Smart Chat settings

Smart Chat lets you chat with models inside Obsidian, with practical controls for:

• Default instructions (so new chats start in the right mode)
• Keyboard send behavior (so writing multiline messages is painless)
• Streaming output (so responses feel fast)
• Context lookup (so retrieval pulls the right amount and type)
• Default chat model (so new threads use the model you expect)

Jump to:

• Smart Chat
• Lookup context
• Chat models

Where to find these settings in Obsidian

Open Settings -> Community plugins -> Smart Chat.

Smart Chat

Custom instructions

What it is

• Default custom instructions added to new chat threads.
• The UI copy notes: "This can be edited per-thread."

When to use it

• You want all new chats to start with a consistent stance (tone, format, constraints, or a project role).

Example (keep it short)

Be concise.
Ask 1-2 clarifying questions when needed.
Prefer actionable next steps.

Send modifier key

What it is

• The keyboard modifier required to send a message.

Why it matters

• If you write multi-line prompts often, this prevents accidental sends and makes "Enter" behave like a newline.

What it changes

• Example shown in the UI: Ctrl (Windows) / Cmd (Mac)

  • Typically means: press the modifier + Enter to send.

Stream responses

What it is

• When enabled, Smart Chat displays partial responses as they are generated.

Tradeoffs

• On: feels faster, you can start reading immediately.
• Off: you only see the final response once generation completes (can feel calmer if streaming distracts you).

Lookup context

These settings control how many items Smart Chat retrieves when you run a context lookup action, and what type of results it returns.

Results limit

What it is

• "Maximum number of context results to retrieve during context lookup action."

How to tune it

• Lower values: faster, less noise, smaller prompt payload.
• Higher values: more recall, better coverage, more tokens.

Result type

What it is

• Controls the kind of items returned during context lookup.
• The screenshot shows the dropdown set to Sources.

Typical meaning

• Sources: whole notes (coarser, usually smaller count).
• If your setup supports it, other granularities may appear (for example, block-level results).

Chat models

This section controls which model new chat threads use by default, and lets you add/edit/test model configurations.

Default chat model

What it is

• "Used as the default chat model when no other is specified."

How it behaves

• New chat threads start with this model unless you explicitly choose another model.

Current model card

What you see

• A "Current: provider - model" label (example shown: ollama - gemma3:4b)

• Provider and Model fields (example shown: Provider: ollama, Model: gemma3:4b)

• Buttons:

  • Edit: modify the current model configuration
  • Test: verify the current configuration works (connectivity/auth/model availability)

Add another model

• Click + New to create an additional chat model configuration.
• After adding multiple configurations, use the Default chat model dropdown to choose which one becomes the default.