39 lines
1.5 KiB
Markdown
39 lines
1.5 KiB
Markdown
|
# Use The AI
|
||
|
|
|
||
|
|
Tolaria gives you two ways to ask for AI help: open the AI panel for an ongoing conversation, or prompt directly from the editor with `Cmd+K` followed by a space.
|
||
|
|
|
||
|
|
## Choose How To Prompt
|
||
|
|
|
||
|
|
- **AI panel** is best for longer conversations, agent work, and requests that need visible back-and-forth.
|
||
|
|
- **Inline prompt** is best when you are already writing. Press `Cmd+K`, type a space, then write the prompt you want the AI to handle from the current note context.
|
||
|
|
|
||
|
|
## Choose A Target
|
||
|
|
|
||
|
|
Open Settings and choose the default AI target:
|
||
|
|
|
||
|
|
- **Coding agent** for tool-backed vault editing through Claude Code, Codex, OpenCode, Pi, or Antigravity CLI.
|
||
|
|
- **Local model** for Ollama or LM Studio chat over note context.
|
||
|
|
- **API model** for OpenAI, Anthropic, Gemini, OpenRouter, or an OpenAI-compatible endpoint.
|
||
|
|
|
||
|
|
If a coding agent is missing, install it and reopen Tolaria or switch to another target.
|
||
|
|
|
||
|
|
## Permission Mode
|
||
|
|
|
||
|
|
Coding agents support per-vault permission modes:
|
||
|
|
|
||
|
|
- **Vault Safe** keeps agents limited to file, search, and edit tools.
|
||
|
|
- **Power User** can allow shell commands for agents that support them.
|
||
|
|
|
||
|
|
Direct model targets always stay in chat mode. They can use note context, but they cannot edit vault files through tools.
|
||
|
|
|
||
|
|
## Good Requests
|
||
|
|
|
||
|
|
- "Find notes related to this project."
|
||
|
|
- "Summarize what changed in this note."
|
||
|
|
- "Draft a weekly review from these linked notes."
|
||
|
|
- "Update this checklist based on the current project status."
|
||
|
|
|
||
|
|
## Review Changes
|
||
|
|
|
||
|
|
AI edits are file edits. Review them with Tolaria's diff and Git history before committing.
|