Obsidian AI With DeepSeek: Connect Your Notes to a Model

DeepSeek provides an API that is compatible with the OpenAI API format, which means you can use it with any Obsidian AI plugin that accepts a custom API base URL. You configure the plugin to point at DeepSeek's endpoint, enter your DeepSeek API key, and choose which model to use. Tools like fenbs complement this by holding the project decisions and rules that should reach every AI app, not just the Obsidian plugin, so approved choices travel with your work rather than staying inside the vault.

This article walks through choosing a compatible plugin, configuring DeepSeek, deciding what vault content the model should see, and keeping shared project decisions in fenbs.

Choose a compatible Obsidian plugin

Look for an Obsidian AI plugin that lets you set a custom API base URL. Several popular ones do. The setting is usually labeled "API Base URL" or "Custom endpoint." Enter DeepSeek's API base URL (available in DeepSeek's own documentation), your API key, and the model name you want to use. The plugin will then route requests to DeepSeek instead of OpenAI.

If a plugin only supports OpenAI and does not expose a custom URL field, it will not work with DeepSeek without modification. Check the plugin's settings screen or README before installing.

Configure the API key safely

Your DeepSeek API key should stay in the plugin's settings, not in a vault note or a shared file. Most Obsidian AI plugins store keys in the plugin data folder, which lives outside the main vault and is not synced by default. Check your sync settings to confirm the plugin data folder is excluded from any cloud sync.

• Store the API key only in the plugin settings, not in a note or a markdown file.
• Confirm that your sync service excludes the .obsidian/plugins folder or its data files.
• Rotate the key if you accidentally commit or upload a file that contains it.
• Use a separate key for Obsidian rather than the same key you use for other projects.

Control which notes DeepSeek can read

By default, some Obsidian AI plugins send the active note or a selected block to the model. Others can index the vault and send broader context. Before enabling vault-wide indexing or sending, decide which notes are appropriate to share with an external API. DeepSeek's API sends content to DeepSeek's servers; check their data-handling documentation before sending anything sensitive.

A practical approach is to create a folder for AI-safe notes: drafts, research, and templates you are comfortable sending to an external API. Keep personal notes, client records, and financial information in a separate folder and confirm the plugin is not indexing it.

Keep shared project decisions in fenbs

A decision you reach during a DeepSeek session in Obsidian stays in that session unless you copy it somewhere. If you also use other AI apps on the same project, they will not see it. Keeping decisions in fenbs means any connected assistant reads them at the start of a session, whether in Obsidian or elsewhere.

This is especially useful for rules that apply across multiple AI tools: approved phrasing for a client, a confirmed technical approach, or a standing instruction that should not be overridden. Record those in Decisions and rules in fenbs so every assistant starts with the approved position. For more on managing AI context across tools, see what an AI brain is.

How to do this with fenbs

Use DeepSeek in Obsidian for note-based AI work and fenbs for the shared decisions and project record.

1. Create a free account and add a project for the work your Obsidian notes support.
2. Write the key project decisions in Decisions and rules, especially any that should apply in all AI sessions, not just the Obsidian plugin.
3. Add AI-safe notes as a context note: which vault folder the assistant may use and what it contains.
4. Connect a supported AI app to fenbs via Integrations so the decisions travel to every session.
5. After each Obsidian session, copy any new decisions that matter to fenbs before they get lost in chat history.

Troubleshooting a DeepSeek connection in Obsidian

If the plugin returns an error after you configure DeepSeek, check the API base URL and model name first. DeepSeek's model identifiers differ from OpenAI's; entering an OpenAI model name will fail. The plugin's error message often shows the raw API response, which will usually say the model was not found or the key was rejected.

• Verify the API base URL matches the one in DeepSeek's documentation exactly.
• Check that the model name is a valid DeepSeek model identifier.
• Confirm the API key is active and has not expired or been revoked.
• Test the API key with a direct request using a tool like curl to isolate whether the issue is the plugin or the key.