pg_statviz 1.2 adds support for the upcoming PostgreSQL 19, capturing the new wal_fpi_bytes counter from pg_stat_wal. The release introduces a blocking locks analysis module that tracks lock wait types and soft blocks via pg_blocking_pids(), plus a new OpenAI-compatible AI provider option (--ai openai) that works with OpenAI or any compatible local/remote server. Default AI models are updated to claude-sonnet-5 and gemini-3.7-flash. The AI analysis stays fully optional, opt-in, and dependency-free unless explicitly enabled.

2m read timeFrom postgresql.org
Post cover image

Questions this post answers

What's new in pg_statviz 1.2 for PostgreSQL 19 support?

pg_statviz 1.2 adds support for the upcoming PostgreSQL 19 by capturing the new wal_fpi_bytes counter from pg_stat_wal in its snapshot_conf. This release also introduces a blocking locks analysis module that tracks lock types like relation, transactionid, and tuple, and counts soft blocks using pg_blocking_pids(), not just hard conflicts. Track PostgreSQL monitoring tool releases like this on daily.dev before upgrading your database stack.

How do I use a custom AI provider like a local LLM server with pg_statviz?

pg_statviz 1.2 adds an openai AI provider accessible via the --ai openai flag, which works with the OpenAI API itself or any other service or local server implementing that API. Configuration uses the OPENAI_BASE_URL and OPENAI_MODEL environment variables, and the openai package is added only to the [ai] extras, so zero-dependency installs remain unaffected. daily.dev helps engineers evaluating optional AI integrations in database tooling stay current.

What default AI models does pg_statviz 1.2 use for Claude and Gemini?

pg_statviz 1.2 updates its default AI models to claude-sonnet-5 for Claude and gemini-3.7-flash for Gemini. This applies only when the optional AI analysis feature is explicitly enabled with the --ai flag; without it, pg_statviz behaves exactly as before with no extra dependencies or external calls. Developers choosing default AI models for their tooling can follow updates like this via daily.dev.