bump-downstream

name	bump-downstream
description	Tag the latest atuin-ai-core commit and update the dependency pin in the downstream consumers (atuin_hub's gleam_cli_chat and atuin-ai-server), cutting a branch in each. Use when a new atuin-ai-core version needs to be rolled out to its consumers.
disable-model-invocation	true

Bump atuin-ai-core in downstream projects

Release the current atuin-ai-core HEAD as a tagged version and update the two downstream consumers to point at it:

• atuin_hub — gleam_cli_chat/gleam.toml pins atuin_ai_core as a git dependency by ref (a vX.Y.Z tag); gleam_cli_chat/manifest.toml locks the commit SHA.
• atuin-ai-server — mix.exs pins :atuin_ai_core as a git dependency by tag:; mix.lock locks the commit SHA.

Arguments

ARGUMENTS: $ARGUMENTS

The user may pass project directories and a version bump, in any order, as key=value pairs or plain prose:

• core=<dir> — atuin-ai-core checkout (defaults to the current repo if it is atuin-ai-core, i.e. its gleam.toml has name = "atuin_ai_core")
• hub=<dir> — atuin_hub checkout
• server=<dir> — atuin-ai-server checkout
• bump=<major|minor|patch> or an explicit version like v0.3.0

If the hub or server directory is not given and cannot be found at an obvious sibling path of the core checkout (e.g. ../atuin_hub, ../atuin-ai-server — verify by checking for gleam_cli_chat/gleam.toml and mix.exs respectively), ask the user for it. Do not guess.

Steps

1. Tag atuin-ai-core

In the core directory:

1. Confirm the working tree is clean and the checkout is on main, up to date with origin/main (git fetch first). If there are unpushed or uncommitted changes, stop and ask the user how to proceed.
2. Check whether HEAD is already tagged: git tag --points-at HEAD.
   • If it has a vX.Y.Z tag, use that tag and skip to step 2 — no new tag.
   • Otherwise, find the latest version tag (git tag --sort=-v:refname | head -1) and compute the next version. Use the bump the user specified; if they didn't specify one, ask whether this is a patch, minor, or major release — don't assume.
3. Create the tag on HEAD (git tag vX.Y.Z) and push it (git push origin vX.Y.Z). The downstream updates fetch from GitHub, so the tag must be pushed before continuing.

Record the tag name and the commit SHA it points at.

2. Update atuin_hub

In the hub directory:

1. Confirm a clean working tree; fetch and start a branch off the latest default branch, e.g. git switch -c <user>/bump-ai-core-vX.Y.Z origin/main. Follow the repo's existing branch-naming convention if one is evident from recent branches.
2. In gleam_cli_chat/gleam.toml, update the ref of the atuin_ai_core dependency to the new tag.
3. From gleam_cli_chat/, refresh the lockfile: gleam deps update atuin_ai_core (fall back to gleam deps download if that fails). Verify manifest.toml now records the expected commit SHA from step 1.
4. Build to confirm the update compiles: gleam build in gleam_cli_chat/.
5. Commit gleam.toml and manifest.toml with a message like Bump atuin_ai_core to vX.Y.Z.

3. Update atuin-ai-server

In the server directory:

1. Same as above: clean tree, branch off the latest default branch.
2. In mix.exs, update the tag: of the :atuin_ai_core dependency to the new tag.
3. Refresh the lockfile: mix deps.update atuin_ai_core. Verify mix.lock records the expected commit SHA.
4. Compile to confirm: mix compile.
5. Commit mix.exs and mix.lock with a message like Bump atuin_ai_core to vX.Y.Z.

4. Report

Summarize: the tag created (or reused) and its SHA, and the branch + commit in each downstream repo. Do not push the downstream branches or open PRs unless the user asks.

Notes

• If a downstream repo already pins the target tag, say so and skip it rather than making an empty commit.
• If any build or lockfile refresh fails, report the error and stop rather than committing a broken pin.