AI Features

SoloMD works fully without AI. Turn it on and it works inside your notes — with your key, or entirely on your machine.

Two ways to connect a model

Agent Panel: chat with your vault

• Open the right-sidebar Agent pane (requires an open workspace folder). Streamed multi-turn chat; every tool call the agent makes — reading a note, searching, writing — shows as an inline card.
• Writes are off by default. Enable “allow write” explicitly; a tool-loop cap keeps runs bounded.
• Insert drops a reply into the active note at the cursor; Copy takes it to the clipboard.
• Every run is recorded — the Trace View replays what the agent read and changed, and Git history lets you roll any change back.

Scheduled recipes

Recipes are saved agent prompts that run on a schedule (daily digest, weekly review, inbox triage…). Runs are logged like any other agent session and produce notes in your vault.

Inline writing tools

• AI rewrite — select text, press Ctrl/⌘ J, choose a rewrite (polish, shorten, translate, fix grammar…). The result replaces the selection; undo works normally.
• CJK proofread — flags common Chinese typos with one-click fixes.

Semantic (RAG) search

SoloMD can build a local vector index of the workspace and answer natural-language questions with citations to your own notes. The index lives on disk next to the workspace; re-index from the command palette after large imports.

Privacy stance

• No AI call ever happens without your explicit action; there is no background upload of note content.
• BYOK requests go directly to the provider you configured, with your key.
• The Ollama path keeps everything local.

Prefer driving from outside? The same vault is scriptable via the CLI and any MCP client via the MCP server.