v0.9.0

✨ New: AWS Bedrock provider

ai-i18n now supports AWS Bedrock as a model provider, alongside Anthropic, OpenAI, and Ollama (#12).

It uses the model-agnostic Bedrock Converse API, so any Converse-capable model works — Anthropic Claude, Meta Llama, Amazon Nova/Titan, Mistral, Cohere, and more.

Usage

- uses: aws-actions/configure-aws-credentials@v4
  with:
    role-to-assume: arn:aws:iam::123456789012:role/bedrock-translate
    aws-region: us-east-1

- uses: i18n-actions/ai-i18n@v0.9
  with:
    provider: bedrock
    aws-region: us-east-1
    model: anthropic.claude-3-haiku-20240307-v1:0   # or meta.llama3-70b-instruct-v1:0, etc.
    target-languages: de,fr,es
    files: 'src/locales/**/*.xliff'

Details

• Auth: aws-region is required. Explicit aws-access-key-id / aws-secret-access-key / aws-session-token inputs are optional and fall back to the default AWS credential chain (so aws-actions/configure-aws-credentials / OIDC works out of the box).
• Default model anthropic.claude-3-haiku-20240307-v1:0; per-provider rate limits added.
• Reuses the existing prompt-building, JSON parsing, batching, rate limiting, and retry logic.

Note: Ensure the target model is enabled in your account's Bedrock Model access for the chosen region. Some models require a cross-region inference-profile ID (e.g. us.anthropic.claude-…) rather than the plain foundation-model ID.

Full changelog: v0.8.0...v0.9.0