Steward

Steward is a plugin that utilizes Large Language Models (LLMs) to interact with your Obsidian Vault. It provides commands like search, move, copy, create, etc, as building blocks to create your own sophisticated commands, prompt chaining, and automation for your specific tasks.

Features

• Built-in Search Engine: A TF-IDF based search with relevant scoring and typo tolerance that is significantly faster than the native Obsidian search
• Command-based Interaction: Support for standard commands like search, create, update, delete, move, audio, image generation, and user-defined commands
• Multi-language Support: Use Steward in your preferred language
• Privacy-focused: Most actions are executed in the front-end using Obsidian API to avoid exposing your data to LLMs (except for your queries)
• Interactive Chat UI: An interactive chat interface using your current theme that leverages Obsidian's editor features to input any supported Markdown syntax. You can start a conversation, whether in the Chat sidebar or directly in the current editor.
• Model Flexibility: Use your favorite AI models including OpenAI, DeepSeek, and Ollama local models
• Intent Caching: Utilizes embeddings to cache similar queries so subsequent requests don't require LLM processing

Development Status

🚧 Active Development 🚧

This plugin is currently under active development. New features and improvements are being added regularly. While the core functionality is stable, you might encounter occasional issues or changes as development progresses. Feedback and bug reports are welcome!

Standard (Built-In) Commands

Steward can be used through the command palette directly in the editor or by opening the chat interface. Here are some example commands:

Command Flow Visualization

The following diagram illustrates how commands are processed in Steward:

Usage

• /search Notes tagged #Todo in the root folder
• / Add tag #Done to all notes of the search results and move them to the Archived folder
• / Write a poem about Angular in a new note then move it to the Generated folder
• / Update the list above to a numbered list
• /audio "project" as a noun and a verb using 11Labs
• / I don't like your name, you are "Joe" from now on

Tips to reduce the number of input tokens

• Use a specific command directly instead of a general command (/ ), e.g, /search, /audio, etc.
• For the /search command, wrap the keyword in quotation marks for searching the vault without the LLM's help, e.g, /search "my cat"
• Create User-Defined commands
• Provide feedback by clicking thumbs up or down on any Steward's answer to help it classify accurately #Todo.

User-Defined Commands

You can create your own User-Defined Commands to automate workflows and combine multiple built-in or other User-Defined commands into a single, reusable command.

How It Works

• User-Defined Commands are defined as YAML blocks in markdown files inside the Steward/Commands folder.
• Each command can specify a sequence of built-in or user-defined commands to execute.
• You can specify if user input is required for your command using the query_required field.
• These commands are available with autocomplete and are processed just like built-in commands.

Example: User-Defined Command YAML

command_name: clean_up
description: Clean up the vault
query_required: false
model: gpt-4o # Optional: Specify a default model for all commands
commands:
  - name: search
    system_prompt:
      - Notes in the root folder
    query: Notes name start with Untitled
  - name: delete_from_artifact
    query: Delete them
    model: gpt-3.5-turbo # Optional: Override the model for this specific step

• command_name: The name you will use to invoke the command (e.g., /clean_up)
• query_required: (optional, boolean) If true, the command requires user input after the prefix
• model: (optional, string) The model to use for all commands in this user-defined command
• commands: The sequence of built-in or user-defined commands to execute
  • system_prompt: The system prompts that allows you to add additional guidelines to LLMs to the current command
  • query: (required if the query_required is true, string) The query to send to LLMs, put the $from_user as a placeholder for your input
  • model: (optional, string) The model to use for this specific command step (overrides the command-level model)

Using Links in System Prompts

You can reference the content of other notes in your vault by using Obsidian links in the system_prompt array:

command_name: search_with_context
description: Search with predefined context
query_required: true
commands:
  - name: search
    system_prompt:
      - '[[My Context Note]]'
      - Additional instructions
    query: $from_user

When the command is executed:

1. The link [[My Context Note]] will be replaced with the actual content of that note
2. This allows you to maintain complex prompts or contexts in separate notes
3. You can update the linked notes independently of your command definition

Usage

1. Create a note in Steward/Commands and add your command YAML in a code block.
2. In any note or the Chat, type your command (e.g., /clean_up #Todo) and press Enter.
3. The command will execute the defined sequence, using your input if required.

Validation

• The system validates your User-Defined Command YAML:
  • command_name must be a string
  • commands must be a non-empty array
  • If present, query_required must be a boolean
  • Each command step must have a name (string) and query (string)
• If validation fails, the command will not be loaded and an error will be logged.

Why User-Defined Commands?

• Automate repetitive tasks
• Create macros for your workflow
• Share and reuse command sequences

Creating Commands with LLM Assistance

You can ask Steward to help create user-defined commands using natural language, even without knowing YAML syntax:

1. Simply share the User-Defined Command Guidelines with Steward
2. Describe what you want your command to do in plain language
3. Steward will create commands with the proper YAML structure for you
4. Review, modify if needed, and save to your Commands folder

Installation

1. Download the plugin from the Obsidian Community Plugins browser
2. Enable the plugin in your Obsidian settings
3. Configure your API keys in the plugin settings

Settings

• API Keys:

  • OpenAI API Key (for OpenAI models and embeddings)
  • ElevenLabs API Key (for audio generation)
  • DeepSeek API Key (for DeepSeek models)

• LLM Settings:

  • Chat Model: Choose between various models from OpenAI, DeepSeek, or Ollama
  • Temperature: Controls randomness in the output (0.0 to 1.0)
  • Ollama Base URL: For local Ollama models (default: http://localhost:11434)

• Steward Folder: The folder where Steward' related notes will be stored (default: Steward)

• Debug Mode: Enable detailed logging for troubleshooting

Development

This plugin uses TypeScript and follows the Obsidian plugin architecture.

Building

1. Clone this repository
2. Run npm install to install dependencies
3. Run npm run build to build the production version

License

MIT

TODOs

• Multiple lines command.
• Moving notes to multiple destinations.
• Search in one or many specific folders.
• Command options: --help, --model, --explain,...
• The reversion of command's actions.
• Reminder.
• Steward can provide information about its functionalities, limitations,...
• Provide any information, usage, and guidance about Obsidian.
• Autocompletion and automation.
• Unused media actions: Move or remove.
• Cache previous generated contents.
• Respect user's language.
• User-defined commands and actions, for example: Don't do anything if not entering the correct password.
• User confirmation or clarify when the AI is unsure about its response.
• User confirmation when the AI performs a large side effect action.
• Traceability
• MCP support.
• Remember user's intent.