![[promo-embed-home]] In the ever-evolving landscape of personal knowledge management (PKM), few tools have captured the attention and loyalty of users quite like Obsidian. Obsidian has carved out a significant niche for itself with its [File over app philosophy](https://stephango.com/file-over-app), which promotes [[Obsidian - Understanding its Core Design Principles|local storage, privacy, and the portability of markdown]]. Unlike many cloud-based alternatives, it keeps notes securely stored on the user’s device while leveraging Markdown for streamlined and efficient note management. Yet, as the PKM community increasingly embraces artificial intelligence (AI), Obsidian stands apart for its absence of native AI features. This article explores Obsidian’s unique stance, the broader evolution of PKM tools in the AI era, and a proposal for potential pathways for integrating AI into Obsidian without compromising its [[Obsidian - Understanding its Core Design Principles|core principles]] of privacy and user control. >[!info] If you want to give feedback on this article, let us continue the conversation at [reddit](https://www.reddit.com/r/ObsidianMD/comments/1gy7f7e/shaping_obsidians_tomorrow_how_obsidian_can/) # PKM in the Age of Artificial Intelligence ![[oiahr-cover-optimized.jpg]] Until recently, many PKM tools promoted their primary selling points as their ability to help users organize files, create wiki-like links, search, and visual graphs. However, as technology advances, the PKM landscape has substantially changed, particularly with the growing interest in artificial intelligence. A few years ago, AI was not a consideration for most PKM users, but today, it has become a key area of innovation and development. Modern PKM applications increasingly integrate AI to enhance user experience, automate tasks, and provide intelligent insights into stored information. (Examples of PKM and AI [1](https://www.notion.com/product/ai)[2](https://reflect.academy/artificial-intelligence)[3](https://tana.inc/ai)[4](https://docs.capacities.io/reference/ai-assistant)[5](https://www.craft.do/blog/craft-ai-assistant)[6](https://evernote.com/features/ai-features)) This shift reflects a broader trend in which users seek more dynamic and interactive ways to manage their knowledge. In other words, why should users be worried about linking notes, organizing files, or formulating complicated search patterns when the software could do it for them? While AI in the PKM arena is still in its early stages, one thing becomes crystal clear: AI-powered features can transform static notes into interactive repositories of knowledge, offering capabilities such as natural language search, automated summaries, intelligent recommendations, and serving as a [[The Power of Artificial Intelligence as a Thinking Partner - Embrace Its Importance|research/brainstorming]] partner. # Obsidian’s Approach to Artificial Intelligence In this context, unlike many competitive PKM apps, the Obsidian team has [deliberately chosen](https://x.com/kepano/status/1799165623395881129) not to incorporate native AI features into its product, opting to leave it to third-party plugin developers to integrate AI into Obsidian. Obsidian’s approach reflects its philosophy of prioritizing user privacy and data control. The theory is that by avoiding direct AI implementation, it aims to give users greater authority over how their data is managed, reducing potential risks tied to AI-driven processes. > The lack of integrated AI in Obsidian appeals to some users, at least for now. User privacy is a fundamental aspect of Obsidian’s strategy. Many users are rightly hesitant to permit AI tools to access and analyze their notes, worried that their data could be exposed or misused without their explicit consent. By making AI capabilities optional and plugin-based, Obsidian offers users the freedom to explore AI features while ensuring they can keep their data completely private. > Opinions in the Obsidian community differ significantly about the role of AI in Obsidian. ![[oiahr-camera.jpg]] In my conversations with the community about AI in Obsidian, I’ve been struck by the wide range of opinions: - Some users strongly oppose AI integration in their notes. They’re concerned about privacy, the risk of AI-generated inaccuracies or “hallucinations,” and the potential for AI-generated content to mix with their original ideas. - On the other end of the spectrum, some users fully embrace AI. They rely on it to write and rewrite notes, conduct research, [[The Power of Artificial Intelligence as a Thinking Partner - Embrace Its Importance|brainstorm]], and generate Markdown or Dataview queries. - Between these extremes lies a curious but cautious group of users. These users aren’t currently using AI but are intrigued by its potential benefits and are wary of its risks. They may feel they’re missing out or don’t know where to begin integrating AI into their workflows. This diverse landscape highlights the varying needs and expectations within the Obsidian community regarding AI. # Community Plugins to the Rescue ![[oiahr-ai-plugins.jpg]] While Obsidian does not include native AI features in its product, its vibrant plugin ecosystem has produced numerous AI-driven plugins. These enhancements integrate sophisticated AI functionalities to improve users' interactions with their notes. There are many popular and useful AI plugins. Let me list just a few examples: - [Smart Connections](https://smartconnections.app) and [Copilot for Obsidian](https://www.obsidiancopilot.com/en) - are plugins that give you a ChatGPT-like experience inside Obsidian, with the ability to chat with your notes and vault. - [Caret](https://caretplugin.ai) uses Obsidian Canvas to create a conversation with AI visually. This is a unique way to use AI chat, allowing you to go down many paths in a conversation and to follow the paths visually. Additionally, [Cannoli](https://github.com/DeabLabs/cannoli) also uses Canvas to create no-code workflows. - [Text Generator](https://text-gen.com), [QuickAdd](https://quickadd.obsidian.guide), and [AI for Templater](https://tfthacker.com/AIT) - plugins to populate templates with AI responses. - And the list goes on and on with AI plugins for [tagging](https://github.com/lucagrippa/obsidian-ai-tagger), [image analysis](https://github.com/Swaggeroo/obsidian-ai-image-analyzer), [latex](https://github.com/aaaaayushh/ai-latex-generator), [summarization](https://github.com/irbull/obsidian-ai-summary), [flashcard generation](https://github.com/irbull/obsidian-ai-summary), [cognitive behavioral therapy](https://github.com/clairefro/obsidian-chat-cbt-plugin), etc. - There are numerous more I don't list here. When you search the community plugin list for AI plugins, one fact is clear: t**here are many AI plugins.** This proves that the _Obsidian community is intensely interested in using AI in PKM_. Before going further, it is worth mentioning that these plugins do a great job of trying to help the user manage what information from their vault is shared with LLMs. In many cases, it is a matter of properly configuring the plugin to limit the scope of its access. I wrote about this here: [[AI and Your Obsidian Vault - Practical Advice to Protect Your Privacy]] >[!info] If you want to give feedback on this article, let us continue the conversation at [reddit](https://www.reddit.com/r/ObsidianMD/comments/1gy7f7e/shaping_obsidians_tomorrow_how_obsidian_can/) ## Challenges with Current Plugins ![[oiahr-complexity.jpg]] Nevertheless, from my personal experience, the current state of AI in Obsidian presents several challenges. Many AI plugins for Obsidian are still in their infancy, often requiring users to navigate complex setup processes and configurations. The variety of plugins means that each developer might have a different approach to integrating AI, leading to inconsistencies in functionality and user experience. Additionally, some plugins can be difficult to use, making them less accessible to the average user, who may not have the technical expertise with AI to troubleshoot issues. Many of these plugins have invented their own way of storing additional data they need (embeddings, chat conversations, etc). While no harm is done, you can end up with weird folders and custom file formats after installing some plugins. It is like the Wild Wild West. These plugins often assume a great deal of knowledge about how LLMs (large language models) work, how to set up API paths and keys, tweak LLM parameters, and choose from various models. While this is not impossible to learn, it can overwhelm users. I find it exhausting having to configure each of these plugins, often giving the same information to each plugin, but having to learn how the developer wants that information in their plugin. These challenges highlight a gap in the current PKM Obsidian landscape. While AI offers promising enhancements to note-taking and knowledge management, the lack of standardized, user-friendly AI integration in Obsidian can be frustrating. Users who desire AI functionalities might find the existing plugins cumbersome, potentially limiting their ability to leverage AI in their PKM workflows fully. ![[promo-embed-newsletter]] # Opportunities for Enhancing AI Integration in Obsidian I believe the Obsidian team has a unique opportunity to lead the way in defining how AI can be thoughtfully incorporated into their product. By taking a proactive approach, they could make AI features accessible and intuitive for users while empowering developers to focus on creating value added features rather than reinventing each AI plugin's foundational “plumbing” code. Let me outline two things I'd like to see in time from the Obsidian team. ## 1. Development of a Core “AI Plumbing” Plugin ![[oiahr-plumbing.jpg]] It would be wonderful if the Obsidian team would develop a centralized core plugin that would provide a standardized framework for connecting to various AI engines (e.g., OpenAI, Claude, LLaMA). Currently, each AI plugin in Obsidian implements its own code for handling these integrations, resulting in redundancy, inefficiency and potential security issues. This core plugin could provide a single location for all settings and options and a shared API for communicating with LLMs. Obsidian would streamline the development process by offering a core plugin, allowing plugin authors to focus on creating innovative features rather than duplicating tedious plumbing/framework development. > This shared foundation would enhance consistency, reliability, safety, and ease of maintenance across all AI-related plugins. Users could opt into this feature, ensuring that AI integration remains entirely under their control. In other words, with a switch, AI can be turned off or on in a vault. This could be a core plugin or a community plugin maintained by Obsidian, thus giving users even more control if AI is permitted in the vault. ## 2. Introduce a User-Friendly “Batteries-Included” AI Plugin ![[oiahr-batteries.jpg]] To make AI accessible to a broader audience unfamiliar with the complex landscape of LLMs, Obsidian could develop a simple, point-and-click plugin offering essential AI functionalities with minimal configuration. > The goal would be to require very little knowledge by the user to use AI in their vault. This plugin could include basic features like generating text, summarizing notes, or making contextual AI suggestions, providing an easy on-ramp for users who are intimidated by complex setups. Obsidian could consider monetizing this plugin, following the model of tools like Notion that charge for AI services. I believe many users would pay $10 to $12 a month for such a service. Obsidian could clearly document its AI providers, negotiate terms with AI providers that don't allow LLM providers to train on user data, and implement best practices that ensure security and transparency. (Notion has defined what I think is the best disclosure approach for this. See these links: [1](https://www.notion.com/help/notion-ai-security-practices), [2](https://www.notion.com/security), [3](https://www.notion.so/notion/Notion-s-List-of-Subprocessors-268fa5bcfa0f46b6bc29436b21676734)) ### Couldn't the Community Build these Plugins???? No. Or at least, I don't think so. Community plugins require a lot of work and time. The "AI Plumbing" core plugin is too big, sensitive, and crucial to be developed without deep insight into Obsidian. The Obsidian development team is in the best position to do this. Many choices need to be made about what to include, but even more importantly, what not to include. Again, the Obsidian development team is best positioned to do this. Having said that, they could open-source the plugin and enlist the community's help while providing thoughtful and clear guidance. ### A Collaborative Model for AI Integration This two-plugin model—one for foundational AI infrastructure and another for basic functionality—would create a cohesive framework for users and developers. Community plugins could build on the “AI Plumbing” plugin to offer advanced capabilities, ensuring flexibility and innovation while maintaining a streamlined, user-friendly experience. This approach balances user control, accessibility, and the potential for future AI advancements in Obsidian. # Looking Ahead: The Future of Obsidian and AI We need to ask ourselves: Is artificial intelligence going away? Is there any future in which it won't play an essential role in knowledge management?  These questions motivate further discussion of Obsidian's core functionality. Obsidian’s deliberate decision to exclude native AI integration poses both opportunities and challenges. With thoughtful development and community collaboration, Obsidian can meet user demand for AI-driven features while maintaining its hallmark focus on privacy and user control—ensuring its continued relevance in the changing PKM landscape. AI’s role in knowledge management is undeniable, and its advancement will increasingly shape user expectations. The question is not whether AI will matter but how tools like Obsidian can integrate it while remaining true to their principles. ![[promo-embed-home]]