[GNU ELPA] Minuet version 0.9.0

Version 0.9.0 of package Minuet has just been released in GNU ELPA.
You can now find it in M-x list-packages RET.

Minuet describes itself as:

  =========================
  Code completion using LLM
  =========================

More at https://elpa.gnu.org/packages/minuet.html

## Summary:

  - [Minuet](#minuet)
  - [Features](#features)
  - [Requirements](#requirements)
  - [Installation](#installation)
  - [Quick Start: LLM Provider Examples](#quick-start-llm-provider-examples)
    - [Ollama Qwen-2.5-coder:3b](#ollama-qwen-25-coder3b)
    - [OpenRouter Deepseek-V4-Flash](#openrouter-deepseek-v4-flash)
    - [Opencode Go Deepseek-V4-Flash](#opencode-go-deepseek-v4-flash)
    - [Deepseek Deepseek-V4-Flash](#deepseek-deepseek-v4-flash)
    - [Llama.cpp Qwen-2.5-coder:1.5b](#llamacpp-qwen-25-coder15b)
  - [API Keys](#api-keys)
  - [Selecting a Provider or Model](#selecting-a-provider-or-model)
    - [Understanding Model Speed](#understanding-model-speed)
  - [Prompt](#prompt)
    - [Prefix-First vs. Suffix-First](#prefix-first-vs-suffix-first)
  - [Configuration](#configuration)

## Recent NEWS:

# Version 0.9.0 (2026-07-31)

## Features

- **Recent Edit History for Duet**: Duet can now use recent edits to provide
  more relevant next-edit predictions.

# Version 0.8.0 (2026-05-18)

## Breaking Changes

- Make `minuet-before-cursor-filter-length` and
  `minuet-after-cursor-filter-length` dynamic depending on provider: FIM
  providers now default to `0` (no filtering), while chat providers retain their
  previous defaults.

## Features

- **Experimental Next Edit Prediction**: Add `minuet-duet` for experimental next
  edit prediction, and `minuet-duet-active-mode` to enable it.
- `transform` option can now be used with all providers (`openai`, `gemini`,
  `claude`), not just `openai_fim_compatible`.
- Change default `openai-compatible` model to `deepseek-v4-flash` with
  prefix-first style.
- Change default `openai` provider model to `gpt-5.4-nano`.

# Version 0.7.1 (2025-12-15)

## Breaking Changes

- Rename `minuet-auto-suggestion-block-functions` to
  `minuet-auto-suggestion-block-predicates`

## Features

- Preserved Suggestions During Multi-line Typing: Prefix-matched suggestions
  persist even while the user is typing multi-line text, to avoid unnecessary
  LLM requests and conserving resources.

# Version 0.7.0 (2025-12-11)

## Breaking Changes

- Default value for `minuet-before-cursor-filter-length` has been changed to `2`

## Features

- completion: Preserve remaining suggestion when using
  `minuet-accept-suggestion-line` for continuous suggestion acceptance
- claude: Change default max tokens to 256.
- claude: Update default model to `claude-4.5-haiku`.

## Fixes

- display: Keep user selected index when displaying more results (#47).
- display: Keep completion items order when new items arrived (#46).

# Version 0.6.0 (2025-08-11)

## Breaking Change

- Improve completion filtering with before/after context:
  - Refactors the completion filtering logic to be based on the longest common
    match.
  - Add a new `minuet-before-cursor-filter-length` config option to trim
    duplicated prefixes from completions based on the text before the cursor.
- Change default few-shot example: The default few-shot example has been updated
  to require the AI to combine information from before and after the cursor to
  generate the correct logic.
- Update default system prompt: The system prompt is refined to be more concise
  and provide clearer instructions to the AI on handling various completion
  scenarios like code, comments, and strings.

## Other

- Updated Gemini Authentication: Switched to using the `x-goog-api-key` header
  for Gemini API requests to align with upstream changes.

# Version 0.5.5 (2025-07-24)

This is primarily a maintenance release with minor feature updates and
improvements.

## Features

- Configurable Endpoints for Gemini and Claude: Users can now specify custom API
  endpoints for Gemini and Claude providers.
- The `chat-input-template` can now be a list of strings, which will be
  constructed into multi-turn conversations, allowing for more organized and
  complex prompt structures.
- Added `minuet-set-nested-plist` for easier manipulation of nested plists in
  configurations.
- The default model for the `openai-compatible` provider is now
  `devstral-small-2505`.
- The default model for the `openai` provider has been updated to
  `gpt-4.1-mini`.

## Fixes

- Switched from `delete-process` to sending a `SIGTERM` signal to gracefully
  cancel in-flight requests.

# Version 0.5.4 (2025-04-13)

## Features

- Add option to show error message on minibuffer and improve error message

## Bug Fixes

- Fix overlay positioning when not at the end-of-line

# Version 0.5.3 (2025-04-08)

## Features

- Added `transform` option for OpenAI-FIM-compatible providers.

  This feature enables support for non-OpenAI-FIM-compatible APIs with
  OpenAI-FIM-compatible provider, such as the DeepInfra FIM API. Example
  configurations are available in [recipes.md](./recipes.md).

# Version 0.5.2 (2025-04-03)

This maintenance release focuses on reducing the ELPA tarball size.
...
...