⚠️ Forward looking statements
Natural Language Graph Construction with Copilot Integration
Overview
Unigraph’s Copilot Integration enables users to construct, edit, and navigate complex graphs using natural language. This system serves as a conversational assistant that turns plain English commands into structured graph operations, allowing users to build and manage information without needing to understand internal schemas or UI complexities.
This feature dramatically lowers the barrier to entry for new users, speeds up prototyping, and augments expert workflows with an intuitive, low-friction interface.
Why This Capability Matters
As knowledge graphs grow in complexity, managing them through traditional interfaces becomes time-consuming and often intimidating. Manual UI interactions can be tedious for:
- Adding multiple interconnected nodes,
- Renaming or restructuring portions of a graph,
- Exploring hidden or complex relationships,
- Modifying types or metadata in bulk.
Natural language interfaces solve this by turning the intent of the user directly into graph operations, bypassing the need for low-level manipulation. Unigraph’s Copilot is not just a command parser — it is a graph-literate assistant that understands:
- Graph structure,
- Type hierarchies,
- Entity relationships,
- User-created models and data.
Core Capabilities
| Copilot Feature | What It Does |
|---|---|
| Graph Creation | Describe your idea in natural language, and Copilot will generate the corresponding nodes, edges, types, and initial layout. |
| Graph Editing | Ask Copilot to add, rename, connect, or delete graph elements. It understands references like “the node about climate change” or “connect John to the project node”. |
| Exploratory Questions | Query the graph conversationally: “What are the most connected nodes?”, “Show me all tags related to AI”, or “Highlight projects assigned to Sarah”. |
| Perspective Generation | Ask Copilot to create different views of the graph, e.g., “Create a timeline of events involving Company X” or “Make a view that only shows nodes with centrality > 0.8”. |
| Multi-modal Commands | Combine natural language with clicks and selections: e.g., “Connect this node to everything related to education.” |
| Data-aware Suggestions | Copilot leverages your graph’s types and entities to make intelligent suggestions, helping users refine and extend their structures naturally. |
Use Cases
| Use Case | Description |
|---|---|
| Non-technical user onboarding | New users can create knowledge graphs without understanding schemas, config files, or syntax. |
| Rapid prototyping | Quickly generate graph structures during brainstorming or ideation. |
| Conversational modeling | Build a model by describing relationships between people, organizations, concepts, etc., in a human-friendly way. |
| Editing at scale | Update names, tags, types, or connections in bulk by describing what needs to change. |
| Graph discovery | Ask natural-language questions to uncover structures, bottlenecks, missing links, or anomalies in a graph. |
| Semantic scaffolding | Use Copilot to draft the high-level structure of an ontology or model before refining it in the UI. |
Workflow Overview
Start Conversing
- Launch Copilot from any graph view or global prompt.
- Example input:
“Create a node called ‘Photosynthesis’ and link it to ‘Plants’ and ‘Sunlight’.”
Context-Aware Understanding
- Copilot uses the surrounding graph context, entity types, and previous queries to interpret your command.
Execution Preview
- Before applying changes, Copilot previews the impact: “This will create 1 new node and 2 new edges. Proceed?”
Apply and Continue
- The graph is updated live, and users can continue the conversation with follow-ups:
“Now group them into a cluster called ‘Energy Cycle’.”
Advanced Actions
- Generate alternative perspectives, set filters, attach metadata, define types — all through plain language.
Why Unigraph’s Copilot is Unique
Many AI assistants offer basic command-line parsing. Unigraph’s Copilot is fundamentally different:
✅ Graph-native: It understands graph structure, centrality, types, relationships, and layout.
✅ Semantically aware: It leverages type information, entity tags, and metrics to ground user intent. ✅ Contextual memory: Operates within the current workspace, using what’s already present in your graph. ✅ Perspective-driven: Supports dynamic creation of alternate views without duplicating the model. ✅ Composable: Combine with other Unigraph features (e.g. image annotation, dataset import, timeline mode) through language.
Example Prompts
“Create a new node for ‘Artificial Intelligence’ and link it to ‘Machine Learning’, ‘Neural Networks’, and ‘Ethics’.”
“Color all nodes tagged ‘Important’ in red.”
“Make a perspective that shows all research papers from 2021 connected to John.”
“List the 5 nodes with the highest centrality.”
“Convert this group into a typed subgraph called ‘Team A’.”