Skip to main content
Obsidian Copilot is an open-source AI assistant plugin for Obsidian with a clean, privacy-first design. It lets you connect any OpenAI-compatible model service with your own API key, create custom prompts, and chat with your entire vault to get answers and insights from your personal knowledge base. Through APIYI, you can access 400+ mainstream models — GPT, Claude, Gemini, DeepSeek, and more — inside Obsidian with a single account, no need to register on multiple platforms.

​
Key Features

Chat with Your Vault

Vault QA mode retrieves across your entire vault via vector indexing, answering questions from your personal knowledge base

Custom Prompts

Built-in commands for summarizing, translating, and rewriting, plus custom prompts to process selected text in one click

Privacy First

Open-source plugin, local data storage, and API keys kept only in your local configuration

Flexible Model Access

Works with any OpenAI-compatible endpoint — switch freely between 400+ models via APIYI

​
Quick Start

​
Step 1: Get Your APIYI Key

  1. Open the APIYI website and sign up (or log in if you already have an account)
  2. Go to the “Tokens” page in the console and create a new API key
  3. Copy the key (it starts with sk-...) for later use

​
Step 2: Install the Obsidian Copilot Plugin

1

Install Obsidian

Download and install Obsidian from the official site: obsidian.md
2

Open the Community Plugin Market

Go to Obsidian Settings → Community plugins, turn off “Restricted mode”, and click “Browse”
3

Install and Enable Copilot

Search for Copilot (by Logan Yang), then install and enable it

​
Step 3: Configure an APIYI LLM Model

1

Open Copilot Settings

Go to Settings → Copilot and switch to the Model tab
2

Add a Custom Model

In the Chat Models section, click Add Custom Model and fill in:
FieldValue
Model NameA model name, e.g. gpt-5.2 or claude-sonnet-5
ProviderSelect 3rd party (openai-format)
Base URLhttps://api.apiyi.com/v1
API KeyYour APIYI key (sk-...)
3

Verify and Add

Click Verify to test connectivity, then click Add Model to finish
Configuration Notes
  • The Base URL must include the /v1 suffix: https://api.apiyi.com/v1
  • The model name must exactly match a model supported by APIYI — check the model list
  • You can add multiple models and switch between them anytime in the chat panel

​
Step 4: Configure an Embedding Model (Required for Vault QA)

To use Vault QA (knowledge-base Q&A) mode, you also need an embedding model:
  1. In the Embedding Models section of Copilot settings, click Add Custom Model
  2. Enter an embedding model name: text-embedding-3-small (cost-effective) or text-embedding-3-large (higher accuracy) recommended
  3. Select 3rd party (openai-format) as the provider
  4. Set Base URL to https://api.apiyi.com/v1 and enter your APIYI key
  5. Click Add Model to finish

​
Step 5: Save and Start Using

Select the newly added model as your default, then click Save and Reload. You can now:
  • Click the Copilot icon in the left sidebar to open the chat panel
  • Talk to the model directly in Chat mode
  • Ask questions across your entire vault in Vault QA mode (the first use requires waiting for the index to build)

​
Supported Models

Obsidian Copilot supports 400+ mainstream AI models through APIYI, including OpenAI, Claude, Gemini, DeepSeek, and Chinese models.

View Latest Model Recommendations

See the latest model recommendations, performance comparisons, and usage advice. The model list is continuously updated so you always use the newest and strongest AI models.
Why don’t we list specific models here?AI models iterate very quickly. To ensure you get the most accurate recommendations, we maintain the latest model list, performance data, and usage advice on the model recommendation page.
Scenario-based Suggestions
  • Everyday note Q&A: pick a fast, low-cost lightweight model
  • Long-form summarizing / deep writing: pick a flagship model with a long context window
  • Vault QA embedding: text-embedding-3-small covers the vast majority of knowledge-base scenarios

​
Advanced Features

​
Custom Commands and Prompts

Copilot lets you save frequent operations as custom commands:
  1. Open the Commands section in Copilot settings
  2. Create a custom prompt, e.g. “Rewrite the selection as a weekly report”
  3. Select text in the editor and invoke it via the command palette (Ctrl/Cmd + P)

​
Working with Selected Text

Select any passage in a note and call the built-in commands directly:
  • Summarize: summarize the selection in one click
  • Translate: translate into a specified language
  • Simplify / Fix grammar: simplify wording or fix grammar
  • Generate table of contents: create a TOC

​
CORS Compatibility Mode

If chat requests fail after configuration, enable the CORS option when adding the model.
With CORS mode enabled, Obsidian does not support streaming output — the reply appears all at once after generation completes. APIYI’s standard endpoint usually does not need this option; only try it if requests fail.

​
Troubleshooting

  • Check that the Base URL is https://api.apiyi.com/v1 (including /v1)
  • Make sure the API key was copied correctly with no extra spaces
  • Confirm your account has sufficient balance
  • If it still fails, try enabling the CORS option in the model settings
  • The model name must exactly match a name supported by APIYI (case-sensitive)
  • Check the exact model name in the model list
  • Make sure a separate embedding model is configured (an LLM model cannot double as the embedding model)
  • Indexing a large vault for the first time takes a while — please be patient
  • After changing the embedding model, rebuild the index (Force Re-index)
  • Switch to a faster lightweight model
  • Reduce the conversation context length
  • In Vault QA mode, lower the number of retrieved chunks

​
Tips

  1. Divide work between models: add multiple models — use a lightweight model for everyday Q&A and switch to a flagship model for deep writing to save cost
  2. Scope the index: exclude attachments and template folders in Copilot settings to reduce embedding cost and improve retrieval quality
  3. Leverage custom prompts: turn high-frequency operations (e.g. “organize meeting notes”) into commands — configure once, reuse forever
  4. Rebuild the index regularly: run Force Re-index after major vault changes to keep Vault QA retrieval accurate

Model Recommendations

See the latest model list and scenario-based recommendations

Getting Started

Register an APIYI account and create a key in 3 minutes

Base URL Configuration

Learn how to correctly fill in the API endpoint in different tools

Cherry Studio

Another powerful desktop AI chat client
Obsidian Copilot open-source repository: github.com/logancyang/obsidian-copilot