacy experts say users have few reliable ways to find out whether their data is in model training sets or to force its removal.

U.S. and China revisit Artificial Intelligence emergency talks

Washington and Beijing are exploring renewed talks on an emergency communication channel for Artificial Intelligence as fears grow over the capabilities of Anthropic’s Mythos model. The shift reflects rising concern in both capitals that competitive pressure is outpacing safeguards.

Artificial Intelligence divides employers as hiring and headcount shift

U.S. hiring beat expectations in April, but employers remain split on whether Artificial Intelligence should drive layoffs, productivity gains, or internal redeployment. At the same time, candidate use of Artificial Intelligence is outpacing employer adoption in hiring, adding new pressure to screening and entry-level recruiting.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.