Skip to content

v0.1.6

Choose a tag to compare

View all tags
@github-actions github-actions released this 25 Apr 07:13
· 109 commits to main since this release
v0.1.6
ea9a0f5
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

What Changed

This release repositions NEXUS as a unified AI toolkit layer — one config for every AI CLI you use — and ships foundational features for the v0.2.0 observability milestone.

New Features

  • TUI: Task Log screen — view MCP tool invocations with model, latency, and pass/fail status. Access from the main menu → Task Log.
  • TUI: GPU auto-detection — Health Check now detects your GPU (NVIDIA, AMD, Apple Silicon) and recommends the right model band for your hardware.
  • MCP: Fast-path routing — commit messages for trivial diffs (renames, version bumps, lock files, deletions) are generated instantly without calling Ollama.
  • MCP: Structured output — commit message generation uses Ollama's native JSON schema support for more reliable output.
  • MCP: Task logging — all tool invocations are logged to ~/.config/nexus/logs/mcp-tasks.jsonl with timing and status data.
  • Onboarding Guide persona — new persona that walks users through hardware detection, model selection, and setup verification. See personas/onboarding-guide.md.

Documentation

  • README rewritten — subtitle, Philosophy, and "What NEXUS Does" sections reframed around config unification and subtraction-first messaging.
  • Cloud Billing Guide — new guide explaining cost optimization through local model routing, with links to official API pricing pages. See docs/cloud-billing-guide.md.

Fixes

  • Close HTTP response body in health check (connection leak)
  • Handle Intel Macs in Apple Silicon detection (no false positives)
  • Handle multi-GPU NVIDIA systems (parse first GPU, split by last comma)
  • Truncate long tool/model names in Task Log to preserve column alignment
  • Safe .env update in onboarding persona (no destructive overwrite)
  • Secure-by-default Ollama install recommendation (inspect-first)

Infrastructure

  • CI parallelized — test-install-cycle no longer waits for build-and-test, cutting wall-clock CI time roughly in half.
  • macOS added to CI build and install test matrix (v0.1.5).
  • Windows binaries removed from releases until #18 ships.

Hardware Benchmarks

  • ⏳ Low tier — pending
  • ⏳ Mid tier — pending
  • ⏳ High tier — pending

Benchmark infrastructure is tracked in #75. Results will be added to these notes as they become available.

Full Changelog: v0.1.5...v0.1.6

Assets 7
Loading