Read me first

A plain-language introduction to SynerWiki. No technical background assumed. If you only read one document, read this one.

---

1. Who this is for

People who are responsible for what their organisation knows.

In practice that usually means someone in engineering, customer support, HR, or operations who has watched the same question get asked five times, found three different answers written down, and had no way to tell which one was still true.

You do not need to be technical to use this software. Somebody technical will need to install it, and there are notes for them elsewhere in this folder, but the day-to-day work of writing, organising, finding and reviewing knowledge is ordinary desk work.

2. What it does

It is a shared place to write things down and find them again.

You create documents, group them into collections, and everyone who should see them can. Several people can edit the same page at the same time. Every change is kept, so you can always look back at what a page used to say. You can search across everything, star the pages you use constantly, and — importantly — see which documents have quietly gone out of date so somebody can fix them.

That last part is the difference between a knowledge base and a pile of old pages.

3. The domain: why written knowledge decays

Every organisation accumulates knowledge that lives in people's heads: how the billing system really works, what to tell a customer who asks about refunds, what happens on someone's first day, which server to restart at 3am.

When that knowledge is written down at all, it tends to be scattered — a document here, a chat message there, a spreadsheet someone made two years ago. Three problems follow, and they are the reason this software exists.

Nobody can find it. The answer exists, but the person who needs it does not know the right words to search for, or does not know it was written down at all.

Nobody knows if it is still true. This is the dangerous one. Documentation does not announce that it has expired. Wrong instructions sit there looking exactly as authoritative as they did the day they were correct, and people follow them.

Nobody owns it. Writing documentation is everyone's job, which means it is nobody's job. Reviewing old documentation is worse, because there is no moment when it obviously needs doing.

A wiki solves the first problem. This software is built to take the second and third seriously too: it tracks how fresh each document is and shows you what needs attention, so keeping knowledge trustworthy becomes a visible task rather than a vague sense of guilt.

4. Where it fits

One organisation per installation. This is deliberate. Your knowledge base is yours alone — it is not a shared service with other companies' data sitting beside yours in the same system.

Who uses it. Anywhere from a handful of people to a large department. Everyone signs in with their existing work identity through a shared identity service, so there is no extra password for anyone to forget or for you to administer.

What it works alongside. As shipped, it can connect to:

• Slack and Google Chat. Your team can search the knowledge base, ask it a question, or link a document from inside a chat conversation, without switching windows.
• Automated user management (SCIM). If your organisation already manages staff accounts centrally, it can create and remove accounts here automatically as people join and leave.
• Other systems, via webhooks and an API. When content changes, this software can notify another system so it can react. And anything you can do in the screens, another program can do through the interface described in api-reference.md.
• Existing wikis and files. You can import a Confluence export, a zip archive of pages, or individual Markdown files, and export back out as Markdown or HTML. More on that below.

The AI features are optional and switched off by default. Search by meaning, the question-answering assistant, and the automatic concept map all need a key from an AI provider, which you supply and pay for. Until you do, nothing calls an outside service and nothing costs you anything — keyword search and everything else work normally. This is worth knowing before you evaluate: if a screen tells you a feature is "not enabled", that is why.

5. First run

Sign in. Whoever signs in first becomes the administrator.

That is the whole first step, and it matters more than it sounds. On a brand-new installation the first person to sign in is made the administrator of the workspace. So the first sign-in should be you, or whoever is meant to own this, and it should happen promptly after installation rather than a week later.

Everyone who signs in after that arrives without access until an administrator gives it to them. Nothing is lost if the wrong person goes first, but sorting it out is more annoying than getting it right.

If you are looking at the preview version to decide whether you want this at all, none of the above applies — it opens straight into a ready-made example workspace. See "The example data in the preview" below.

6. Setting up

Work through this in order. None of it is urgent on day one except the first item.

1. Invite the people who will use it. Ask them to sign in, then set each person's role (see "Who can do what" below). If your organisation manages accounts centrally, turn on automated user management instead and skip the manual step.
2. Create your collections. These are the big buckets — Engineering, Support, People, Operations. Resist the urge to make thirty of them. Five or six broad collections that everybody understands beats a precise structure nobody can navigate.
3. Nest pages where it helps. A document can sit inside another, so a long topic can have child pages instead of one unreadable wall of text.
4. Decide who can see what. Access can be granted per collection, either to an individual or to a team, at three levels: read, read and write, or full control. Most collections should simply be open to everyone — restrict only what genuinely needs it, such as HR material.
5. Bring your existing documents in. If you are moving off Confluence, export it from there and import the result directly. Zip archives of pages and individual Markdown files also import. Anything else is usually faster to copy and paste into a new page.
6. Connect your chat tool, if you use Slack or Google Chat. This is the single change most likely to get people actually using the knowledge base, because it meets them where they already are.
7. Only then consider the AI features. Get the basics working first. Search by meaning and the answering assistant are genuinely useful, but they are an improvement to a working knowledge base, not a substitute for one.

7. Day to day

Writing. Create a page in a collection and write. The editor works the way you expect: headings, lists, tables, links, images, code. Typing / opens a menu of things to insert, and typing @ mentions a colleague or links to another document. If someone else is editing the same page, you will see their cursor and your changes will merge — no locked files, no "document in use".

Reading and finding. Search from the top of any screen. You can search by keyword, by meaning, or both at once, and each result tells you why it matched. Star the pages you keep coming back to and they collect on your Starred page, which is yours alone.

Discussing. Comment on a document rather than starting a separate chat thread, and reply to comments to keep a thread together. The conversation stays attached to the thing it is about, where the next person will find it.

Correcting. Just fix it. Every version is kept, so a mistaken edit is never destructive and you can see what changed and when.

8. Ongoing care

Little and often. Roughly:

Every week or two — review what has gone stale. The Content governance screen lists documents that have not been touched in a long time (ninety days by default). Work down the list: confirm the ones still correct, fix the ones that are not, and retire what is no longer needed. Twenty minutes a fortnight keeps the whole thing trustworthy; six months of neglect produces a knowledge base nobody believes.

Occasionally — check who has access. People change roles and leave. If you use automated user management this mostly takes care of itself.

Regularly — make sure someone is backing up the database. This is a job for whoever installed the software, and it is the one piece of ongoing care that is genuinely technical. Ask them to confirm it, and ask them to confirm they have tested restoring it. An untested backup is a hope, not a backup.

When you want a copy — export it. You can export a page or an entire collection as Markdown or HTML at any time. Useful for sharing outside the system, for archiving, and for reassurance that your knowledge is not trapped in here.

---

Glossary

• Workspace — your whole knowledge base. One organisation, one workspace.
• Collection — a top-level grouping of documents, like a folder.
• Document — a single page. Documents can be nested inside each other.
• Revision — a saved earlier version of a document. Kept automatically.
• Stale — a document nobody has touched for a long time, and which therefore may no longer be true. Being stale is a prompt to review, not proof of error.
• Team — a named group of people, so access can be granted to the group instead of person by person.
• Permission — who may see or change a collection or document: read, read and write, or full control.
• Semantic search / search by meaning — finding documents about the same idea even when they use different words. Needs an AI provider key.
• Answer bot — asks your knowledge base a question and answers in prose, citing the documents it used. Needs an AI provider key.
• Concept map / knowledge graph — an automatically inferred picture of how topics across your documents relate. Needs an AI provider key.
• Webhook — an automatic message sent to another system when something changes here.
• API key — a credential that lets another program use this software on your behalf.

The other documents in this folder

• getting-started.md — the practical first steps once the software is running. Read after this one.
• useful-commands.md — day-to-day commands for whoever operates the installation. Technical.
• production-initialization.md — how to prepare a brand-new real installation for its first use. For whoever installs it, and worth reading before they do: it describes a step that erases data.
• api-reference.md and openapi.yaml — the programming interface, for connecting other systems.
• testing.md — how the software is tested. Of interest if you are assessing how carefully it was built.

Who can do what

Everyone has one workspace-wide role:

• Administrator — full control, including people and access.
• Member — the normal role. Reads and writes where permitted.
• Viewer — reads, does not change.
• Guest — like Viewer, treated as read-only. The software makes no further distinction between the two today, so use whichever label better describes the person.

On top of that, individual collections and documents can be shared with a person or a team at read, read-and-write, or full-control level. A workspace can never be left with no administrator at all.

The example data in the preview

The preview version ships with a ready-made example workspace for a fictional company called Northwind — several collections, a few dozen documents, and a handful of invented people. It exists so you can judge the software with something in it. Every name, document and email address is made up.

It is a demonstration, not a starting point. When you set up a real installation you begin with an empty knowledge base rather than deleting Northwind's content one page at a time — whoever installs it has a documented step for exactly that (production-initialization.md).

Common questions

Do people need another password? No. Sign-in uses your organisation's existing identity service.

What happens if two people edit the same page at once? Both edits are kept and merged, and you can see each other working. This is normal use, not a problem to avoid.

I searched for something I know is written down, and got nothing. Search needs submitting — type your query and press Enter. Also check whether the document is in a collection you have access to.

A screen says a feature is not enabled. That is an AI feature and no provider key has been supplied. Everything else works without one.

Is my content locked in? No. Export any page or collection as Markdown or HTML whenever you like.

Something is definitely wrong. Nothing is destroyed by editing — check the document's earlier versions first, since the previous text is still there.

Where to get help

This application is entirely AI-generated, and what you have is deliberately the core: the functionality is real, but there is no branding, styling, or company-specific vocabulary. It is meant to be shaped into what you actually need.

If you know your field and are comfortable working with AI, you can do that shaping yourself. If you would rather someone did it for you, or you want to ask a question, suggest an improvement, or see what other people have built, visit the community at wbsp.ai. Members have built many versions of this and of other applications — one of them may already be much closer to what you want than starting over.