v0.2.0

Download binaries below or see README for install

Highlights

⚠️ This is a pre-release and is not considered stable. CLI structure and output may change before GA.

New Commands

• model-api: list and call Baseten Model APIs (shared inference endpoints)
  • model-api list: list added Model APIs (--all browses the full catalog)
  • model-api describe --model <name>: show details for one Model API
  • model-api predict: send an inference request via --content (chat-completions from a message), --data (inline JSON), or --file; supports text, JSON, and JSONL streaming output
• model watch: watch a local model directory and live-patch its development deployment on each change, no full rebuild (--hot-reload, --no-keepalive)
• model deployment metrics: fetch deployment metrics with --mode current|summary|series (series renders as sparklines, or a table with --no-chart); window via --start/--end or --since (max 7d); filter with repeatable --metric
• org billing usage: render a usage table broken down by category (Dedicated, Model APIs, Training) with CATEGORY, MINUTES, TOTAL, CREDITS, SUBTOTAL columns and an "All" total row; --since accepts a duration (e.g. 30d)

New Flags

• model push: --develop (push to the mutable dev slot), --watch (live-patch after push; implies --develop), --watch-hot-reload, --watch-no-keepalive
• model deployment logs: --min-level, --includes/--excludes (repeatable substring filters), --search-pattern (RE2 regex), --replica, --request-id
• auth login: --profile (name the stored profile), --remote-url, --no-switch

Improvements

• Named auth profiles: auth moved from a host/label model to named profiles, selectable via --profile or BASETEN_PROFILE; auth switch changes the current profile and auth logout/auth status operate per-profile
• auth login: browser-based login now opens the browser automatically using the full device-flow verification URI

💥 Backwards-Incompatible Changes

• model push --promote removed: use --environment production instead
• auth login --label renamed to --profile
• Auth reworked around named profiles: auth logout/auth switch/auth status now operate on profiles rather than host/user pairs
• Commands renamed (fetch becomes describe):
  • model fetch becomes model describe
  • model deployment fetch becomes model deployment describe
  • model environment fetch becomes model environment describe

What's Changed

• Update README for v0.1.0 release by @cretz in #12
• Open browser and use complete verification URI for auth login by @cretz in #14
• Add per-platform install one-liners to README and CI version check by @cretz in #15
• Add filter flags to model deployment logs by @cretz in #16
• Replace host/label auth model with named profiles by @cretz in #17
• Add model-api commands (fetch, list, predict) by @cretz in #18
• Remove --promote flag from model push by @cretz in #19
• Add model deployment metrics command by @cretz in #20
• Teach org billing usage to stop panicking about money by @quinnypig in #21
• Add model watch and push --watch live-patch loop by @cretz in #22
• 💥 Rename fetch command verb to describe by @cretz in #23

Full Changelog: v0.1.0...v0.2.0