feat(cli-v2): support --output with GitHub PR URLs in fern generate

[Swimburger]

Description

Refs internal Slack discussion (Niels)

Allows fern generate --output https://github.com/owner/repo/pull/123 to resolve the PR's head branch and push generated code directly to it, updating the existing PR instead of creating a new one. This eliminates the manual workflow of editing generators.yml to set mode: push and branch: <pr-branch> when iterating on a PR.

Changes Made

• gitUrl.ts: Added isGithubPrUrl() and parseGithubPrUrl() helpers; modified isGitUrl() to exclude PR URLs (both start with https://github.com/)
• resolveGithubPrBranch.ts (new): Fetches the PR's head branch and head repository via the GitHub REST API using fetch() with GITHUB_TOKEN/GIT_TOKEN. Uses head.repo.full_name for the URI so fork-based PRs resolve to the correct repository.
• command.ts:
  • Made parseTargetOutput() and getTargets() async to support the GitHub API call
  • When a PR URL is detected, resolves to { mode: "push", branch, uri } output config
  • Forces local generation when --output is a PR URL (same as git URL output)
  • Hoisted parseTargetOutput call out of .map() in getTargets() to avoid redundant API calls
  • Added isRemoteOutput guard so git/PR URLs are not resolved as filesystem paths in checkOutputDirectory or pipeline.run
• parseOutputArg.ts: Same PR URL handling for the no-config (flags-only) code path
• Updated --output CLI help text to mention PR URL support
• Added changelog entry

Human Review Checklist

• forceLocal propagation: In handleWithWorkspace, validateArgs receives args.local || forceLocal but runGeneration receives original args + separate forceLocal param — verify runGeneration respects the forceLocal flag independently (it does via isLocal on line 250)
• parseOutputArg.ts usage: Grep shows no imports of this file — confirm whether it is dead code or dynamically referenced
• outputPath: undefined: When isRemoteOutput is true, pipeline.run receives outputPath: undefined. Verify this doesn't cause issues downstream in resolveTargetOutput — the git output config on target.output.git should be the primary output mechanism in this case
• No retry/rate-limit handling on the GitHub API call — acceptable for v1?

Testing

• Unit tests added/updated
• Lint passes (pnpm run check)
• Manual testing completed — needs a real PR URL + GITHUB_TOKEN to verify end-to-end

Link to Devin session: https://app.devin.ai/sessions/9ec43fea26df4a57b4357ec778c86d8a
Requested by: @Swimburger

---

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[claude]

Claude Code Review

This repository is configured for manual code reviews. Comment @claude review to trigger a review and subscribe this PR to future pushes, or @claude review once for a one-time review.

_{Tip: disable this comment in your organization's Code Review settings.}