FAQ
Frequently asked questions about OWL.
General
What is OWL?
OWL is a local AI daemon with persistent memory, project awareness, and an evolving personality. It runs entirely on your machine using Ollama.
Is my data sent anywhere?
No. Everything runs locally:
- Ollama for LLM inference
- SQLite for storage
- ChromaDB for vectors
No external API calls are made.
What models work with OWL?
Any Ollama-compatible model:
llama3.2(default)mistral,mixtralcodellamallama2,llama3- And many more
Can I use OpenAI/Anthropic APIs?
Currently only Ollama is supported. API providers may be added in the future.
Installation
How do I install OWL?
One command:
curl -sSL https://kavin-ps.io/owl/install.sh | bash
This installs Ollama, pulls models, and sets up OWL globally.
How do I upgrade OWL?
uv tool upgrade owl
How do I uninstall OWL?
uv tool uninstall owl
rm -rf ~/.owl # Optional: remove data
Usage
How do I switch models?
Edit ~/.owl/config.yaml:
llm:
model: mistral:latest
Then restart OWL:
owl stop
owl
How do I clear conversation history?
/new # Start fresh session
Can I use OWL with multiple projects?
Yes. Switch with /project <path>. History is scoped per project.
How do I backup my data?
Copy the ~/.owl directory:
cp -r ~/.owl ~/owl-backup
Memory & Learning
What does OWL remember?
- Memory file: Facts you explicitly tell it (
/remember) - Conversation history: Recent messages
- Learnings: Auto-extracted insights
- Knowledge base: Documents you've added
How does evolution work?
After 10+ learnings accumulate, use /evolve to trigger character evolution. OWL synthesizes learnings into an updated personality description.
Can I edit the soul directly?
Yes:
vim ~/.owl/soul.yaml
owl stop && owl # Restart to reload
Tools
Why can't OWL write files?
Check your profile:
/tools
Switch to coding or full:
/tools coding
Are shell commands safe?
Shell commands (run_command) are only available in full profile. Use with caution - OWL can run any command you could.
Can I add custom tools?
Not yet through configuration, but you can modify owl/tools/file_ops.py and register new tools in owl/tools/registry.py.
Technical
Where is data stored?
~/.owl/
├── config.yaml # Configuration
├── soul.yaml # Personality
├── memory.md # Memory file
├── memory/owl.db # SQLite database
└── knowledge/ # ChromaDB vectors
How much disk space does OWL use?
Minimal. Typical usage:
- Base: ~1MB
- Per conversation: ~10KB
- Knowledge base: ~1KB per chunk
Can I run multiple OWL instances?
Not recommended. The daemon uses a single socket. Multiple CLI instances can connect to one daemon.
How do I debug issues?
- Check daemon output for errors
- Enable debug logging in
~/.owl/config.yaml:
daemon:
log_level: DEBUG
- Restart OWL
Contributing
How can I contribute?
OWL is open source. Contributions welcome:
- Bug reports
- Feature requests
- Pull requests
- Documentation improvements
Where do I report bugs?
GitHub Issues: github.com/kavinps/owl/issues