Skip to content

v0.1.2

Latest

Choose a tag to compare

@IsuraManchanayake IsuraManchanayake released this 17 Sep 06:07

Local 0.1.2

Local 0.1.2 puts thinking under your control, keeps long coding sessions
alive, adds a recoverable Trash for your sessions, and shows at a glance
which sessions are still working.

Highlights

  • Choose how much a model thinks. A Thinking control beside the model
    selector sets the reasoning level for your next reply, for models whose
    chat template supports it: off, on, or a level such as low, medium and
    extra high. Local reads the available levels and the model's own default
    from the loaded model itself, remembers the level you last used with each
    model, and says plainly when a model has no thinking controls or is still
    loading.
  • Coding sessions that go the distance. The engine now sizes the context
    window itself, and window sizes pinned in earlier releases are cleared
    once so they can't hold a session back. When a long session fills its
    window, the coding agent summarizes its earlier steps and carries on
    instead of going quiet, and the transcript shows that work as it happens.
    Failures say what went wrong.
  • A recoverable Trash for sessions. Deleting a chat, coding session or
    meeting moves it to the Trash, with Undo and thirty days to change your
    mind. Restore one, restore all, delete permanently, or empty the Trash.
    Select many at once with Command-A, Shift-click and the Delete key.
    Coding project folders and their files always stay on disk.
  • See what's still working. In the sidebar's recent list and on the
    sessions page, a session's icon breathes while it's generating and stays
    lit when a reply, a coding turn or a meeting write-up finished while you
    were looking elsewhere, until you open it.
  • Math that covers what models actually write. Chat math now typesets
    far more LaTeX: matrix and alignment environments, cases, cancels and
    boxed answers, labelled arrows, modular arithmetic, colored expressions,
    and a much wider range of symbols. Primes and native fractions survive
    intact, and display math wider than the transcript scrolls, with a cue at
    its edge.
  • Runs on macOS 14 and newer. Chat and Coding work on any Apple Silicon
    Mac running macOS 14 or later. Recording a meeting needs macOS 26, whose
    Speech framework provides the on-device transcription; on older systems
    Meetings says so.

Fixes and polish

  • Thinking blocks use the same Markdown renderer and animations as replies,
    sit where they belong, and keep consistent spacing in saved conversations.
  • One signal per activity: the status line steps aside when a reasoning
    header or a running tool row already shows the work, and reasoning
    headers count tokens live and keep their total.
  • Coding tool rows say what a call does, such as the file and its size, the
    number of edits, the search pattern or the address, with the tool's name
    set apart from its arguments, and without a progress counter that isn't
    really counting.
  • Answers in coding transcripts get breathing room from the tool runs and
    reasoning headers around them.
  • Every completed coding response has a Copy control, including older
    sessions saved without statistics.
  • A finished reply shows a placeholder while it's still being parsed instead
    of a blank.

Requirements

  • Apple Silicon Mac, macOS 14 or newer. Recording meetings needs macOS 26.
  • Local installs and manages the BaseRT runtime for you.

Local checks for updates and tells you when a new version is available.


Signed and notarized for macOS 14+ (Apple Silicon).

Download: https://downloads.basecompute.co/local/v0.1.2

a8372a75629bf9de8993a9cf60471d53aa88d943900b568f9fc3f10bd060d322  Local-0.1.2.dmg
a8372a75629bf9de8993a9cf60471d53aa88d943900b568f9fc3f10bd060d322  Local.dmg