Skip to content

v0.9.63 — AI translation: commit to translated Play Store notes in one flag

Choose a tag to compare

View all tags
@yasserstudio yasserstudio released this 20 Apr 12:05
· 222 commits to main since this release
v0.9.63
f89280b

What's Changed

gpc changelog generate can now translate non-source locales through your own LLM key. Add --ai to the existing play-store flow — the [needs translation] placeholder that v0.9.62 emitted becomes real translated text.

gpc changelog generate --target play-store --locales auto --ai
gpc changelog generate --target play-store --locales fr-FR,es-ES --ai --provider openai

The resolver auto-detects whichever provider key you have set — AI_GATEWAY_API_KEY, ANTHROPIC_API_KEY, OPENAI_API_KEY, or GOOGLE_GENERATIVE_AI_API_KEY — and picks sensible per-provider model defaults (claude-sonnet-4-6, gpt-4o-mini, gemini-2.5-flash). Override with --provider and --model. Reasoning-model defaults are explicitly avoided so you don't pay for thinking tokens on a translation task.

Gateway-primary routing: if AI_GATEWAY_API_KEY is set, requests route via the Vercel AI Gateway — 20+ providers through a single interface, aggregate cost in USD reported per run, auto-fallback via models: [...]. Otherwise falls back to direct SDKs.

All four AI SDK deps are lazy-loaded. gpc changelog generate without --ai imports none of them — the cold-start budget is preserved and users without an LLM key pay zero startup cost.

Third in the changelog-generation series. v0.9.64 closes the loop with --apply to write the translated bundle straight into your draft release.

  • feat: --ai flag on gpc changelog generate --target play-store — translate non-source locales via your own LLM key
  • feat: --provider <anthropic|openai|google> and --model <id> — override the auto-detected provider and per-provider defaults
  • feat: --dry-run — print the resolved config and prompt that would be sent, without calling the LLM or spending tokens
  • feat: --strict extended — exits 1 if any locale fails to translate (network, rate limit, safety, auth)
  • feat: Gateway-primary routing via the Vercel AI Gateway when AI_GATEWAY_API_KEY is present — unlocks 20+ providers plus cost-per-run tracking
  • feat: JSON output adds an ai block with path, provider, model, tokensIn, tokensOut, plus runId and costUsd on the Gateway path
  • feat: structured error-reason classification — rate_limited, auth, safety_blocked, timeout, network, no_source, unknown. Sensitive error details never reach your Play Store draft.
  • feat: new @gpc-cli/core exports — translateBundle, createTranslator, resolveAiConfig, fetchAggregateCost, classifyError, formatPathLabel

Tests: 2,037 (+38). Endpoints: unchanged at 217. Runtime deps: four added (all lazy-loaded) — ai, @ai-sdk/anthropic, @ai-sdk/openai, @ai-sdk/google.

Full Changelog: v0.9.62...v0.9.63

Assets 8
Loading