Graphmemory
Build and query embedded GraphRAG databases with DuckDB vector search
What You Can Do
Store entities and relationships in a persistent, embedded graph database. Run vector, full-text, and hybrid semantic searches across your graph structure. Extract entities and relations from unstructured text using DSPy, deduplicate nodes and edges intelligently, run graph algorithms like PageRank and centrality analysis, and visualize your knowledge graph interactively in a browser—all within a single Python package with zero server dependencies.
Features
semantic similarity search over embedded entities and relations
combine vector, full-text (BM25), and structured filters in one query
follow relationships across multiple levels to retrieve connected knowledge
automatically extract entities and relations from raw text with built-in deduplication
merge duplicate nodes by name similarity threshold (e.g., 0.9) without manual conflict resolution
PageRank, centrality measures, connected components via NetworkX integration
chainable API for intuitive graph queries without SQL
render your knowledge graph in the browser with zero dependencies
Example Output
Extract and store a knowledge graph:
Input: "Claude was created by Anthropic in 2022. Anthropic focuses on AI safety research."
Output nodes: [Entity(name="Claude", type="AI Model", created_by="Anthropic"), Entity(name="Anthropic", type="Company", founded=2022)]
Output edges: [Relation(src="Claude", relation="created_by", tgt="Anthropic")]
Semantic search and traversal:
Query: graph.search(query="AI safety company", top_k=5)
Result: [Anthropic (score: 0.94), Claude (score: 0.87), ...]
Multi-hop: graph.traverse(start="Claude", relation="created_by", depth=2)
Result: [Claude → created_by → Anthropic → founded_in → San Francisco]
What's Included
- SKILL.md: Complete API reference and decision table for insert, merge, search, and dedup operations
- Extraction template: DSPy-based prompt for entity and relation extraction from text
- Query builder examples: Copy-paste patterns for vector search, hybrid search, and multi-hop traversal
- Deduplication workflow: Step-by-step guide for fuzzy matching and node/edge merging
- Visualization snippet: HTML/JavaScript template to render your graph in an interactive browser view
Who It's For
- Data engineers and ML engineers building RAG systems that require graph traversal for retrieval
- Knowledge management teams extracting and organizing entities from research papers, documents, or domain-specific text
- Prototyping teams needing a lightweight knowledge graph without managing external databases
- AI researchers experimenting with DSPy-based information extraction pipelines
- Compliance and risk teams building interconnected entity relationship maps from unstructured data
Best For
- Knowledge graph construction and semantic search over entities and relationships
- Retrieval-augmented generation (RAG) where graph structure improves answer quality
- Entity and relation extraction from text with automatic deduplication
- Graph algorithms (PageRank, centrality analysis) on your stored data
- Interactive exploration and visualization of domain knowledge networks
You might also like
You can input client briefs, building dimensions, departmental requirements, and adjacency needs to receive a comprehensive space program that quantifies square footage by function, identifies spatial conflicts, calculates efficiency metrics, and prioritizes areas by business criticality. This becomes your design contract—a documented baseline that prevents scope creep, justifies design decisions, and enables accurate cost estimation throughout the project lifecycle.
by bradagi
You can analyze any NoSQL query—whether MongoDB, DynamoDB, Cassandra, or Firestore—to identify performance bottlenecks, missing indexes, and inefficient data access patterns. You'll receive structured recommendations with before/after comparisons, cost estimates, and refactored queries that dramatically improve performance and reduce database costs.
You can create systematic healthcare space plans that simultaneously satisfy building codes, infection prevention requirements, clinical workflow patterns, and evidence-based therapeutic design. Claude helps you audit existing layouts against regulatory checklists, generate multiple compliant variations for stakeholder review, and identify design conflicts before specifications are finalized. You'll coordinate infection control zones, accessibility pathways, MEP infrastructure, and staff workflow patterns to deliver functional, code-compliant healing environments.
by bradagi
You can submit architectural design briefs—regardless of project type or complexity—and receive a comprehensive validation that identifies gaps in technical specifications, flags compliance and regulatory risks, surfaces conflicting requirements, and generates clarifying questions for stakeholders. This catches issues when they're cheapest to fix, preventing costly design iterations and scope creep during development phases.
by bradagi
Get expert guidance on designing efficient NoSQL schemas tailored to your access patterns and scaling requirements. You'll analyze query bottlenecks, receive specific optimization strategies with concrete examples, and understand trade-offs between consistency, availability, and performance across MongoDB, DynamoDB, and Cassandra.
You can rapidly develop comprehensive space programming documents that quantify spatial requirements, define operational workflows, and establish design criteria before schematic design begins. This skill translates your hospitality venue's business objectives and guest experience vision into precise area schedules with net/gross calculations, adjacency matrices showing spatial relationships, and guest flow diagrams that guide all subsequent design decisions and ensure efficient space allocation.
by bradagiYou'll learn to architect Claude-powered automation systems that handle errors gracefully, integrate seamlessly with your stack, and optimize costs at scale. Build workflows that run 24/7 with minimal supervision, from proof-of-concept to production deployment. This skill covers prompt engineering, async patterns, error recovery, monitoring, and cost optimization so your automation stays reliable and affordable as it grows.
by bradagi
You can generate multiple furniture arrangement options for any residential space by inputting room dimensions, permanent fixtures, and design requirements. The skill produces annotated layout recommendations with specific furniture placement reasoning, traffic pattern analysis, functional zone definitions, and design narratives tailored for client presentations. It handles challenging scenarios like unusual room shapes, multiple entry points, and small-space optimization to maximize both functionality and aesthetic appeal.
by bradagi