Included Tools

GoodLife ships a curated AI research stack pre-configured and offline-capable on first boot. Everything below is included by default; you don’t need to pacman -S your way through a setup. All tooling is local-first; no telemetry, no cloud defaults.

Local LLM inference

Tool	Notes
Ollama	Pre-configured. qwen2.5:7b and llama3.2:3b pre-pulled (~6.5 GB) — usable offline on first boot
llama-cpp	Fallback inference engine for restricted-RAM nodes
vLLM (opt-in)	High-throughput batching server for multi-user setups
synos-cortex-q	v53 Quantumweave MPS (matrix-product-state) inference path

ollama run qwen2.5:7b              # interactive chat, local
ollama list                        # see what's pulled
synos-alfred chat --backend ollama # ALFRED on top of Ollama

Notebook & analysis

Tool	Use
JupyterLab	Default notebook IDE, pre-configured kernels
Quarto	Reproducible analysis docs, publish to PDF/HTML
VSCodium	Editor with notebook support
Polars	DataFrames — faster than pandas for large data
DuckDB	In-process OLAP database; SQL on parquet/arrow
Pandas	Available; not the default for new code

Deep learning frameworks

Framework	Notes
PyTorch	Default. CUDA + ROCm + CPU builds available
TensorFlow	Available
JAX + Flax	Available
scikit-learn	Classical ML

GPU support is automatic when a discrete GPU is present; CPU-only is the fallback path. The Salvage Yard mesh deployment scenario (4× i5-class laptops, no GPU) is a first-class supported config.

NLP & embeddings

Tool	Notes
Transformers (HF)	Hugging Face model hub client + local cache
sentence-transformers	Embedding models with offline cache
synos-nlp-pipeline	Syn_OS-native pipeline with hippocampus integration
synos-datalake	Curated local corpus for RAG
LanceDB	Embedded vector DB
Qdrant (containerised)	Standalone vector DB option

Voice & audio

Tool	Use
whisper.cpp	Local speech-to-text — fast on CPU
Piper TTS	Local text-to-speech
PortAudio + SoX	Audio capture / processing

ALFRED in GoodLife defaults to text I/O; voice is opt-in.

ALFRED (research-mode)

~/.config/alfred/research.toml

[mode]
default = "advisory"
allow_promotion = false

[backends]
default = "ollama"

[privacy]
job_hunt_mode = true
context_isolation = "tenant"

Cloud backends (Claude, OpenAI, Gemini, DeepSeek) are available — but off until you configure keys. Privacy-first job-hunt mode keeps each conversation in its own tenant scope so résumé context doesn’t bleed into general chat.

Mesh tooling (ARCANUM Hive)

Tool	Use
synos-hive-bootstrap	Bring up a mesh from an empty node (Stoneglass v55)
arcanum-controller (k8s)	Workload scheduler across the mesh
Tailscale + WireGuard	Mesh transport
Ollama hive mode	Sharded inference across mesh nodes

Bevy desktop plugins (GoodLife subset)

Plugin	Function
Cutscene	Typewriter dialogue / onboarding
Mindmap	Visualise concept graphs from notes / transcripts
RetroFilter	Aesthetic CRT/scan-line overlay (toggle)
Cyberspace	3D file/process visualisation
SkillTree	Capability map (research-mode variant)
Rehoboam	System monitor overlay (CPU/RAM/GPU/inference stats)
Twin (v51)	Digital-twin substrate visualisation

The FactionHQ plugin is disabled on GoodLife (no GRIMOIRE faction system on this profile).

What’s not installed (by design)

• offensive security tooling (metasploit, aircrack, the C2 frameworks)
• Distrobox/BlackArch on-demand expansion (Master only)
• the GRIMOIRE lab catalogue
• the Master-tier RaaS engine + finding ranker + digest renderer

If you want any of these, you’re in the wrong profile — see Three ISOs.

Where to dig in next

• GoodLife Overview → — what this profile is for
• ARCANUM Mesh → — distributing inference across salvaged hardware
• ALFRED → — the consciousness fusion engine in research-mode