Wiki and Knowledge Graph
Introduction
The Sutram Wiki is a layer of knowledge over your project. While the Content tab organizes files and folders (and can grant access to governed content), the Wiki organizes ideas and how they connect: concepts, themes, and documents become nodes linked to one another by mentions and relations, forming a navigable knowledge graph.
More than an internal encyclopedia, the Wiki is the project's structured memory — the foundation that lets both your team and AI assistants find, understand, and connect what has already been produced, without depending on whoever "was there when that was decided."
When to use the Wiki
The Wiki adds value when a project has:
- Knowledge that accumulates over time — decisions, concepts, and definitions that need to be found again months later
- Many documents that reference each other — regulations that amend others, reports that cite exams, specifications that depend on drawings and other specifications
- Multiple participants — people who come and go during the project and need to understand the context quickly
- AI use — when you want an assistant to answer questions anchored (guard-rails) in the project's real material
If your project is just a file repository with no relations between its items, the Content tab is enough; the Wiki comes into play when the ontology and the connections matter.
Requirement: The Wiki is a feature of the paid plans (starting with Pro) and must be enabled by the owner in the project settings.
Core Concepts
Wiki Node (WikiNode)
The node is the Wiki's unit: a referenceable entity that represents a concept, a theme, or a document. Unlike a file, a node does not store bytes — it stores meaning: a name, a slug (short identifier), an optional synthesis, and metadata.
There are two main types of node:
| Type | What it is | How it arises |
|---|---|---|
| Concept node | An idea or theme (e.g., "Customer onboarding", "Bioequivalence") | Created deliberately (via AI/MCP) or as a stub when it is mentioned before it exists |
| Document-mirror node | The reflection of a project Record in the Wiki | Generated by the AI from the Records; carries the document text extracted from the document's files (PDF, Office, etc.) |
Mentions [[slug]]
A mention is a reference to another node, written as [[slug]] inside a node's text. It becomes a clickable link on the node's page (or accessible directly by the AI). You can also display a label different from the slug with [[slug|display text]].
When you mention something that does not yet exist, the mention stays pending (marked as broken) and creates a stub — a ghost node that resolves as soon as the real target is created. This lets you build the network incrementally, citing concepts before documenting them.
Backlinks
The backlink is the inverse of the mention: if node A mentions node B, then A appears in B's backlinks — the "Mentioned in" panel. This is how you discover every place where a concept appears in the project, without searching file by file. Each node displays its backlink count.
Typed relations
A simple mention says "A cites B." A typed relation says how A relates to B: regulates, revokes, amends, depends on, and so on. The label appears on the graph edge and gives the connections semantics — turning a list of links into a network with meaning.
Synthesis vs. Full text
Each node can have two levels of content:
- Synthesis — a curated, editorial summary of what the node represents (the "entry").
- Full text — in mirror nodes, the complete content of the source document, indexed for search.
The Wiki search covers both, so a question finds both the summary and the body of the document.
The Knowledge Graph
When nodes connect through mentions and relations, they form a graph — visible in the Wiki's Graph section. There you see the nodes as points and the relations as labeled lines, and you can:
- Search and filter by node name or type
- Open a node to read its synthesis and follow its connections
- Explore a node's neighborhood via the ego-graph (the radial graph centered on it)
- Rearrange the layout and navigate through clusters of related themes
The graph is not decorative: it is a way to see the structure of the project's knowledge — which concepts are central, what is isolated, how things link together.
The Wiki as memory for AI
The Wiki was designed to be the source an AI assistant consults when answering about the project. The typical flow is:
- Ask — the AI runs a full-text search of the Wiki based on your question
- Read — it opens the most relevant nodes, with synthesis and full text
- Navigate — it follows mentions and backlinks to gather the surrounding context
- Answer — it synthesizes the answer anchored in the sources it found
This is what sets the Sutram Wiki apart from a generic chatbot: the answers come from your project's real material, with a trail back to the source documents. For the step-by-step of how an AI operates the Wiki, see the Sutram MCP Server Guide.
Wiki, Content, and Records
The Wiki does not replace the Content tab — it complements it:
| Layer | Stores | Answers |
|---|---|---|
| Content | Files and folders | "Where is file X?" |
| Records | Structured data with metadata | "Which documents in category Y?" |
| Wiki | Concepts, relations, and syntheses | "What is X and how does it connect to Y?" |
The Records feed the Wiki: each Record can have a mirror node that brings its text into the search and into the graph. The mirroring is done by an AI assistant via MCP — it does not happen on its own. This way, what you organize as structured data also becomes navigable knowledge.
Roles and Permissions
Access to the Wiki follows the project roles:
| Action | Owner / Admin | Member | Viewer |
|---|---|---|---|
| Navigate, search, and read nodes | Yes | Yes | Yes |
| Explore the graph and backlinks | Yes | Yes | Yes |
| Create and edit nodes, synthesis, and relations | Yes | Yes | No |
Reading the Wiki is available to any project participant, including viewers. Writing (creating/editing nodes, defining relations) follows the editing roles and is done via an AI assistant / MCP Server.
Best Practices
- Let knowledge emerge from the documents (bottom up) — start by mirroring Records into the Wiki; the network grows as the texts cite each other.
- Name concepts consistently — the slug is the node's address; consistent names avoid duplicates and broken mentions.
- Prefer typed relations where the "how" matters — "revokes", "depends on" say far more than a simple mention.
- Use mentions even for what does not yet exist — the pending mention becomes a visible reminder that a concept needs to be documented.
- Write short, objective syntheses — the entry is a starting point, not a copy of the document; the full text already covers the detail.
Frequently Asked Questions
Q: What is the difference between the Wiki and the Content tab?
A: Content stores folders, files, and links; the Wiki stores concepts and connections. Content answers "where is the file?"; the Wiki answers "what is this and how does it relate to the rest?". The two coexist — Records from the Content tab feed mirror nodes in the Wiki.
Q: Who builds and maintains the Wiki?
A: AI assistants, via the Sutram MCP. All Wiki maintenance — creating mirror nodes from the Records, writing concept nodes and syntheses, defining typed relations, syncing the document text — is done by an assistant connected to the project. None of this is created on its own by the system, and the web interface does not offer that functionality: it is the place to navigate, search, and read. In practice, you talk with the assistant and it builds the network; afterward, any participant consults the result through the interface.
One partial exception: the [[slug]] mentions written in a node's body create stubs for concepts not yet documented — but that too only happens when the node is saved via MCP.
Q: What is a "pending" or "broken" mention?
A: It is a [[slug]] mention whose target does not yet exist. It stays marked as broken and creates a ghost node (stub); when the real target is created, the mention resolves normally.
Q: Can I edit the Wiki through the web interface?
A: The web interface is geared toward navigating, searching, and viewing the graph and the nodes. Creating and editing nodes, syntheses, and relations is done via the Sutram MCP Server (by an AI assistant). Viewers can navigate, but not edit.
Q: Does the Wiki search cover the content of the documents?
A: Yes. In mirror nodes, the document's full text is indexed, so a search finds both the synthesis and the body of the original document.
Document Version: 1.0 Last Updated: July 2026 Author: Sutram Development Team