TarsLab/obsidian-tars

What it solves

Tars is an Obsidian plugin that integrates AI text generation and multimodal capabilities directly into your notes. It allows users to interact with various AI assistants without leaving their knowledge base, enabling a seamless workflow for drafting, brainstorming, and analyzing documents.

How it works

The plugin uses a tag-based trigger system. Users can initiate AI responses by typing a specific tag (e.g., #Claude :) followed by a space, or by selecting tags from a command list. It follows a structured conversation format where system, user, and assistant messages are separated by blank lines. The plugin supports a wide array of AI providers (including OpenAI, Claude, Gemini, DeepSeek, and Ollama) and handles multimodal inputs such as embedded images and PDFs for visual understanding.

Who it’s for

Obsidian users who want to integrate LLMs into their personal knowledge management system for content generation, document analysis, and conversation logging.

Highlights

• Broad Provider Support: Integrates with numerous services including Azure OpenAI, Claude, DeepSeek, Gemini, and local models via Ollama.
• Tag-Driven Interaction: Uses Obsidian's native tag completion to trigger AI responses quickly.
• Multimodal Capabilities: Supports image generation, image analysis, and PDF interpretation.
• Dataset Export: Allows exporting conversations to JSONL format for use with fine-tuning frameworks like ms-swift.
• Reasoning Visibility: Displays Chain-of-Thought (CoT) reasoning in callout format for supported models like DeepSeek and LongCat.