0.3.8

Full Changelog: 0.3.7...0.3.8

0.3.7

Bump version to 0.3.7

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

0.3.6

What's New

Separate model configuration for Smart Search

You can now configure a different model for Smart Search queries vs. extraction:

• Use case: Use fast/cheap models (GPT-4o-mini) for KG extraction, and better models (GPT-4o, Claude Sonnet) for search
• New settings:
  • "Use separate model for smart search" toggle
  • Smart Search provider selection
  • Smart Search model dropdown with custom input

How to use

1. Go to Settings → Simple Graph Builder → Smart search model
2. Enable "Use separate model for smart search"
3. Select your preferred provider and model for search queries

---

Full Changelog: 0.3.5...0.3.6

v0.3.4

Changes

Obsidian Review Bot Fixes

• Sentence case fixes for UI text (16 changes)
• Remove deprecated RelationshipType exports
• Change "General" heading to "Provider" in settings
• Fix confirm-modal callback type to allow async
• Use proper vault.delete(file, true) for trash behavior

Full Changelog

0.3.3...0.3.4

0.3.3

Changes in v0.3.3

Obsidian Review Bot Fixes

• Use sentence case for all UI text (commands, settings headings, button labels)
• Remove unnecessary type assertions where type guards already narrow types
• Add proper type checks before string conversion on JSON-parsed values
• Fix type union redundancy in function signatures
• Add eslint-disable comments for intentional deprecated usage (migration code)
• Change console.log to console.debug for migration message
• Handle promises properly with await/void operators
• Use CSS classes instead of inline styles for graph tooltips

Graph Visualization Improvements

• Add "Minimum connections" setting to filter nodes with few connections from the graph view
• Reduce edge color intensity (#cbd5e1, opacity 0.4) for better node label readability

Code Quality

• Extract helper functions to reduce code duplication
• Move constants to module level for better performance
• Simplify functions with early returns and object lookups

0.3.2

Add v3 data model with EntityType and relationship fields

- Add EntityType union type with 10 fixed entity types
- Add entityType field to OntologyNode (replacing flexible labels)
- Add relationship field to OntologyEdge (free-form verbs)
- Add helper functions: getEntityTypeColor, getNodeEntityType, getEdgeRelationship
- Add labelToEntityType for legacy v2 data migration
- Update prompts for v3 extraction format
- Update search to use new type helpers

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

v0.3.1 - Hybrid Entity Resolution

What's New

Hybrid Entity Resolution (Opt-in)

A multi-stage entity deduplication pipeline inspired by recent knowledge graph research (KGGen, NeurIPS 2025). Ensures "AI", "artificial intelligence", and "Artificial Intelligence" merge into a single node.

Resolution Pipeline:

Stage	Method	Speed
1. Persistent cache	Previously resolved tokens	O(1)
2. Session cache	Same name resolved this session	O(1)
3. Exact name	Hash lookup on canonical name	O(1)
4. Alias match	Hash lookup on stored aliases	O(1)
5. Embedding similarity	Cosine similarity > 0.90 = auto-merge	O(n)
6. LLM verification	Ambiguous matches (0.80-0.90) verified by LLM	API call
7. Create new	No match found	-

New Features

• Embedding-based resolution: Support for OpenAI, Gemini, and Ollama embedding providers
• Binary embedding storage: Efficient embeddings.bin format (~6KB per node)
• Manual entity merge: Right-click nodes in graph view to merge duplicates
• Entity Resolution settings: Configure thresholds, providers, and cache management

Settings

• Enable/disable embedding-based resolution (default: off to avoid API costs)
• Choose embedding provider (can differ from main LLM provider)
• Configure similarity thresholds (high: 0.90, low: 0.80)
• Enable/disable LLM verification for ambiguous matches
• Compute embeddings for existing nodes
• Clear resolution cache

Documentation

• Updated README with entity resolution docs
• Added research references (LightRAG, Microsoft GraphRAG, KGGen, Neo4j GraphRAG, pgvector)

Installation

Using BRAT (Recommended)

1. Install BRAT from Community Plugins
2. Open command palette → "BRAT: Add a beta plugin"
3. Enter: junhewk/simple-graph-builder

Manual

Download main.js, styles.css, and manifest.json below and place in .obsidian/plugins/simple-graph-builder/

0.3.0

Update with lightweight ontology construction (see README.md)

0.2.0

Obsidian plugin simple-graph-builder 0.2.0 for community release