Getting Started with Smart Connections

Connections view

Use the Connections view to see relevant content based on the current note.

Opening the Connections view

Click the Connections view icon (circle) or open the command palette to select one of the Connection view commands (rectangle).

Using the Connections view

Connections view results update automatically when you change notes. The name of the current note is located in the bottom-left of the Connections view.

Result score (underlined)

The result score is based on the semantic similarity between the result and the current note. The value and range will change depending on the embedding model being used.

Show/hide content (expand or collapse)

Results can be expanded or collapsed within the Connections view. The button in the row at the top can be used to expand/collapse all results.

Updating the results (refresh)

Use the refresh button to update the embedding for the current active note and re-generate the connections results.

Lookup query (semantic search)

Opens the Lookup pane to make a semantic query.

Interacting with the Connections view

Creating links from the Connections view is as easy as dragging a result into an open note. Holding ⌘/ctrl while hovering the mouse over a result will show the Obsidian native Hover Preview.

Hide connections

Right-click a connection to open the menu and hide it from the results.

Unhide connections

Right-click any connection to open the menu and unhide all previously hidden connections.

Inline connections (Early Release)

Hover the connections icon that appears inline with your content to surface a pop‑over of semantically related blocks from across your vault.
Hold ⌘/Ctrl while hovering over a result (blue highlight) to trigger Obsidian’s Hover Preview (displayed in pink). Click a result to open its note in a new tab.

Smart Chat

Opening the Smart Chat

Click the Chat view icon (circle) or open the command palette to select the Chat view command (rectangle).

Using the Smart Chat

Smart Chat leverages context from your notes. Context can be added both manually and through AI guided semantic lookup.

New chat (green squares)

Open notes are automatically added to the chat context

Chat history and conversation names (orange squares)

Chats can be named directly in Smart Chat. Click the chat history icon to access past chats.

System prompt (blue square)

System prompts can be added alongside the user chat input and will be handled differently depending on the currently configured chat model.

Adding context to the Smart Chat

Open the context builder by clicking the "Edit context" button and then select "Done" to insert the selected context into the conversation.

Show connections (connections icon)

Click the connections icon next to any context item to show a list of connections that can be added as context.

Show links (link icon)

Click the link icon next to any context item to show a list of links that can be added as context.

Remove context (x)

Individually remove context items by clicking the "x" button to the left of the item. Clicking the "New" button will remove all context items.

Trigger retrieval with self-referential pronouns

Use a self-referential pronoun to trigger note retrieval (lookup context action) from within a conversation.

Review the retrieved context and send

Retrieved notes are displayed before the AI finishes responding. Context items may be added or removed before continuing. Click the "Send" button (blue square) to complete the request based on the context.

The language setting determines which pronouns are detected to trigger the lookup. Both self-referential pronoun detection and reviewing context before completion are also configurable via the settings.

Drag files into the conversation to add as context

Include files in the chat using drag and drop.

Notifications & Settings

Process Notifications

Gain insight into what's happening under-the-hood.

Pause embedding process (green)

After clicking pause a new notification with a "Resume" button will appear.

Mute notification (blue)

Most notification are mutable. Notifications can be un-muted from the settings.

Muted notices in the settings

Muted notices will appear at the bottom of the Smart Connections settings. Use the "Remove" button to un-mute a notice.

Settings

Customizing the Connections view

Show full path

Toggle this to show the full file path for each result in the Connections view. Helpful for disambiguating similarly named notes in different folders.

Render markdown

When enabled, renders markdown (e.g., bold, links, lists) directly in the preview of connection results, making context easier to interpret at a glance.

Results limit

Set a cap on how many connection results are shown per note. A smaller number can improve clarity and performance; a higher number surfaces more context.

Exclude inlinks (backlinks)

Hides notes that link to the current note. Useful if you want to see only AI-suggested relationships and ignore existing manual backlinks.

Exclude outlinks

Hides notes that the current note already links out to. Keeps your Connections view focused on new or unlinked associations.

Include filter

Restrict results to only those whose file path contains this value. Example: entering projects/ will only show notes from the projects folder.

Exclude filter

Hides results with a file path containing this value. Example: entering archive/ will hide archived notes from appearing in the Connections view.

Smart Chat settings

Chat interface

Model configuration

Smart Environment settings

Action buttons

Show stats

Displays key metrics and diagnostic stats about your current Smart Environment (e.g. number of blocks, sources indexed, excluded items). Useful for debugging or performance tuning.

Reload sources

Reprocesses all sources in the environment, including notes, folders, and files. Handy after significant file changes or if context seems stale.

Clean-up data

Runs a safe cleanup process to remove orphaned or obsolete blocks and ensure metadata integrity. Does not affect your source files.

Clear sources data

Wipes all Smart Environment data (sources, blocks, embeddings, metadata) without touching original files. Ideal when switching datasets or fixing corrupted states. Does not delete settings.

📁 Excluded Folders

Lists folders that should be completely ignored by the Smart Environment. Any content in these folders won’t be indexed or included in prompts.

• Click “Add excluded folder” to select a folder.
• Click (x) to remove it from the exclusion list.

📄 Excluded Files

Manually exclude individual files from being processed by Smart Context. Useful for private drafts or noisy files that distort context.

• Use “Add excluded file” to select files.
• Click (x) to remove a file from exclusion.

🗂 Show Excluded

Opens a panel showing all currently excluded files and folders, giving you a full overview of what’s opted-out of the Smart Environment.

Sources and Blocks settings

Smart Sources

Minimum length

Minimum number of characters an entity (e.g., paragraph, heading) must have to be embedded. Shorter entries are ignored to reduce noise.

Embedding model platform

Choose where embeddings are generated.

• Transformers (Local, built-in) uses bundled models for full local processing.
• Other options may appear if remote APIs or custom models are integrated.

Embedding model

Choose which model to use for generating embeddings.
Example: BGE-micro-v2 is lightweight and fast, suitable for small-to-medium vaults.

Legacy transformers (no GPU)

Enables v2 transformer logic, which may be more compatible on systems without GPU acceleration. Toggle this if local embeddings fail to load or crash.

Smart Blocks

100% embedded (2636/2636)

Displays the current embedding coverage: how many content blocks have been processed.

Utilize Smart Blocks

When enabled, large documents are split into fine-grained blocks (based on headings and structure), improving semantic search and context relevance.

Minimum length (for blocks)

Sets the minimum character length for a block to be embedded. Avoids generating embeddings for very short or irrelevant sections.