v0.3.0

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

• ssh: new top-level command group for SSH access to running workloads; run ssh setup once, then connect to inference deployments, environments, and training jobs over *.ssh.baseten.co
• model environment logs: full parity with model deployment logs (--tail, --start/--end/--since, --min-level, --includes/--excludes, --search-pattern, --replica, --request-id, text/json/jsonl)
• model environment metrics: full parity with model deployment metrics; metrics are aggregated across every deployment active on the environment, and series mode splits the window at each promotion
• org user, org team, whoami: read-only org user list / org user describe --user-id <id|me>, org team list / org team describe --team-id <id|name>, and a top-level whoami for the authenticated user

New Flags

• model deployment logs / model environment logs: --limit (default 5000, 0 = no limit) caps total lines pulled while paging backward through the window; rejected with --tail
• org user describe: --user-email as an alternative to --user-id
• org team describe: --team-name as an alternative to --team-id
• model deployment commands and model predict: --deployment-name as an alternative to --deployment-id
• model-api list: --added-only to restrict the listing to added Model APIs (see breaking changes)

Improvements

• Backward log pagination: model deployment logs / model environment logs now page backward through the full window instead of silently returning a single 500-line page; the default no-flag window resolves to the last 30 minutes, and a stderr note fires when --limit trims older lines
• Server-side name resolution: model, deployment, user, and team name/email lookups now resolve via server-side ?name=/?email= filters instead of listing-and-looping client-side; all resolvers error on both zero and multiple matches
• Richer model push summary: a facts card plus grouped, code-styled hints for logs, invocation, and SSH, with environment-scoped hints tagged (once deployed) when the deployment targets an environment
• Clickable output URLs in supporting terminals
• Homebrew install: releases now publish a formula to the basetenlabs/baseten tap
• model deployment logs (and model push --tail) moved off the deprecated POST logs endpoint to the GET endpoint

Fixes

• model push for scattered trusses: pushing a Truss with external_package_dirs no longer fails server-side build validation; the flag is stripped from the archived config.yaml (contents are already inlined under packages/), mirroring the Python CLI

💥 Backwards-Incompatible Changes

• model-api list --all removed: listing now defaults to the full catalog; use --added-only to restrict to added Model APIs

What's Changed

• Update README install version to v0.2.0 by @cretz in #24
• Add baseten ssh commands for workload SSH access by @cretz in #25
• Add CI, release, and license badges to README by @cretz in #26
• 💥 Replace -all with --added-only on model-api list by @cretz in #27
• Update README.md by @saimaddali in #29
• Add org user, org team, and whoami CLI commands by @cretz in #30
• Simplify README product name and pre-1.0 warning wording by @cretz in #31
• Publish Homebrew formula to basetenlabs/baseten tap on release by @cretz in #28
• Add model environment logs command and switch deployment logs to GET by @cretz in #32
• Paginate deployment and environment logs backward with a --limit cap by @cretz in #34
• Add environment SSH hostname form and richer push summary by @cretz in #35
• Add name/email flags and server-side filters to CLI commands by @cretz in #36
• Add model environment metrics command by @cretz in #39
• Clear external_package_dirs from pushed config so server build validation passes by @cretz in #37
• Make output URLs clickable in supporting terminals by @cretz in #38

Full Changelog: v0.2.0...v0.3.0