Build a Personal Knowledge Base with WorkBuddy + ima – A Tutorial Anyone Can Follow

When most people think of a personal knowledge base, Obsidian is the first thing that comes to mind.

It’s certainly great: local Markdown, bidirectional links, plugins, folders, templates, graph views – once you go deep, it’s incredibly powerful.

But the problem is that for most ordinary people, the bottleneck isn’t “my knowledge base isn’t advanced enough.”

The real bottleneck is: Your materials are too scattered.

A bunch of articles saved in WeChat Favorites, a pile of docs in Tencent Docs, dozens of PDFs in computer folders, meeting notes everywhere, videos watched without subtitles, ideas written in quick notes – and in the end, nothing can be found.

If you ask a beginner to jump straight into learning Obsidian, Markdown, templates, bidirectional links, and plugins, it’s actually a surefire way to scare them off.

So if your main materials live in a Chinese office environment, I’d actually recommend starting with a combo better suited for domestic users:

✅ WorkBuddy + ima

• ima handles putting materials into the knowledge base.
• WorkBuddy handles reading, organizing, categorizing, summarizing, and generating reviews.
• You make the final judgment: which content is worth keeping, and which is just temporary material.

This is how I see a personal knowledge base built from 0 to 1.

Don’t think about building a complex system right away. First, let someone finally help you process that first pass of materials scattered across WeChat, docs, PDFs, meetings, and videos.

1. Why Not Start with Obsidian Right Away?

Obsidian’s advantages are clear: it’s great for long-term accumulation, for building your own structure, and for weaving knowledge into a network bit by bit.

But it has a practical problem: It’s more like “a really powerful warehouse.”

But the warehouse itself won’t help you organize the shelves.

For many people, their Obsidian doesn’t become a second brain in the end. It becomes a second trash bin. Web clippings go in, PDFs get dumped in, daily notes get written in, meeting minutes get stuffed in. It feels satisfying at first, but after a while, you don’t even want to open it.

The reason is simple: Once materials come in, no one does the second step of processing.

• Should a fleeting idea be kept?
• Which project does an article relate to?
• Which pages of a PDF are useful?
• What conclusions did a meeting actually yield?
• Can a piece of video content become a topic?

If these questions aren’t addressed, the bigger the knowledge base gets, the heavier the burden becomes.

So for ordinary people building a personal knowledge base, the first step shouldn’t be chasing sophistication.

Chase one thing first: Get someone to do the initial processing once materials come in.

That’s the value of WorkBuddy + ima.

2. How Does This Combo Actually Divide the Work?

I break it down into three roles:

This division is crucial. Because many people, when building a knowledge base, want to hand the entire process over to AI immediately – automatic reading, automatic categorization, automatic summarization, automatic archiving, automatic note generation. It sounds awesome, but in practice, it’s dangerous ⚠️.

Because long-term knowledge isn’t formatted text; long-term knowledge requires judgment.

• Which materials are worth keeping?
• Which viewpoints just seem correct?
• Which content will you actually use in the future?
• Which materials were only useful at the time and will never be opened again?

These cannot be completely handed over to AI.

A more stable approach is:

WorkBuddy handles pre-processing. ima handles accumulation. You handle confirmation.

Don’t mess up this order. If you do, your knowledge base will become a pile of beautifully generated nonsense from AI.

3. Starting from 0: Complete the Connection in 5 Minutes

Step 1: Install and Log in to ima

• Open https://ima.qq.com/ in your computer browser.
• Download the Mac or Windows client and log in by scanning a QR code with WeChat.
• Mobile: Search for the “IMA Knowledge Base” mini-program in WeChat and log in with the same WeChat account.

Important: ima and WorkBuddy must be logged in with the same WeChat account, otherwise the knowledge base won’t be accessible.

Step 2: Connect ima in WorkBuddy

• Open the WorkBuddy client.
• Click “Material Library” in the left sidebar, then select “ima Knowledge Base”.
• Click “Go to Authorize”.
• A WeChat QR code scan interface will pop up. Log in to the ima knowledge base by scanning with WeChat.
• Click “Confirm Authorization”.

After authorization is complete, WorkBuddy will have the following permissions:

• View your knowledge base list in WorkBuddy.
• View or search materials in the knowledge base.
• Add knowledge base files to tasks or save files to the ima knowledge base.

Verify the connection: Type “Help me list all the knowledge base names in the ima knowledge base” in the WorkBuddy dialog. If it returns your knowledge base list, the connection is successful.

Step 3: Understand the Two Ways to Reference Materials

After a successful connection, there are two ways to reference ima knowledge base content in a task:

Method A – Add from Material Library: In “Material Library → ima Knowledge Base”, select the corresponding file and click “Add to Task”.

Method B – Add in the Dialog: Click the “Upload File” icon above the input box in the new task window, open the ima knowledge base. You can select an entire knowledge base or specific files. Click “Confirm” to add them to the conversation.

Step 4: Understand the “Save Output Back” Function

After WorkBuddy processes materials, the outputs can be saved back to ima:

• Find the desired file in the output area on the right.
• Click “Upload to Cloud”.
• Select “ima Knowledge Base” to save.
• Supports saving to Personal / Shared / Subscribed knowledge bases.

This completes the full loop: Read → Process → Save Back.

4. For Beginners: Just Build Three Knowledge Bases

The biggest problem with many knowledge base tutorials is designing things too complex from the start. Material library, project library, domain library, card library, output library, archive library, idea library, reading library, client library, product library, retrospective library…

It sounds professional, but for beginners, the more categories there are, the easier it is to give up.

For the first version, just build three:

01-Material Library

Used for storing external materials.

Articles, PDFs, reports, web pages, courses, video subtitles, interview transcripts, meeting audio transcriptions.

02-Project Library

Used for storing things you are actively working on.

A public account topic, a client project, a product idea, a course plan, a business proposal.

03-Output Library

Used for storing your own creations.

Article drafts, topic ideas, retrospectives, summaries, viewpoint cards, methodology insights.

Why only three?

Because in the beginner stage, the most important thing isn’t fine-grained categorization. It’s first cultivating a habit: Once materials come in, don’t scatter them everywhere. First, let them enter a place where they can be read and processed by AI.

Once this step works, you’ve already surpassed most people’s “bookmark knowledge base”.

How to create a knowledge base in ima: Open the ima client → “Knowledge Base” on the left → Click “+” to create a new knowledge base → Enter a name → Creation complete.

5. Don’t Import Too Many Materials at First

The easiest mistake when building a knowledge base is wanting to dump everything in on the first day. Hundreds of WeChat Favorites, hundreds of browser bookmarks, dozens of PDFs in computer folders, tons of historical notes. Importing it all looks spectacular.

But this isn’t building a knowledge base. This is moving a trash heap into a new house.

Keep the first batch of materials small. I recommend no more than 20 items:

Why so few? Because the first phase isn’t about “building a massive knowledge base”. It’s about verifying three questions:

• Can ima store these materials properly?
• Can WorkBuddy read and organize these materials effectively?
• After you see the AI-generated results, are you willing to continue using it?

If these three questions aren’t answered successfully, more materials just mean more chaos. First, test it with a small batch of real materials. Once it works, then expand.

How to Add Materials to ima:

• Public account articles: Copy the article link → Open ima client/mini-program → Enter the corresponding knowledge base → Paste the link to save.
• WeChat chat files: In the knowledge base, select “Upload WeChat Files” → Choose a contact → Select files → Add to library.
• Local files: Open ima client → Enter the corresponding knowledge base → Upload the file.
• Photos/Screenshots: Open the IMA mini-program → “+” → Take a photo → Automatically OCR and add to library.

Note: Sending files directly in the dialog only makes them temporary files. You must go into the “Knowledge Base” section to upload for them to be truly stored.

6. After the Connection is Established, First Run Three Minimum Tasks

Once WorkBuddy and ima are connected, don’t rush to automation. First, run three minimum tasks. Once these three tasks work, the prototype of your personal knowledge base is ready.

Task 1: Organize the Material Library

Let WorkBuddy read your 01-Material Library. You can say something like this:

Please read the “01-Material Library” in the ima knowledge base. Help me organize the recently uploaded materials. Group the output by topic. For each item, include: Title, one-sentence summary, core viewpoint, suggested topic it belongs to, whether it’s worth close reading. Do not generate a final article. Do not delete, move, or overwrite any materials.

The purpose of this task isn’t for AI to give you a conclusion. It’s to first turn a pile of materials into a checklist you can quickly scan. After you read it, you’ll at least know:

• Which ones are worth close reading.
• Which ones are just temporary references.
• Which ones can be read later.
• Which ones shouldn’t have entered the knowledge base at all.

This step is crucial. Because the first checkpoint for a personal knowledge base isn’t collection – it’s filtering.

Task 2: Generate a Project Material Pack

Let WorkBuddy organize materials around a specific project:

Please read the “02-Project Library” in the ima knowledge base. Compile a project material pack related to “My Current Project”. Output: Project background, existing materials, key questions, available resources, missing information, next steps. Do not make the final decision for me. Do not generate a formal plan.

This task is extremely practical. Because many times, the problem isn’t that we don’t have materials. It’s that we don’t know the relationships between them. A client requirement is in a Tencent Doc, an idea is in a phone memo, an industry article is in WeChat Favorites, a meeting minute is in a folder.

WorkBuddy’s role is to first pull them into the same perspective for you. Then you judge: What is this project missing now? What should be supplemented next? Which materials can be used immediately? Which are just noise?

This is much more reliable than having AI write a plan directly.

Task 3: Generate Content Topics

If you write for a public account, X (Twitter), Xiaohongshu (Little Red Book), or video scripts, this combo is great for building a topic pool:

Please read the “01-Material Library” and “03-Output Library” in the ima knowledge base. Help me generate 10 topics suitable for posting on a public account or X. Each topic includes: Title, core viewpoint, quotable materials, suitable angle, risk points. Do not write them directly as full articles.

The key point in this instruction is the last sentence: Do not write them directly as full articles.

Many people using AI for content writing immediately say “Help me write a viral article.” The result is usually a bunch of clichés. A more stable approach is to have AI first generate topic candidates, material packs, viewpoint comparisons, and risk points. You still decide how the final content is written.

7. The Correct Process: Materials Don’t Become Articles Directly

Many people misunderstand an AI knowledge base as an automatic writing machine: Put materials in → AI reads them → Directly generates an article → Copy and publish.

This process is tempting, but it’s prone to problems. Because:

• Materials ≠ Viewpoints
• Viewpoints ≠ Articles
• Articles ≠ Content ready for publication

A more stable process should be:

Materials enter ima ↓ WorkBuddy generates a review ↓ Human filters valuable content ↓ Then generate topics ↓ Then generate an outline ↓ Finally, write the main text

This process seems to have more steps, but it actually saves more time. Because it avoids a much bigger pitfall: You publish something that looks complete, but you haven’t truly judged it yourself.

This is also why I don’t recommend letting WorkBuddy generate the final knowledge directly. It can help you read, organize, categorize, find connections, and generate candidates – but the final step must be confirmed by you.

Especially for content involving viewpoint judgment, business judgment, or high-risk areas like law, medicine, or investment – it cannot be left entirely to AI.

8. The Most Important Thing About a Personal Knowledge Base Isn’t “Storing”, It’s “Reusing”

Many people think the core of a personal knowledge base is storing materials. Actually, it’s not. Storing is just the first step. The real value lies in reusing.

For example, today you read an article about AI Agents and put it in 01-Material Library. Next week, when you write a WorkBuddy tutorial, it becomes a reference. Next month, you do an AI tool sharing session, it becomes a case study. Three months later, you write about your personal methodology, it becomes supporting evidence.

That’s a real knowledge base. If materials are just stored and never looked at again, it’s just a bookmark collection.

I recommend doing a weekly review:

Please read the materials added this week in the ima knowledge base. Generate a weekly review for me. Include: New content added this week, recurring themes, materials worth close reading, topics that can be developed into articles, information that needs to be supplemented, action suggestions for next week. Do not delete, move, or overwrite any materials.

This weekly review is the real sign that your personal knowledge base has started running. Because it forces you to process materials every week – not just keep throwing things in, but regularly filter, review, and use them.

9. Remote Triggering: Handle Low-Risk Tasks Even When Away from Your Computer

Compared to ordinary knowledge base tools, WorkBuddy has a feature well-suited for domestic users: It can trigger tasks remotely via WeChat, WeCom, QQ, DingTalk, Feishu (Lark), and other entry points.

This capability is great for handling low-risk knowledge base tasks.

For example, you’re out and suddenly think of a question:

Search the ima knowledge base for materials related to “AI workflow”. Read only, do not modify. Output: The titles of relevant materials, a one-sentence summary, their source location, and the 5 most worth continuing to examine.

Or, after a meeting, you want your computer to organize the meeting notes first:

Read the meeting transcription content added today. Organize: Meeting topic, discussion points, clear conclusions, action items, responsible persons, deadlines, and pending questions. Save the result as a review. Do not generate formal documents.

Remote tasks must be used with restraint.

Principle: Only do low-risk processing remotely. Formal modifications must be confirmed back at your computer.

10. What Kind of Materials Are Suitable for ima?

Category 1: Articles

Public account articles, web articles, industry analyses, product introductions, tutorials, case studies.

Suitable for placing in 01-Material Library. WorkBuddy can help you extract titles, summaries, core viewpoints, quotable content, and related topics.

Category 2: PDFs

Industry reports, white papers, course handouts, product documentation, research materials.

Don’t read the entire PDF carefully from the start. First, have WorkBuddy extract the table of contents, chapter summaries, key data, page numbers, and risk warnings. Especially note: Critical data must retain its source. Data without a source should not be directly used for publication.

Category 3: Meeting Minutes

Meeting records are the easiest to scatter. An audio recording today, a document tomorrow, a WeChat message the day after. It’s recommended to standardize and put them all into 02-Project Library.

Have WorkBuddy output: Meeting topic, background, discussion points, clear conclusions, action items, responsible persons, deadlines, pending questions.

This way, meeting minutes don’t just become “heard and forgotten.”

Category 4: Video Subtitles

Courses, interviews, launches, podcasts, tutorial videos – they can all be converted to text first and then placed in the material library. WorkBuddy can help you organize chapters, key concepts, action suggestions, timestamped sections worth re-watching, and potential follow-up topics.

Category 5: Your Own Ideas

This type of content is even more important. Because external materials are just raw material. Your own judgment is the most valuable part of the knowledge base.

You can put them in 03-Output Library first and organize them slowly later. For the first version, don’t make it complicated. Just let them enter the system.

11. What Content Should NOT Be Put in Directly?

Not all content is suitable for a knowledge base. Especially the following categories must be handled with caution:

• Client data that has not been desensitized.
• Sensitive information like contracts, quotes, ID card numbers, phone numbers, bank card numbers.
• Internal company secrets.
• Final judgment materials for medical, legal, or investment decisions.
• Unverified second-hand information.
• Large collections of web bookmarks with no clear purpose.

An AI knowledge base is not a safe. Once sensitive data enters an automated workflow, you need to consider model calls, third-party services, logs, sync directories, and permission boundaries.

If you’re not sure whether something can be put in, don’t put it in. If it must be processed, desensitize it first:

• Change client names to “Client A”.
• Delete phone numbers.
• Obscure contract amounts.
• For internal information, keep only the structure, not the original text.

For beginners, the safest principle is: Content that can be published publicly can be prioritized for storage. Content that cannot be published publicly should not be stored yet.

12. Keep File and Knowledge Base Names Simple

Many people’s knowledge bases are messy, not because the tool is bad, but because their naming is too casual. Today it’s called “Materials Organize”, tomorrow it’s “AI Summary”, the day after it’s “Final Version”. A week later, even you don’t know which is which.

Naming for the first version doesn’t need to be fancy, but it needs to be stable:

The purpose of naming isn’t to look good. It’s so you can find things later. The worst thing for a personal knowledge base isn’t having too few materials – it’s having materials inside but not being able to find them.

13. 5 Instructions Beginners Can Copy Directly

Instruction 1: Organize the Material Library

Please read the “01-Material Library” in the ima knowledge base. Organize the recently added materials by topic. For each item, output: Title, one-sentence summary, core viewpoint, quotable content, related topics, whether it’s worth close reading. Do not delete, move, or overwrite any materials.

Instruction 2: Organize Project Materials

Please read the “02-Project Library” in the ima knowledge base. Compile a project material pack related to “[Project Name]”. Output: Project background, existing materials, key questions, available resources, missing information, next steps. Do not generate a final plan.

Instruction 3: Generate Topics

Please read the “01-Material Library” and “03-Output Library” in the ima knowledge base. Generate 10 content topics related to “[Topic]”. Each topic includes: Title, core viewpoint, quotable materials, suitable angle, risk points. Do not write full articles directly.

Instruction 4: Organize a Meeting

Please read the latest meeting record. Organize: Meeting topic, background, discussion points, clear conclusions, action items, responsible persons, deadlines, and pending questions. Output the result as a review. Do not generate formal external documents.

Instruction 5: Weekly Review

Please read the materials added this week in the ima knowledge base. Generate a weekly review. Include: New content added this week, recurring themes, materials worth close reading, topics that can be developed into articles, information that needs to be supplemented, action suggestions for next week. Do not delete, move, or overwrite any materials.

Master these 5 instructions first. Don’t chase automation from the start.

14. After It Runs Smoothly, Then Consider Automation and Skills

If you only organize materials occasionally, just sending a task directly is enough. If you need to organize them weekly, turn them into fixed templates. If the templates run stably many times, then consider automation. If automation is also stable, then distill it into a Skill.

The correct progression order is:

One-time Task → Fixed Instruction → Fixed Template → Weekly Review → Automation → Skill

Don’t skip steps. Many people fail because they want to build an “All-Powerful Personal Knowledge Base Assistant” on day one – it needs to read materials, categorize, summarize, write articles, do project management, and archive automatically. It will inevitably become a disaster.

The first Skill shouldn’t be all-powerful. Solve only one clear problem:

• ima Material Library Organization Skill
• Meeting Minutes Review Skill
• PDF Reading Notes Skill
• Content Topic Organization Skill
• Project Material Pack Skill

If you can run one stably, then build the second. A knowledge base system isn’t built all at once. It grows slowly, starting from one small reusable process.

Automation Example: In the “Automation” section on the left in WorkBuddy, create a new task. Set it to trigger every Friday at 17:30. Use the review instruction as the prompt. WorkBuddy will then automatically read the ima knowledge base, generate a weekly report, and save it back to ima every week. This is a truly long-running knowledge management system.

15. Common Problem Troubleshooting

Problem 1: Knowledge Base Can’t Be Read After Authorization

Cause: You have multiple ima accounts, and the authorized one is not the one containing the knowledge base.

Solution: Re-authorize in the WorkBuddy connector settings, ensuring you select the correct ima account. ima and WorkBuddy must be logged in with the same WeChat account.

Problem 2: WorkBuddy’s Responses When Reading the Knowledge Base Are Low Quality

Cause: The instruction is too vague. The AI doesn’t know what content to read or which dimensions to extract.

Solution: Make the instruction specific regarding “which knowledge base to read”, “which dimensions to focus on”, and “what format to output”.

• ❌ Bad: Analyze the content in the knowledge base for me.
• ✅ Good: Access the “Product Requirements” knowledge base. Extract requirement points related to “User Experience”. Rank them as High, Medium, or Low priority. Output as a list.

Problem 3: ima Knowledge Base Shows 0 Items

99% of the time, it’s one of these 4 reasons. Check them in order:

• The file didn’t enter the knowledge base, only existed in a temporary conversation → Go to the corresponding knowledge base in the ima client and confirm if the file is there.
• The file is still being parsed → Wait 1–3 minutes after uploading. It needs to show “Parsing Complete” to be searchable.
• Wrong library linked or wrong key → Send “View ima configuration” to check the information.
• Cache hasn’t updated → Send “Refresh ima knowledge base cache” and check again.

Problem 4: Saving Output Back to ima Fails

Cause: The ima knowledge base’s permissions are set to “Only I can write”, and the connector’s authorization isn’t sufficient.

Solution: In ima, confirm the knowledge base permission is set to “Allow writing”, or when re-authorizing, check the “Write permission” box.

16. First, Run One Minimum Workflow

If you’ve read everything and only want to do one thing, I recommend this:

Step 1: Create three knowledge bases in ima: 01-Material Library, 02-Project Library, 03-Output Library.

Step 2: Put in the first batch of materials – 5 articles, 3 PDFs, 3 video subtitles, 5 idea notes, 3 project materials.

Step 3: Have WorkBuddy read the material library and generate a review.

Step 4: Human check – verify if sources are retained, summaries are accurate, tags are useful, topics have supporting materials, and action items are clear.

Step 5: Only retain the valuable content.

Don’t aim for perfection in one go. Don’t try to empty all your bookmarks from the start. First, process a small batch of real materials and see if this combo reduces your organization burden. If it works, then expand.

Conclusion: A Second Brain Isn’t About Having More Materials

I increasingly feel that the term “Second Brain” can be misleading.

Many people think a second brain means storing everything – web pages, PDFs, screenshots, meetings, videos, notes – the more, the better.

But the real questions are:

• Once materials are in, who processes them?
• Who filters them?
• Who reviews them?
• Who judges if they’ll be useful in the future?

Without this step, even the most advanced knowledge base is just a warehouse.

The value of WorkBuddy + ima isn’t giving you a fancier bookmark collection. It’s breaking down the personal knowledge base into a process more suitable for beginners:

ima handles storage. WorkBuddy handles organization. You handle judgment.

I won’t let WorkBuddy take over the entire knowledge base. I only let it do three things first: read materials, generate reviews, and retain sources. The final decision on whether something enters the formal knowledge base is still mine to make.

This is how I understand building a personal knowledge base from 0 to 1.

It’s not building a complex system. It’s first ensuring that every batch of materials entering the system gets processed, gets filtered, and gets a decision on whether it’s worth retaining.

About the Author

Punk | MBA from USTC | Chief Designer at HerName | Executive Dean at Stanley Business School |

| AI Prompts | Sharing AI / money-making methods that beginners can understand and replicate easily |

@AdrianPunk115