Original video in Chinese.
Update note (2026-10-03): This article preserves my workflow record from when I used Omnivore and Readwise. The official Omnivore repository states that its cloud service was deprecated in November 2024, and the current project is offered in self-hosted form. The Omnivore cloud account and sync steps in this article should not be used as a current installation guide. If you now want to save web pages to Obsidian, you can first look at the official Web Clipper instructions, my site’s clipping practice, and the Obsidian AI workflow.
Key Takeaway
- The author upgraded the Obsidian note system with the Omnivore and Readwise plugins, solving the pain point of syncing external information into Obsidian.
- Omnivore is responsible for syncing web articles into Obsidian, and supports both desktop and mobile.
- Readwise is responsible for saving article highlights and Twitter posts, and can consolidate different highlights from the same article into a single note.
- As the core of the AI knowledge base, Obsidian itself has AI plugins and a rich community plugin ecosystem, supports both cloud and local models, and syncs smoothly.
- With these two plugins, Obsidian achieves efficient input of external information, making it a more complete form of AI knowledge base.
Full Content
I made a small upgrade to my AI knowledge base. Although I only added two plugins, they have had a big impact on the whole workflow.
I previously released a video introducing my note system, which is the AI knowledge base. It consists of two major parts: first, processing external information, with AnythingLLM as the core; second, recording and accumulating notes, with Obsidian as the core.
To be honest, this setup can only be considered a beta version, and I’m not very satisfied with it. Because in actual use, I still had to move the useful content from AnythingLLM into Obsidian, which was too troublesome and not smooth enough. So that led to the two plugins I want to recommend today.
One is called Omnivore, and I have it specifically handle synchronizing web articles into Obsidian exactly as they are.
On desktop, I installed the Chrome plugin, so as long as I click the button in the browser, those WeChat public account articles and the like will be saved into my Omnivore account, and then, according to the settings I configured in advance, synced to a designated folder in Obsidian.
It can also sync on mobile, but the operation is different: you need to install the Omnivore app, then share from the browser to the app, and that’s it.
Besides saving entire articles, Omnivore can also save a certain passage from an article. However, I set a Template, so even when saving just a passage, it still brings along the full text afterward — which feels a bit untidy to me. So I let another plugin handle this job, called Readwise.
Readwise is simple to use: you install the browser plugin first, then select a piece of text, right-click, choose “Save Hightlights to Readwise,” and finally it automatically syncs to a designated folder in Obsidian according to the preconfigured format.
What I like most about this feature is that it can consolidate different Highlights from the same article.
A slightly longer article often has several Highlights that I want to keep. Many tools handle this by making one Highlight correspond to one note. After reading one article, you end up generating many notes, and then I still have to deliberately gather them into a single note and delete the redundant ones.
Readwise is very considerate in this respect. For example, I’ve already saved several Highlights from this article, and they’ve been synced to Obsidian to generate a note. Later, if there’s a new Highlight, it will supplement the existing note. Once an article has been digested, one corresponding note is generated — that feels much more comfortable.
Besides saving article excerpts, I also use Readwise to save tweets on Twitter. It supports two methods: @Readwise in a reply and ask it to save, or send it the tweet link via direct message. I don’t want to type, so I chose the direct-message method.
Readwise itself is an aggregator, and it connects with the mainstream tools on the market. It can even sync notes from Kindle, which is especially powerful — everyone should give it a try.
So, with these two plugins, I adjusted the folders in my Obsidian vault and added an Input directory, which contains three subfolders:
Articles is specifically for articles coming from Omnivore. In the plugin settings, I filled in the folder path in the Folder option.
Highlights and Tweets hold Readwise content. In addition to filling in the folder names in the plugin settings, I also changed the name of the Category Folder in the website backend, changing Articles to Highlights, otherwise it would overlap with Omnivore.
Both of these plugins support custom content formatting. I referred to templates from experts online and then modified them myself. I’ve posted all of these formats in the newtype community, and friends on Knowledge Planet or Patreon can take them and use them directly.
After all of these settings and adjustments, Obsidian became the core of my AI knowledge base:
First, it already has AI plugins, such as Copilot and Text Generator I introduced before. Both cloud and local models are supported, and there’s no problem at all with things like dialogue, content generation, and knowledge base queries.
Second, it’s already a very useful note-taking app, with an extremely rich ecosystem of community plugins. Syncing across PC, Mac, and iPhone through iCloud is also very smooth.
After adding the external-information sync piece through the two plugins I just introduced, this is the form of AI knowledge base I’m more satisfied with.
OK, that’s all for this episode. If you have anything you want to talk about, come find me at newtype — I’m always there. See you next time!
Keep Practicing
This article records how information enters the note vault. Next, you can keep reading according to the task you want to accomplish:
- How Obsidian Connects to an AI Workflow: choose a plugin, file system, or MCP integration approach.
- How Creators Build a Wiki: organize scattered materials into a knowledge system with sources and links.
- Obsidian PDF AI Workflow: when handling PDF materials, check whether text extraction, page numbers, and citations correspond.