Skip to content

Releases: sudoeren/eckra

v1.2.9

31 May 16:24

Choose a tag to compare

v1.2.7 - Improved AI Commit Message Generation

31 May 14:15

Choose a tag to compare

What's Fixed

AI now consistently generates multi-line commit messages with detailed body sections.

Problem

AI was sometimes generating short, single-line commit messages instead of the expected multi-line format with subject and body.

Solution

Strengthened the AI prompt with:

  • 3 complete examples (few-shot learning) showing proper format:

    • JWT token validation
    • Dark mode contrast fixes
    • Database query optimization
  • Critical format requirements with explicit instructions:

    • Subject line (max 50 chars)
    • Empty line
    • Body with 2-4 bullet points
  • Stronger language: 'MUST have', 'Do NOT write single-line messages'

  • Increased token limit: 600 → 1000

  • Enhanced system message: 'ALWAYS generate multi-line messages'

Result

AI now reliably generates professional commit messages like:

`
feat(auth): add JWT token validation

  • Implement token verification middleware
  • Add refresh token rotation for security
  • Store tokens in httpOnly cookies
    `

Installation

npm install -g eckra@1.2.7

Changelog

  • fix: strengthen AI prompt for multi-line commit messages (#14)

Full Changelog: v1.2.6...v1.2.7

What's Changed

  • fix: strengthen AI prompt for multi-line commit messages by @sudoeren in #14

Full Changelog: v1.2.6...v1.2.7

v1.2.6 - VSCode Copilot-Style Commit Messages

31 May 14:08

Choose a tag to compare

What's New

VSCode Copilot-style commit messages with subject line and detailed body.

Multi-line Commit Messages

AI now generates professional commit messages in the format:

`
feat: add autocomplete search

  • Implement type-to-search in provider selection

  • Fetch models from each provider's API

  • Add fallback to manual input when API fails
    `

  • Subject line: Max 50 characters, imperative mood

  • Body: Explains what changed and why (bullet points, 72 char wrap)

Automatic Retry

AI automatically retries when it returns short or invalid responses:

  • Max 2 attempts with 500ms delay
  • Better error handling

UI Improvements

  • Only subject line shown in suggestion list
  • Full message (subject + body) displayed after selection
  • Edit only subject line, body is preserved
  • Same improvements in easy workflow mode

Prompt Enhancements

  • Detailed format instructions with examples
  • Increased token limits (100 → 400, 300 → 600)
  • Strengthened system message

Installation

npm install -g eckra@1.2.6

Changelog

  • feat: enhance commit message generation with detailed instructions and retry logic (#13)
  • feat: improve AI commit message generation and user interaction (#13)

Full Changelog: v1.2.5...v1.2.6

What's Changed

  • feat: improve AI commit message generation by @sudoeren in #13

Full Changelog: v1.2.5...v1.2.6

v1.2.5 - Autocomplete Search for Providers & Models

31 May 13:55

Choose a tag to compare

What's New

Autocomplete search for provider and model selection.

Provider Selection

Type to search through providers in Settings and Onboarding.

Model Selection

Models are fetched from each provider's API and can be searched in real-time:

  • OpenAI, Anthropic, Gemini, Ollama, OpenRouter, LM Studio
  • Falls back to manual input if API is unreachable

CI

  • Removed Node.js 18 from CI matrix (EOL April 2025)

Installation

npm install -g eckra@1.2.5

Changelog

  • feat: add autocomplete search for provider and model selection (#11)
  • chore: remove Node 18 from CI matrix (#12)

Full Changelog: v1.2.4...v1.2.5

v1.2.4

28 May 20:38

Choose a tag to compare

Full Changelog: v1.2.3...v1.2.4

v1.2.0

28 May 20:21

Choose a tag to compare

Full Changelog: v1.0.8...v1.2.0