LLM Flex 0.2.3

Added

• Fresh Futureformed look. New app icon, a custom menu-bar glyph, and the
  brand purple (#6127F5) accent throughout the popover.
• Real app icons for Codex and Claude Code on the status cards (with an SF
  Symbol fallback).

Changed

• Clearer status cards. Codex and Claude Code now read as peer cards, each
  tagged FLEXED (currently switched by LLM Flex) or DEFAULT (on its own
  settings). Removed the endpoint line from the active card to tighten it up.

Fixed

• Profile row menu showed a redundant dropdown chevron next to the "…"
  button; removed it.

LLM Flex 0.2.2

Fixed

• Couldn't delete a profile. The delete confirmation used a system alert,
  which steals focus from the menu-bar popover and dismisses it before the
  Delete button's action runs. Replaced with an inline "Delete / Cancel"
  confirmation that stays within the popover.

LLM Flex 0.2.1

Fixed

• Crash on opening the menu in distributed builds. ProviderBadge loaded
  icons via SPM's generated Bundle.module, whose accessor fatalErrors when
  its hardcoded candidate paths miss — which they always do in a downloaded
  .app (the bundle lives in Contents/Resources, and the fallback path is the
  build machine's). Icons are now located resiliently and fall back to a
  monogram instead of crashing. Affected 0.1.0 and 0.2.0.

LLM Flex 0.2.0

Added

• Update notifications — on open, the app quietly checks GitHub for a newer
  release and shows a banner linking to the download when one is available.
  Pre-releases are included so alpha builds are covered.
• In-app feedback — an envelope button in the menu opens a feedback form
  that composes an email to llmflex@holdtight.cc, pre-tagged with your app and
  macOS version.
• Step-by-step getting-started guide in the README covering how to obtain a
  provider API key and use the app.

Changed

• Clearer status panel — each target (Codex, Claude Code) now shows the
  active provider, model, and endpoint on labelled rows, so it's obvious at a
  glance what's currently live.
• Menu polish — the feedback button is now labelled "Send feedback" (was a
  bare envelope icon), a "Help" link in the footer opens the GitHub guide, and a
  stronger tinted divider separates the status panel from the profiles list.

LLM Flex 0.1.0

First public alpha. A macOS menu-bar app for switching the model/provider used
by Claude Code and OpenAI Codex without hand-editing config files.

Added

• Profiles — define a provider, base URL, and model; switch between them
  from the menu bar. API keys are stored in the macOS Keychain, never in plain
  config or UserDefaults.
• Claude Code target — writes ~/.claude/settings.json (model + auth env),
  deep-merging so your other settings (permissions, hooks) are preserved.
• Codex target — writes a managed block in ~/.codex/config.toml, updates
  ~/.codex/auth.json, and sets OPENAI_API_KEY in the launchd environment.
  Blocks switching while Codex is signed in via ChatGPT to avoid clobbering.
• Providers — OpenAI, OpenRouter, LM Studio, Anthropic, Gemini, Ollama,
  Opencode Go, plus OpenAI-compatible and custom endpoints.
• Per-target compatibility chips — flags provider/target combos that won't
  work without a translating proxy (e.g. OpenAI ↔ Claude Code).
• Curated model catalog with a Quick Pick picker in the profile editor.
• Test Connection — validates a profile against the provider using its
  real auth scheme and surfaces the provider's error message on failure.
• Snapshot & restore — captures your original config files on first touch
  so changes can be fully rolled back.
• DMG packaging (scripts/package-dmg.sh) and tag-triggered GitHub Releases.