The release candidate llm 0.32rc1 finalizes work that began in llm 0.32a0. The update introduces a redesigned schema intended to capture more detailed metadata about prompts and responses produced by recent model families.
The most significant change is the use of content-addressable hash IDs for stored messages. This brings two practical benefits:
- it enables deduplication in the database by storing identical message content only once,
- and it allows llm to represent trees of messages for forked conversations.
Because this is a substantial schema change — implemented by adding new tables only — existing data should not be affected. Nevertheless, the maintainers recommend making a backup of your current logs database before upgrading to the RC. The suggested command is:
llm logs backup logs-backup.db
The RC also adds support for three gpt-5.6 variants: gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna.
In short, llm 0.32rc1 focuses on more accurate internal representation of prompts, responses and conversation structure while extending model compatibility; perform a logs.db backup prior to upgrading as a precaution.
