Building Blocks
The app turns text into structured memory. It extracts entities and relationships from sentences, builds a knowledge graph, and saves it into the memory layer. When you query the system, an LLM synthesizes responses from that stored graph. As users add new facts, the graph is updated and reconciled with existing knowledge.
Deepthink uses multiple reasoning services depending on the task:
- Simple deduction for direct inference.
- Causal reasoning for cause-and-effect explanations.
- Hypothetical (if/else) reasoning for branching scenarios.
These services let the system choose the right inference path while keeping results grounded in the knowledge graph.