Skip to content

v1.2.0 — Ghost Bench + Ollama discovery

Choose a tag to compare

View all tags
@C-K-Loan C-K-Loan released this 18 Apr 08:21
· 536 commits to main since this release
v1.2.0
49dee82
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

✨ What's new

📊 Ghost Bench — In-dashboard benchmark system

New Benchmarks tab in the dashboard. 14 built-in tasks (settings + navigation) you can run against any connected device or emulator. Live progress via Server-Sent Events, live screenshots, per-task pass/fail with reasoning.

This is Phase 1 of the AndroidWorld benchmark integration — Phase 2 will wrap the real google-research/android_world package for the academic standard benchmark.

7 new API endpoints:

  • GET /api/benchmarks/suites
  • GET /api/benchmarks/tasks?suite=ghost_bench
  • POST /api/benchmarks/runs
  • GET /api/benchmarks/runs
  • GET /api/benchmarks/runs/{id}
  • GET /api/benchmarks/runs/{id}/events (SSE live stream)
  • POST /api/benchmarks/runs/{id}/stop

🦙 Ollama quality-of-life

  • Live model discovery — dropdown shows actually-installed models (queries localhost:11434/api/tags on mount)
  • Background model pull — pull new models without blocking the UI, with live status tracking
  • Helpful errors when Ollama isn't running or a requested model isn't installed

📦 Install

pip install ghost-in-the-droid==1.2.0
# or
uvx --from ghost-in-the-droid==1.2.0 android-agent-mcp

Full changelog: v1.1.0...v1.2.0

Assets 2
Loading