Intelligence Engine
Domain-agnostic intelligence engine with schema-driven knowledge graphs.
Build knowledge graphs from any structured domain using YAML schema definitions, then search, explore, and analyze them through hybrid search, a web UI, or an MCP server for AI assistants.
| Key Stats: 1261+ tests | 15 MCP tools | 33 REST endpoints | 8 languages | 2 domains |
Screenshots
 Knowledge Graph — 5000 nodes, 7048 edges |
 Cypher Console — 50 rows in 9ms |
 Dashboard Timeline — 3 indexing runs |
 Dashboard Health — dead code, cycles, hubs |
 Cross-Project Compare |
 AI Memory — KuzuDB + Minna |
Architecture
Domain Schema Extractors Knowledge Graph Hybrid Search MCP / Web UI
(YAML) | | | |
code.yaml +--> Tree-sitter --+--> KuzuDB +--> BM25 (0.35) +--> 15 MCP tools
archaeology | (8 languages) | Entity_code | Semantic | (FastMCP)
your-domain | | Entity_archae.. | (0.40) |
... +--> Custom YAML --+ Rel_code_* | Graph (0.25) +--> 33 REST
| extractors | Rel_archae.._* | | endpoints
| | | 3-way RRF | (FastAPI)
+-------------------+ NetworkX | fusion |
(fallback) +------------------+--> React 18
| Sigma.js
LanceDB | graph explorer
(384-dim vectors) +----
Each domain is defined by a single YAML file that specifies entity types, relationship types, entity properties, search profiles, display configuration, and health checks. The engine creates domain-scoped database tables (Entity_<domain>, Rel_<domain>_*) and routes all operations through the (project, domain) key.
Feature Highlights
- Domain Generalization – Define any knowledge domain via a YAML schema. The engine handles table creation, indexing, search, and visualization automatically.
- Archaeology MVP – First non-code domain: finds, sites, periods, materials, water bodies. Validates the entire domain-agnostic architecture.
- Hybrid Search – 3-way Reciprocal Rank Fusion combining BM25 keyword search, semantic vector search (all-MiniLM-L6-v2), and graph context expansion.
- Graph Visualization – React 18 + Sigma.js 3 force-directed graph explorer with filtering, clustering, and interactive entity detail.
- AI Integration – LLM-powered entity summaries and Q&A (Claude, OpenAI, Gemini, Ollama). Persistent AI data survives re-indexing.
- MCP Server – 15 tools exposing the full engine to AI coding assistants via the Model Context Protocol.
- Multi-Project – Index and search across 100+ projects. Shared database mode enables cross-project Cypher queries and global analysis.
Domains
Code Intelligence (built-in)
Parses 8 languages via Tree-sitter:
| Language | Entity Types |
|---|---|
| Python | function, class, method, module, variable |
| JavaScript | function, class, method, module, variable |
| TypeScript/TSX | function, class, method, module, variable, interface |
| Java | function, class, method, module, interface |
| Go | function, class, method, module, interface |
| HTML | component, template, form, section |
| CSS | selector, css_variable, keyframe, media_query |
Relationships: CALLS, IMPORTS, EXTENDS, DEFINES, METHOD_OF, LINKS_STYLESHEET, REFERENCES_SCRIPT, USES_VARIABLE
Archaeology (first non-code domain)
Custom YAML/JSON extractor for archaeological data:
| Entity Type | Category | Example |
|---|---|---|
| find | artifact | Bronze axe, pottery shard |
| site | location | Burial mound, settlement |
| period | temporal | Bronze Age, Iron Age |
| material | classification | Bronze, flint, ceramic |
| water_body | geography | Lake, river |
Relationships: FOUND_AT, DATED_TO, MADE_OF, NEAR_WATER, ASSOCIATED_WITH
Adding a New Domain
- Create
config/domains/your-domain.yamldefining entity types, relationships, properties, and search profiles - Write an extractor (or use Tree-sitter for code-like domains)
- Index your data – the engine creates the necessary database tables automatically
See docs/architecture.md for the full domain schema specification.
Status
All phases complete + Domain Generalization. v0.21.0, 1261+ tests passing.
| Phase | Feature | Status |
|---|---|---|
| 1 | AST Parsing (Tree-sitter) | Complete |
| 2 | Knowledge Graph (NetworkX + KuzuDB) | Complete |
| 3 | Hybrid Search (BM25 + Graph RRF) | Complete |
| 4 | MCP Server (15 tools via FastMCP) | Complete |
| 5 | Multi-Project (registry, cross-project search) | Complete |
| 6 | Semantic Embeddings (sentence-transformers + LanceDB) | Complete |
| 6.5 | KuzuDB Migration (dual-backend, Cypher queries) | Complete |
| 7 | Visual UI (React + Sigma.js graph explorer) | Complete |
| 8 | Multi-Language Support (Python, JS, TS/TSX, Java, Go, HTML, CSS) | Complete |
| 9 | Incremental Indexing (git diff + hash fallback) | Complete |
| 10 | Performance Dashboard (timeline, phases, health, compare) | Complete |
| 11 | AI-Powered Summaries (Claude, OpenAI, Gemini, Ollama) | Complete |
| 12 | Graph Clustering (Louvain community detection) | Complete |
| – | Batch Summaries, Code Quality Metrics, UI Polish | Complete |
| – | AI Q&A (free-form questions, template prompts, history) | Complete |
| – | Shared DB Architecture (multi-tenant, migration tool) | Complete |
| – | Cross-Project Search + Global Graph Analysis | Complete |
| – | AI Overlay Preservation (data survives re-index) | Complete |
| – | Read-Only Serving + Input Sanitization | Complete |
| – | AI Memory Tab (unified memory browser, export) | Complete |
| – | HTML & CSS Language Support (8 new entity types) | Complete |
| – | Domain Generalization (schema-driven, YAML config) | Complete |
| – | Archaeology MVP (first non-code domain) | Complete |
Quick Start
# Clone and set up
git clone <repo-url> intelligence-engine
cd intelligence-engine
python3 -m venv .venv
source .venv/bin/activate
uv pip install -e ".[dev]"
# Index a project (code domain, auto-detected)
python -m intelligence_engine index /path/to/your/project
# Search
python -m intelligence_engine search data/myproject/parse.json "find authentication"
# Start the web UI
python -m intelligence_engine serve --port 8420 &
cd src/intelligence_engine/web/frontend && npm install && npm run dev
# Open the URL shown by Vite in your browser
Storage Modes
# Per-project (default): each project gets its own databases
python -m intelligence_engine index /path/to/project
# Shared mode: all projects in a single database (enables cross-project queries)
# Set storage.mode: shared in config/config.yaml, then:
python -m intelligence_engine migrate --to-shared
python -m intelligence_engine migrate --verify-only
Web UI
React 18 + Sigma.js 3 graph explorer with 6 dashboard tabs:
- Interactive graph – Force-directed visualization (ForceAtlas2) with node/edge type filtering
- Hybrid search – BM25, semantic, graph, or combined search with dropdown mode picker
- Entity detail – Source code preview, complexity badges, AI summaries, Q&A history
- Cypher console – Direct KuzuDB queries with sticky headers and monospace output
- Performance dashboard – Index timeline, per-phase timing, health snapshots, project comparison, quality metrics
- AI Memory browser – Unified view of all AI-generated data with filters, search, and CSV/JSON export
- Graph clustering – Louvain community detection, color by type or cluster
- Code quality – Composite scores, complexity histograms, coupling analysis
- AI Q&A – Free-form questions about any entity, with template prompts and persistent history
- LLM settings – Multi-provider credential management (Claude, OpenAI, Gemini, Ollama)
- Batch operations – Project indexing, batch summarization with progress tracking
MCP Server
15 tools for AI assistants via the Model Context Protocol:
| Tool | Purpose |
|---|---|
ie_index |
Index a project into the knowledge graph |
ie_query |
Search within a single project |
ie_search_all |
Cross-project semantic search |
ie_context |
Entity context (callers, callees, blast radius) |
ie_detect_changes |
Pre-change risk assessment via git diff |
ie_cypher |
Read-only Cypher queries on KuzuDB |
ie_wiki |
Generate documentation from the graph |
ie_status |
List all indexed projects |
ie_health |
Structural health (dead code, cycles, hubs) |
ie_quality |
Code quality metrics (complexity, docs, coupling) |
ie_summarize |
AI-powered entity summary (single) |
ie_batch_summarize |
AI-powered summaries (project-wide) |
ie_global_analysis |
Cross-project clustering + health (shared mode) |
ie_memory |
Unified AI memory browser |
The server includes self-describing resources (ie://schema, ie://cypher-templates, ie://guide) and pre-built prompt workflows (code review, capability audit, change risk assessment).
Technology Stack
| Component | Technology |
|---|---|
| Language | Python 3.12 (backend), TypeScript (frontend) |
| AST Parser | Tree-sitter (8 language grammars) |
| Graph DB | KuzuDB (default) + NetworkX (fallback) |
| Vector Store | LanceDB (all-MiniLM-L6-v2, 384-dim) |
| Keyword Search | BM25 (rank_bm25) |
| MCP Server | FastMCP |
| Web Backend | FastAPI (33 REST endpoints) |
| Web Frontend | React 18 + Sigma.js 3 + Vite 6 + Tailwind CSS v4 |
| AI Providers | Claude, OpenAI, Gemini, Ollama |
| Build | Hatchling (Python), Vite (frontend) |
| Package Mgmt | uv (Python), npm (frontend) |
| Testing | pytest (1261+ tests) |
Documentation
- Architecture Deep-Dive – Domain schema system, pipeline, storage modes
- Technical Overview – Parser, graph, search, AI, MCP, REST details
- Architectural Decisions – Key design choices and rationale
- Project Context – What it does, why it exists, status
Interactive Demos
- Which Search Strategy? – Choose the right search approach for your query