Your Personal Knowledge Graph: How Ditto Maps Your Thinking
Every conversation you have contains hidden structure. You mention people, reference projects, connect ideas across topics, and build on concepts over weeks and months. Traditional AI assistants throw all of that away. Ditto doesn’t.
Ditto automatically extracts subjects — people, topics, technologies, concepts — from every conversation and weaves them into an interactive knowledge graph. The result is a visual map of your thinking that grows with every message.
No other AI assistant does this. ChatGPT, Claude, and Gemini offer some form of memory, but none give you a navigable, interconnected graph of the ideas and relationships you’ve built over time.
What Is a Knowledge Graph?
A knowledge graph is a network of interconnected concepts. Instead of storing your conversations as a flat list, Ditto identifies the subjects within them and tracks how those subjects relate to each other through your conversations.
Think of it like this:
- Traditional AI memory: A filing cabinet. Conversations go in, and you hope the AI finds the right one later.
- Ditto’s knowledge graph: A mind map that grows automatically. Subjects connect to each other based on how you’ve discussed them together.
When you talk about “React” and “TypeScript” in the same conversation, Ditto links those subjects. Over time, clusters emerge — your work projects, your learning interests, your personal goals — all visible in a single interactive visualization.
How It Works Under the Hood
Automatic Subject Extraction
Every time Ditto syncs your conversations, an extraction pipeline runs in the background:
- Conversation pairs (your message + Ditto’s response) are analyzed
- Subjects are extracted — named entities, technical topics, abstract concepts, people
- Vector embeddings are generated for each subject, enabling semantic search
- Co-occurrence links are created when subjects appear in the same conversation
This happens automatically. You don’t tag, categorize, or organize anything — Ditto builds the graph from the raw signal of your conversations.
Subject Co-occurrence: The Secret Sauce
The most powerful part of the knowledge graph isn’t individual subjects — it’s the connections between them. Ditto tracks which subjects appear together in your conversations through a pair_subject_links system.
When you discuss “API Design” and “Authentication” in the same conversation, those subjects become linked. The more you discuss them together, the stronger the connection. This creates a weighted graph that reflects your actual thinking patterns, not just a list of topics.
These connections power several features:
- Visual clustering in the network view — related subjects naturally group together
- Contextual memory retrieval — when you ask about one subject, Ditto can pull in related subjects
- Subject-based memory filtering — click any subject to see all conversations that mention it
- Thought Trains — curate subjects and memories into focused context collections
Semantic Search Across Your Graph
Every subject in your knowledge graph has a vector embedding, which means you can search semantically — not just by exact keywords. Search for “frontend frameworks” and you’ll find subjects like “React”, “SolidJS”, “Vue”, even if you never used the exact phrase “frontend framework” in conversation.
This is powered by the same vector search infrastructure that drives Ditto’s memory retrieval system, extended to work at the subject level.
The Interactive Network Visualization
Ditto renders your knowledge graph as an interactive network using a force-directed layout. Here’s what you see when you open the Memory Network view:
- Nodes represent subjects — sized by how often they appear in your conversations
- Edges represent co-occurrence — connecting subjects that appear together
- Clusters emerge naturally — your work topics group together, your personal interests form their own cluster
- Click any node to see all conversations mentioning that subject
- Zoom, pan, and drag to explore your entire knowledge landscape
The visualization isn’t just eye candy. It surfaces patterns you might not notice in a flat conversation list:
- Which topics dominate your AI conversations?
- What unexpected connections exist between your interests?
- How has your focus shifted over time?
For example, you might discover that “TypeScript” and “Zod Validation” have become heavily linked in your graph over the past month — reflecting your deep dive into type-safe validation patterns. Or you might notice that “Job Search” and “React” cluster together, revealing that your technical learning is being driven by career goals.
Knowledge Graph + Memory: Better Together
The knowledge graph doesn’t replace Ditto’s memory system — it enhances it. Here’s how they work together:
Smarter Memory Retrieval
When you ask Ditto a question, the memory system doesn’t just search for semantically similar conversations. It can also use subject relationships from the knowledge graph to broaden or narrow its search. If you ask about “OAuth”, Ditto knows from your graph that this is connected to “Firebase Auth” and “API Security”, so it can pull in relevant context from those related subjects.
Subject-Based Filtering
The top subjects dashboard shows your most-discussed topics at a glance. Click any subject to filter your conversation history to just that topic — far more useful than scrolling through hundreds of messages trying to find “that conversation about databases.”
Memory Network Traversal
Ditto’s MCP integration exposes a get_memory_network tool that lets any connected AI assistant traverse your knowledge graph. This means Claude, Cursor, or any MCP-compatible tool can navigate from one memory to related memories through shared subjects — following the same connections you see in the visual graph.
Real-World Examples
The Developer
A software engineer has been using Ditto for three months. Their knowledge graph shows:
- A dense cluster around “React” → “SolidJS” → “TypeScript” → “Vite” — reflecting a frontend migration project
- “PostgreSQL” and “Supabase” linked through multiple database conversations
- “CI/CD” connecting to both “GitHub Actions” and “Docker” — DevOps discussions spanning weeks
- A smaller cluster of “Interview Prep” → “System Design” → “Algorithms” — career development conversations
When this developer asks “What approach did we discuss for the database migration?”, Ditto uses the knowledge graph connections to surface the right conversations — not just any conversation mentioning databases, but the ones specifically linked to their migration project subjects.
The Researcher
A graduate student uses Ditto to process research papers and brainstorm ideas. Their graph shows:
- “Machine Learning” → “Transformers” → “Attention Mechanisms” — core research topics
- “Literature Review” connecting to five different paper titles
- “Thesis” → “Methodology” → “Experiment Design” — a project-planning cluster
- Unexpected connection: “Cooking” linked to “Chemistry” — from a conversation about molecular gastronomy that sparked a research idea
The Creative
A writer brainstorming a novel series sees clusters for each character, plot thread, and world-building element:
- “Protagonist” → “Backstory” → “Motivation” — character development conversations
- “Magic System” → “Rules” → “Limitations” — world-building details
- “Chapter 12” connecting to “Plot Twist” and “Foreshadowing” — structural decisions
How Ditto’s Knowledge Graph Compares
| Feature | Ditto | ChatGPT | Claude | Gemini |
|---|---|---|---|---|
| Cross-conversation memory | ✅ Visual + proactive | 🟡 Silent, basic | ❌ Per-project only | 🟡 Extensions |
| Knowledge graph | ✅ Interactive network | ❌ | ❌ | ❌ |
| Subject extraction | ✅ Automatic | ❌ | ❌ | ❌ |
| Memory visualization | ✅ Force-directed graph | ❌ | ❌ | ❌ |
| Subject-based search | ✅ Semantic | ❌ | ❌ | ❌ |
| Co-occurrence tracking | ✅ Weighted links | ❌ | ❌ | ❌ |
| Graph traversal via MCP | ✅ get_memory_network | ❌ | ❌ | ❌ |
No other consumer AI platform gives you an interactive, automatically-built knowledge graph of your conversations. This is Ditto’s most unique differentiator — and it’s available today.
Getting Started
Your knowledge graph starts building the moment you create a Ditto account. There’s nothing to configure:
- Sign up at assistant.heyditto.ai
- Have conversations — ask questions, brainstorm, debug code, plan projects
- Open Memory Network from the sidebar to see your graph grow
- Click subjects to explore your conversation history by topic
- Connect via MCP to access your knowledge graph from Claude, Cursor, or any MCP-compatible tool (setup guide)
The more you use Ditto, the richer your graph becomes. After a few weeks, you’ll have a visual map of your thinking that no other AI tool can replicate.
Ditto is built by Omni Aura, the team that’s been building memory-first AI since before ChatGPT existed. Read our origin story →