PackageTrack
Sign in Get early access

arb_ai

A CLI and Dart package for AI-powered, build-time translation of Flutter ARB files with smart-diffing and CI/CD ready.

0.2.1 RodotDev/arb_ai

What this package is like to depend on

Last release 2 months ago

03 Jun 2026

Too new to tell

only 1 release windows

Nearly every release is documented

notes for 3 of 3 stable releases

Nothing withdrawn

no release was ever pulled

2 months old

3 releases · first in 2026

3 releases in the last 12 months

see the full history below

Release timeline

3 releases · May 2026 to Jun 2026
Release Pre-release

Releases

latest 3
  1. 0.2.1 03 Jun 2026
    Release notes

    Changed

    • Updated default model from gemini-2.5-flash to gemini-3.5-flash after demonstrating stability in real-world tests.

    Fixed

    • Fixed ICU plural syntax warnings caused by overlapping categories (e.g., having both =1 and one) in target translations.
    • Added base-language fallback (e.g., pt for pt_BR) for target locales to enforce CLDR category validations correctly.
    Open source →
    Release notes

    Added

    Changed

    • Updated default model from gemini-2.5-flash to gemini-3.5-flash after demonstrating stability in real-world tests.

    Fixed

    • Fixed ICU plural syntax warnings caused by overlapping categories (e.g., having both =1 and one) in target translations.
    • Added base-language fallback (e.g., pt for pt_BR) for target locales to enforce CLDR category validations correctly.
    Open source →
  2. 0.2.0 03 Jun 2026
    Release notes

    Added

    • Early "fail-fast" validation for the AI provider API key, ensuring the CLI halts immediately with clear instructions if the key is missing, rather than parsing files and starting the pipeline only to fail midway.
    • Auto-formatting (pretty-print) for raw API JSON error responses in the terminal console to improve diagnostic readability.
    • Progressive batch saving that writes target .arb files and updates the state cache (.arb_ai_state.json) immediately after each batch of translations is completed and validated, preventing data loss and maintaining translation integrity.

    Changed

    • Updated default model from gemini-3.5-flash to gemini-2.5-flash for improved availability and stability.
    • Increased default batch size from 25 to 100 to optimize API quota consumption.
    • Expanded CLDR plural rules validation support to cover more languages: Czech (cs), Slovak (sk), Romanian (ro), Croatian (hr), Bosnian (bs), Serbian (sr), Lithuanian (lt), Latvian (lv), and Slovenian (sl).

    Fixed

    • Added translation retry error feedback, passing validation issues from previous attempts back to the LLM context to enable self-correction.
    • Implemented smart HTTP 429 rate limit backoff parser that respects the API's retryDelay and aborts if it exceeds 120 seconds.
    • Added automated retries for transient server errors (500, 502, 503) and instant fail-fast for fatal API errors (400, 401, 403).
    Open source →
    Release notes

    Added

    • Early "fail-fast" validation for the AI provider API key, ensuring the CLI halts immediately with clear instructions if the key is missing, rather than parsing files and starting the pipeline only to fail midway.
    • Auto-formatting (pretty-print) for raw API JSON error responses in the terminal console to improve diagnostic readability.
    • Progressive batch saving that writes target .arb files and updates the state cache (.arb_ai_state.json) immediately after each batch of translations is completed and validated, preventing data loss and maintaining translation integrity.

    Changed

    • Updated default model from gemini-3.5-flash to gemini-2.5-flash for improved availability and stability.
    • Increased default batch size from 25 to 100 to optimize API quota consumption.
    • Expanded CLDR plural rules validation support to cover more languages: Czech (cs), Slovak (sk), Romanian (ro), Croatian (hr), Bosnian (bs), Serbian (sr), Lithuanian (lt), Latvian (lv), and Slovenian (sl).

    Fixed

    • Added translation retry error feedback, passing validation issues from previous attempts back to the LLM context to enable self-correction.
    • Implemented smart HTTP 429 rate limit backoff parser that respects the API's retryDelay and aborts if it exceeds 120 seconds.
    • Added automated retries for transient server errors (500, 502, 503) and instant fail-fast for fatal API errors (400, 401, 403).
    Open source →
  3. 0.1.0 31 May 2026
    Release notes

    Added

    • Initial release of arb_ai.
    • Core CLI commands and options support (--dry-run, --check, --config, --force, --clean).
    • Smart Diffing engine with cryptographic .arb_ai_state.json hash matching to save translation costs.
    • Direct integration with the Gemini Native REST API (utilizing JSON schema constraints, safety flags, and exponential 429 backoff).
    • Rigid ICU validation supporting plural forms and CLDR rules across languages (Polish, Arabic, Portuguese, Russian, and more).
    • Auto-healing translation retry loop on failing ICU parser outputs.
    • Deterministic ARB writer for clean, git-friendly output file styling.
    Open source →
    Release notes

    Added

    • Initial release of arb_ai.
    • Core CLI commands and options support (--dry-run, --check, --config, --force, --clean).
    • Smart Diffing engine with cryptographic .arb_ai_state.json hash matching to save translation costs.
    • Direct integration with the Gemini Native REST API (utilizing JSON schema constraints, safety flags, and exponential 429 backoff).
    • Rigid ICU validation supporting plural forms and CLDR rules across languages (Polish, Arabic, Portuguese, Russian, and more).
    • Auto-healing translation retry loop on failing ICU parser outputs.
    • Deterministic ARB writer for clean, git-friendly output file styling.
    Open source →

Every package, every release, already written down.

The archive is open and free. Watching your own project is what we are building next.

Browse the archive