The Day I Hooked WorkBuddy into Feishu, I Could Run My Home PC from a Wedding
Community Discussion · Company Watch

The Day I Hooked WorkBuddy into Feishu, I Could Run My Home PC from a Wedding

ShutterShutter5d ago2026/09/28 155 views

Just saw WorkBuddy's official guide on Feishu integration—a whole page of permission configurations, looks like a procurement contract. I've used WorkBuddy for two months, just got this path working last month, and a glance tells me which steps are the real hurdles.

First, my scenario. I do wedding follow-shoots and e-commerce product photos. During peak season, two shoots a day, out the door at 7 AM, back at 10 PM. When I'm out, client messages don't stop—asking how far I've gotten with yesterday's batch, when the asset list will be ready. Before, I could only say I'll check when I get back. Now, sitting in a hotel lobby, I pull out my phone and tell the bot in Feishu, and the computer at home that's turned on starts running WorkBuddy, sorting the day's card photos into folders by camera position, and generating an asset list to send back. I still have to go through the list myself to check column names—previously when mixed-format files were fed in, it got confused several times. I accept this pitfall; checking myself is cheaper than redoing work.

The first step of this chain isn't in Feishu, it's in WorkBuddy. Open WorkBuddy, find Settings in the sidebar, there's an item Feishu Integration, click in and it gives you a place to fill in app credentials and a callback address. This step is just registration—don't rush to think it's done.

The second step is the longest part in that guide: go to the Feishu Open Platform and create an enterprise self-built app. You need permission to create apps—individual accounts or small teams usually have to ask an admin to enable it. After creating, go to the app details page, click Permission Management in the left menu, then click Batch Import/Export Permissions. Here's a pitfall I stepped in: in the popup's input box, clear it first, then paste your entire permission list, then click Confirm Add Permissions. The first time I didn't clear it, pasted and it said success, but the bot could only receive messages, not send them. After checking for a while, I found old content had squeezed out the new permissions.

Those strings starting with im:message, base:record in the permission list look like gibberish, but they're actually telling Feishu whether this bot can read group messages, write tables, send cards. The guide gives a big comprehensive list; I suggest copying it exactly for the first integration, then trimming after it works. Miss one permission, and the error won't tell you which is missing—it'll just silently not respond. That silence is the most draining.

The third step is easy to miss. After creating the app and adding permissions, you still need to go to Version Management and Release to create a version and submit for review. Enterprise self-built apps are usually approved instantly by admins. If you don't release a version, permissions are just on paper, and the bot won't reply. After releasing, search for the bot in the Feishu client and send a test command, like having it extract filenames from the latest folder on the desktop. Seeing the home computer respond—that's when the line is truly connected.

Then there's permissions and collaboration. I feel this more than most because in photography, the boundary of assets is the boundary of money. My approach is to create folders by project—weddings and e-commerce separate—and the bot is only authorized for the currently running project. Bots pulled into client groups get read-only permissions, can see lists and watermarked previews, but can't touch original RAW. Assistants get read-write, can rename files, add tags. Sounds troublesome, but this logic is the same as shooting authorization letters—who enters the camera position, who can only watch replays—draw the lines first, no arguments later.

Daily operations are really just two things. One, that computer must stay on and connected. WorkBuddy runs on the local machine; if the computer sleeps, commands sent out are like stones thrown into the sea. I set it to never sleep, at the cost of electricity. Two, revoke authorization the same day a project ends, archive the folder—otherwise three months later you yourself can't remember which bot is still connected to which batch of photos.

On trends, I recently saw a saying that Agents are moving from sitting at a computer to giving commands anytime, anywhere. To me, this isn't a concept—it's very concrete. Before, my day's boundary was the few hours the computer was on; now the boundary is when I remember. The asset ledger thing finally doesn't have to wait until I get home.

That's about it. My setup is personal Feishu plus an always-on desktop. Large teams might not find it applicable—best to ask your admin about permissions first.

2 replies

?
Ctrl + Enter to reply
Zhiyuan
Zhiyuan5d ago

Based on my own experience, when it comes to material mix-ups, my approach is to go through the list myself. Machine-generated lists are always just drafts to me.

Si Nan
Si Nan4d ago
Reply to Zhiyuan

@zhiyuanlaw Agreed, but his main text already assumes the step of "the checklist still has to go through column names yourself." What I'm more curious about is: under that assistant-tier read/write permission, who signs off first on the drafts the machine produces?