Come for the wool, Tencent Cloud Cloud Studio big wool!!! Four

Building a private knowledge base using Cloud Studio

Cloud Studio + ollama + deepseek + nomic-embed-text + anythingllm + Private Knowledge === AI Private Knowledge Base

The image is the address of the AI-generated image

What is needed has already been mostly mentioned above, let's start our operation.

Still open our old friend Cloud Studio, now some of us refer to Cloud Studio as CS. I hope CS can grow bigger and stronger, creating brilliance again. Build it into our local Colab.

Still choose the homepage banner, directly select the 32b model
Wait for resource allocation, start up

After completion, as shown in the figure below

Cloud Studio has already embedded anythingllm for us. If we need to install and use it on our local computer, we need to download and install anythingllm. Of course, if we just want a simple experience, we don't need to download it, but can directly use the embedded web provided by CS.
We just need to open the embedded anythingllm page in the browser.

The step of pulling deepseek-r1:32b can be omitted because CS has already downloaded the model for us, provided that the 32 plan was selected. If we need to use 32b, we need to download it by pulling ollama pull deepseek-r1

Pull nomic-embed-text ollama pull nomic-embed-text

If using locally, choose to download the anythingllm client
Open initialization

Only local installation of anythingllm requires initialization; using Cloud Studio is already initialized by default. However, remember to execute the pull commands in the Cloud Studio terminal; the speed may be a bit slower since the resources are on foreign servers.

Click to start

Scroll down, select ollama

Do you remember when we mentioned deploying deepseekr1 with ollama to provide the API? The URL here is the one that needs to be filled in for the API address. If you don't know how to fill it, open Cloud Studio and click on the port.

See which port provides the ollama service, click the small globe to open a new page

Copy the URL here, mine is https://ottzkl-pkzulq-8434.app.cloudstudio.work, and paste it into the anythingllm client

Remember, be sure to remove the trailing slash from the URL; it should not have a slash, https://ottzkl-pkzulq-8434.app.cloudstudio.work

After entering the URL, the model above will automatically recognize it. Scroll down to select the appropriate model, for example, we choose 32b. We can keep the other parameters as default; modifying parameters is beyond the scope of this article.

Then keep clicking next until initialization is complete.

Let's start with some simple settings

Click the small wrench icon
2025/2/17 15:20

The operations at this point are consistent with the web page operations.

Click on embedder to embed preferences

Select ollama, the address is the same as the previous API address, paste it in, select nomic-embed-text, and click save changes.

Click on the small wrench icon, which is now a back arrow, click it to return to the homepage.

Create a new workspace on the homepage

Enter the name of the workspace, then click save.

Let's test it first to see if we can call deepseek normally

OK, normal. Click the upload button next to the workspace we created to upload materials

2EEBA100-19F5-4868-B087-BE212F6D09E4

Many file formats are supported, including txt, doc, csv, xls, docx, xlsx, pdf, URLs, etc., and audio and video files are also supported. Text files are the best format as they are easier and faster to recognize accurately.

I randomly found a few PDF files online

Drag, select all, and move to the workspace

Scroll down and select save.

Wait a moment for the upload to update the workspace data
Wait for the success prompt, and then pin the uploaded files so that our materials will be prioritized for calling