Skip to content

Local AI Visual Decision 1.2.0 - multi-model preview

Pre-release
Pre-release

Choose a tag to compare

@fnzbrn fnzbrn released this 13 Oct 10:06

Multi-model preview package 1.2.0 for the reviewed Macro Handler 1.0.52 multitask host.

  • Four separately installed, verified profiles: SmolVLM-256M/500M, LFM2.5-VL-450M and Qwen3-VL-2B.
  • Decision, bounded Number/Text, and explicit Arithmetic OCR tasks with answer variables and existing action branches.
  • 14-language themed settings, model memory guards, strict incomplete-output rejection, no cloud fallback or screenshot history.
  • Actual expected/actual device results are attached. Protocol success is not accuracy success: generative math was unreliable. Qwen message understanding worked on the owned sample but did not meet a universal 1–3 second target.

Read the guide and manifest before importing. The exact reviewed host source is caa175781ba055331bfb969efa43882a82632e8a. Model weights and native engines are not distributed in these assets. LFM requires its own license eligibility/consent. This preview does not replace the stable four-plugin release or claim Google Play publication.