Graph-Loom

A lightweight, local-first graph notebook and visualizer built with Rust and egui. Features physics-assisted layout, OpenCypher query support, embedded API servers, and enterprise-grade analytics.

Key Features

• Physics-Assisted Layout: Auto-clustering with smooth interaction for hundreds of nodes
• OpenCypher Query Console: Full support for MATCH, CREATE, MERGE, SET, DELETE, WHERE, RETURN, WITH, UNWIND
• Graph Algorithms: PageRank, Betweenness Centrality, Shortest Path, A* Pathfinding
• Temporal Queries: Time-travel visualization with timestamp-based filtering
• Local Embeddings: TF-IDF based semantic similarity search (no external API required)
• LLM Integration: Optional OpenAI/Anthropic/Ollama support for entity extraction
• Embedded APIs: HTTP, WebSocket, and gRPC interfaces for automation
• SQLite Storage: ACID-compliant persistence with FTS5 full-text search
• Local-First: All data stored locally with automatic backups

---

Installation

Prerequisites

• Rust toolchain (stable, edition 2024)

Build from Source

    # Clone the repository
    git clone https://github.com/jackpots28/Graph-Loom.git
    cd Graph-Loom
    
    # Build and run (release mode recommended)
    cargo run --release

Build with CLI Shell

    # Build both GUI and CLI tools
    cargo build --release --features cli --bin glsh --bin Graph-Loom

---

Quick Start

Creating Nodes and Relationships

Using the Query Console (sidebar → Query tab):

    -- Create nodes
    CREATE (p:Person {name: 'Alice', role: 'Engineer'})
    CREATE (c:Company {name: 'TechCorp', industry: 'Software'})
    
    -- Create relationship
    CREATE (p)-[:WORKS_AT {since: '2020'}]->(c)

Or use the sidebar tools for point-and-click creation.

Querying Data

    -- Find all people
    MATCH (p:Person) RETURN p
    
    -- Find relationships
    MATCH (p:Person)-[r:WORKS_AT]->(c:Company)
    WHERE p.name = 'Alice'
    RETURN p, r, c
    
    -- Pattern matching with filters
    MATCH (n) WHERE n.name CONTAINS 'Tech' RETURN n

---

Query Language Reference

OpenCypher Support

Clause	Description	Example
MATCH	Find patterns	MATCH (n:Person) RETURN n
CREATE	Create nodes/relationships	CREATE (n:Label {key: 'value'})
MERGE	Create if not exists	MERGE (n:Person {name: 'Bob'})
WHERE	Filter results	WHERE n.age > 30
RETURN	Output results	RETURN n.name, n.age
SET	Update properties	SET n.status = 'active'
REMOVE	Delete properties	REMOVE n.temp_field
DELETE	Delete nodes/relationships	DELETE n
DETACH DELETE	Delete with relationships	DETACH DELETE n
WITH	Chain queries	WITH n ORDER BY n.name
UNWIND	Expand lists	UNWIND [1,2,3] AS x
ORDER BY	Sort results	ORDER BY n.name DESC
SKIP/LIMIT	Pagination	SKIP 10 LIMIT 5
DISTINCT	Unique results	RETURN DISTINCT n.label

WHERE Operators

    -- Comparison
    WHERE n.age > 30 AND n.status = 'active'
    WHERE n.name <> 'Unknown'
    
    -- String operations
    WHERE n.name CONTAINS 'Smith'
    WHERE n.name STARTS WITH 'A'
    WHERE n.name ENDS WITH 'son'
    
    -- Null checks
    WHERE n.email IS NOT NULL
    WHERE n.phone IS NULL
    
    -- List membership
    WHERE n.role IN ['Admin', 'Manager']
    
    -- Logical
    WHERE n.active = true OR n.role = 'Admin'
    WHERE NOT n.archived

---

Graph Algorithms

Access via CALL procedures in the Query Console:

PageRank (Node Importance)

    CALL algo.pageRank()
    -- Returns nodes sorted by importance score

Betweenness Centrality (Bridge Detection)

    CALL algo.betweenness()
    -- Identifies nodes that connect different graph regions

Shortest Path

    CALL algo.shortestPath("node-uuid-1", "node-uuid-2")
    -- Returns path between two nodes

A* Pathfinding

    CALL algo.astar("from-uuid", "to-uuid")
    -- Heuristic-based pathfinding using node positions

All Paths

    CALL algo.allPaths("from-uuid", "to-uuid", 5)
    -- Find all paths up to depth 5

Levenshtein Distance (Fuzzy String Matching)

    CALL string.levenshtein("search text", 2)
    -- Find nodes within edit distance 2 (default)
    
    CALL string.levenshtein("search text", 3, "name")
    -- Search specific field with custom max distance

Returns _levenshtein_distance and _matched_field in metadata.

---

Temporal Queries

Nodes and relationships include created_at and updated_at timestamps.

Timeline

    CALL temporal.timeline()
    -- Returns chronological list of all graph events

Timestamp Range

    CALL temporal.range()
    -- Returns min/max timestamps in the graph

Nodes in Time Range

    CALL temporal.nodesInRange(1704067200000, 1706745600000)
    -- Find nodes created within Unix timestamp range (milliseconds)

Graph State at Time

    CALL temporal.atTime(1705000000000)
    -- Returns graph state as it existed at that timestamp

---

Semantic Search & Embeddings

Local Embeddings (No API Required)

Graph-Loom includes local embeddings for semantic similarity with three options:

• ONNX (default): Best quality using all-MiniLM-L6-v2 transformer model. Downloads ~90MB on first use.
• TF-IDF: Fast, lightweight term-frequency based vectors, no downloads required
• Word2Vec: Learns semantic word relationships from your graph data (pure Rust implementation)

Configure in Preferences → Embeddings to switch between models. On startup, if the node_embeddings table is empty, all nodes are automatically re-embedded using the selected model.

ONNX Examples (Default - Best Quality)

    -- High-quality semantic search using transformer embeddings
    CALL embedding.similar("machine learning engineer", 10)
    
    -- Understands synonyms and related concepts
    CALL embedding.threshold("software development", 0.7)

TF-IDF Examples

    -- Find nodes with similar keywords
    CALL embedding.similar("database engineer", 10)
    
    -- Good for exact term matching
    CALL embedding.threshold("python developer", 0.3)

Word2Vec Examples

    -- Find semantically related nodes (understands word relationships)
    CALL embedding.similar("machine learning", 10)
    -- May return nodes about "AI", "neural networks", "deep learning"
    
    -- Find neighbors of a node by semantic similarity
    CALL embedding.neighbors("node-uuid", 5)
    
    -- Discover related concepts above similarity threshold
    CALL embedding.threshold("data science", 0.5)

Word2Vec learns that words appearing in similar contexts are related, enabling discovery of semantic relationships between nodes even when they don't share exact terms.

Fast Approximate Search (HNSW)

    CALL embedding.ann("search text", 10)
    -- O(log n) approximate nearest neighbor search using HNSW index

Faster than exact search for large graphs. Results include _search_type: hnsw_ann in metadata.

Re-Embedding

    CALL embedding.reembed()
    -- Clears and rebuilds all embeddings using the selected model

Use this after changing the embedding model in Preferences, or click "Re-Embed All Nodes" in Preferences → Embeddings.

Results include _similarity (cosine) and _distance (L2) in metadata.

Full-Text Search

    CALL db.search("search query")
    -- Searches node labels and metadata

Graph Schema

    CALL db.schema()
    -- Returns labels, relationship types, and query suggestions

LLM-Powered Entity Extraction

Configure in Preferences → LLM Settings:

    CALL semantic.extract("John works at Acme Corp as a senior developer")
    -- Extracts entities using configured LLM or heuristic fallback

Supported providers: OpenAI, Anthropic, Ollama (local)

---

API & Integration

Enable APIs

Go to Settings → Preferences → API Settings:

• HTTP/WebSocket: Default 127.0.0.1:8787
• gRPC: Default port 50051
• API Key: Optional authentication

HTTP API

    curl -X POST http://127.0.0.1:8787/api/query \
      -H "Content-Type: application/json" \
      -H "X-API-Key: your-key" \
      -d '{"query": "MATCH (n) RETURN n LIMIT 10"}'

Health Check

    curl http://127.0.0.1:8787/health
    # Returns: {"status":"healthy","components":{...},"version":"1.15.1"}

Prometheus Metrics

    curl http://127.0.0.1:8787/metrics
    # Returns Prometheus-compatible metrics:
    # graph_loom_queries_total{status="success"} 42
    # graph_loom_queries_total{status="error"} 3
    # graph_loom_query_duration_seconds_sum 12.5
    # graph_loom_api_up 1
    # graph_loom_nodes_total 150
    # graph_loom_relationships_total 200

WebSocket REPL

Connect to ws://127.0.0.1:8787/ws for interactive queries.

gRPC

High-performance interface using Protocol Buffers. See proto/graph_loom.proto for schema.

Available RPCs:

• Execute(QueryRequest) -> QueryResponse - Standard request/response
• ExecuteStream(QueryRequest) -> stream QueryRow - Streaming for large result sets

CLI Shell (glsh)

    # Build CLI
    cargo build --features cli --bin glsh
    
    # Connect to running instance
    ./target/debug/glsh --host 127.0.0.1 --port 8787

The CLI provides an interactive REPL for executing queries against a running Graph-Loom instance:

Python Client

Example client in examples/python_client/. See Python Client README.

---

Headless Mode

Run as a pure API server without GUI:

    ./target/release/Graph-Loom --background --api-enable

Command Line Options

Flag	Description
--background, -b	Run without GUI (background/headless mode)
--api-enable	Enable HTTP/WebSocket API
--api-bind <addr>	Bind address (default: 127.0.0.1)
--api-port <port>	HTTP port (default: 8787)
--api-key <key>	Set API authentication key
--grpc-enable	Enable gRPC server
--grpc-port <port>	gRPC port (default: 50051)

---

User Interface

Canvas Navigation

• Pan: Drag background
• Zoom: Scroll wheel (cursor over canvas)
• Select: Click node/relationship
• Multi-select: Rectangle drag or Shift+click

Sidebar Modes

• Tooling: Node/relationship creation, layout controls
• Query: OpenCypher console with autocomplete
• Notes: Markdown notes linked to graph nodes

Node Details

Click on any node to view and edit its details, metadata, and relationships:

Note Editor

Create and manage markdown notes that can be linked to graph nodes:

Keyboard Shortcuts

• Ctrl/Cmd+S: Save
• Ctrl/Cmd+Space: Force autocomplete popup
• Tab/Enter: Accept autocomplete suggestion
• Escape: Dismiss autocomplete/dialogs
• Arrow keys: Navigate autocomplete suggestions

Query Autocomplete

The query console provides IDE-style autocomplete:

• Type to filter suggestions
• Arrow keys to navigate
• Tab/Enter to accept
• Trailing space hides suggestions
• Ctrl/Cmd+Space shows all options

---

Data Storage

SQLite Backend (Default)

• Location: OS-specific app data directory
• File: state.db
• Features: ACID compliance, FTS5 full-text search, R-tree spatial indexing

Performance Optimizations

• WAL Mode: Write-Ahead Logging for 10-100x faster writes and concurrent reads
• Query Plan Caching: LRU cache (1000 plans) reduces parse overhead for repeated queries
• Property Indexes: B-tree indexes on metadata for fast property lookups
• Incremental Embeddings: Add/remove nodes without full index rebuild
• Batch Operations: Optimized bulk imports (1000-10000 items per transaction)

Versioned Backups

Automatic timestamped backups in RON format for human readability.

Settings

Stored in settings.json:

• macOS: ~/Library/Application Support/Graph-Loom/
• Windows: %APPDATA%\Graph-Loom\
• Linux: ~/.config/Graph-Loom/

---

System Tray & Background Mode

When API/gRPC is enabled:

• Close to Tray: Window hides, service continues running
• Tray Menu: Show window or quit application
• Multi-Instance Detection: Second launch brings existing instance to foreground
• CPU Efficiency: Near-zero CPU when backgrounded

---

Contributing

Development Setup

    git clone https://github.com/jackpots28/Graph-Loom.git
    cd Graph-Loom
    cargo build

Running Tests

    cargo test
    cargo test algorithms
    cargo test temporal
    cargo test embeddings

Code Style

• Follow existing patterns in the codebase
• Use cargo fmt before committing
• Run cargo clippy for lints

Pull Requests

1. Fork the repository
2. Create a feature branch
3. Make changes with tests
4. Submit PR with clear description

---

Project Structure

    src/
    ├── api/                    # HTTP, WebSocket, gRPC servers
    │   ├── auth.rs             # Authentication & RBAC
    │   ├── grpc.rs             # gRPC service
    │   └── server.rs           # HTTP/WebSocket server
    ├── gql/                    # Query language
    │   ├── cypher_spec.rs      # OpenCypher parser
    │   └── query_interface.rs  # Query execution
    ├── graph_utils/            # Core graph operations
    │   ├── algorithms.rs       # PageRank, centrality, pathfinding
    │   ├── graph.rs            # Node/Relationship structures
    │   └── temporal.rs         # Time-travel queries
    ├── gui/                    # User interface
    │   └── frontend.rs         # egui application
    ├── persistence/            # Data storage
    │   ├── persist.rs          # Save/load logic
    │   ├── settings.rs         # App configuration
    │   └── sqlite_backend.rs   # SQLite storage
    ├── semantic/               # AI/ML features
    │   ├── embeddings.rs       # Local TF-IDF embeddings
    │   ├── extraction.rs       # Entity extraction
    │   ├── llm_client.rs       # LLM API client
    │   └── rag.rs              # Graph RAG
    └── main.rs                 # Application entry point

---

License

Apache 2.0

---

Acknowledgments

Built with:

• egui - GUI
• rusqlite - SQLite bindings
• tonic - gRPC framework
• actix-web - HTTP server