Skip to content

v0.11.5

Choose a tag to compare

View all tags
@snonux snonux released this 07 Feb 19:48
· 402 commits to main since this release
v0.11.5
6feda08

Release v0.11.5

Release Notes

Overview

This release focuses on further improving configuration flexibility, reliability, and user experience. It adds more powerful environment/model override logic, fixes the default Ollama model, refines unit tests to catch regressions, and enhances internal code structure for maintainability. Users benefit from smoother model switching (even across providers), better error handling, and minor but meaningful bug fixes.

Improvements

Flexible Model Overriding

  • Smarter Model/Provider Selection via Environment Variables:
    The configuration now supports much more flexible model and provider selection. New logic for HEXAI_MODEL_FORCE / HEXAI_MODEL allows you to easily override the model on a per-provider basis or globally, without having to edit config files. This applies even when switching between OpenAI, Ollama, or Copilot providers.

    • Why it matters: Users automating workflows, working with different API keys/environments, or scripting Hexai now have reliable, robust ways to set models via env vars.
    • Migration: No changes required for most users; consult updated docs if you need advanced overrides.

  • OpenAI Model Presets:
    The config now allows [openai.presets] for conveniently mapping friendly names to real model names—great for teams or scripts using multiple models.

Ollama Default Model Correction

  • Fixed Default Model Name:
    The default Ollama model is now correctly set to qwen3-coder:30b-a3b-q4_K_M (removed erroneous backtick).
    • Why it matters: Avoids confusing errors if you rely on the built-in default when no model is provided.

Usability, Consistency, and Developer Experience

  • Cleaner CLI Input Handling:
    The hexai CLI now properly propagates and reports stdin read errors, making problem diagnosis easier for power users and downstream developers.

  • Expanded Unit Test Coverage:
    Many LSP, CLI, and handler tests add coverage for server defaults and new code paths, helping to prevent subtle regressions and solidify overall reliability.

  • Code Quality & Refactoring:

    • Internal code modularization and de-duplication, including:
      • Server helpers now consistently use server configuration (no more global state).
      • Refactored LSP completion logic, improving maintainability and future readability.

Bug Fixes

  • Ollama Default Model Typo Fixed:
    Addresses failed requests to Ollama when no model explicit name is set.

  • Custom Actions/Prompts in TUI & LSP:
    Updates the way selection and prompts are extracted and handled for rewrite and custom actions, eliminating spurious errors or "echoing input" messages in edge cases.

  • Consistent Inline Prompt Handling:
    All completion and inline tag helpers now correctly respect the current server’s configured characters/prefixes, avoiding inconsistencies between tests and runtime.

  • History Handling & Code Actions:
    Chat trigger and code action extraction are configuration-aware, so customized triggers and comments work as expected.

  • Better Error Messages:
    CLI now provides clear error messages if stdin cannot be read, especially useful in scripts or testing environments.

Breaking Changes / Migration

  • No breaking changes expected.
  • Environment Variable Model Selection:
    If you previously used only HEXAI_OPENAI_MODEL, switching providers or models with scripting is now easier and more robust—review the documentation for details.

Documentation & Examples

  • Configuration and usage documentation are updated to reflect the new model selection scheme and provide more examples of environment override behavior.
  • All relevant README/test files and example configs are updated.

Thank you for your feedback and contributions!
For details, advanced configuration, or help, see configuration docs and the project's README.

Assets 2
Loading