Home

Hosomaki Roadmap

1. Short‑term

These are the features that are already taking shape and should land soon.

• doctor
• predict
• audit
• trace
• time‑bounded explain (--since, --until)
• context explain (--context)
• boot‑to‑boot diffs (--diff)

These are the “quality of life” improvements that make Hosomaki feel more like a real assistant than a log viewer.

2. Medium‑term

These features need more thinking, but the direction is clear.

• history
• alias
• watch
• compare
• why
• summarise
• JSON output for scripting and automation

This phase is about making Hosomaki more useful in daily workflows.

3. Memory Layer

• A bigger milestone, but one that unlocks a lot. For local-only.
• Local vector database (pgvector or SQLite + embeddings)
• Behavioural drift detection
• Better explanations for recurring issues
• Predictions based on machine’s own history

This is where Hosomaki starts to feel “aware” of the system over time. No cloud yet.

4. Daemon Mode

A background service (hosomakid) that keeps an eye on the system all the time.

• Real‑time anomaly detection
• Instant status and doctor responses
• Continuous predictions
• Proactive warnings

Now Hosomaki starts to look like a lightweight monitoring layer.

5. System‑level tools

These are smaller but very practical additions in my opinion:

• ports
• crons
• mounts
• timers
• env-check

6. Multi‑provider AI support

The goal is flexibility and maybe optionally using cloud (I am still considering this one).

• Local models (Ollama, the current one)
• Remote providers (OpenAI, Anthropic, etc.)
• Or no AI at all via Insight mode

Users choose the backend, Hosomaki adapts to this choice.

7. Long‑term vision

Just a smart, local companion for Linux systems, with a fully developed UI for people who prefer a visual tool instead of the CLI.