Ontoforge

The ontology mesh for GraphRAG. Mine ontologies from your logs, validate with SHACL, ship a hybrid retriever — without hand-crafting a single Turtle file.

Get started Install from PyPI

---

The guided wizard (Twilight theme) — Define → Enrich → Build → After. Also ships Mono and Light themes.

---

A guided tour

From landing page to ontology-grounded reasoning search, in order.

1. Landing — the marketing entry point.

2. Model your domain — the Twilight wizard walks you from a domain to a generated ontology in four phases (Define → Enrich → Build → After).

3. Run report — every run emits a branded HTML report: competency-question results, governance scorecard, semantic-loss analysis, and orphan-class checks.

4. Ask — pose a question — the read-only, ontology-grounded Ask console. Pick a flavor and provider (Ollama / OpenAI), set your sensitivity-tier ceiling, and optionally build an RDF subgraph for live SPARQL.

5. Ask — a cited, reasoned answer — the answer cites its sources, shows the read-only SQL that ran, materializes the result subgraph, and streams the full reasoning trace (plan → execute → relations → reason → synthesize).

6. Ask — live SPARQL over the result subgraph — interrogate exactly what grounded the answer with a SPARQL SELECT over the materialized RDF triples.

Reasoning Search reference →

---

What ships in the box

8 pipeline phases	Domain → Entities → Events → Relationships → SHACL → Generation → Evolution → Compliance
8 enrichment phases	Log mining L4–L13 (template clustering, HMM regimes, Granger causality, PMI co-occurrence, switching SSM, …)
10 starter industries	Telecom, healthcare, finance, manufacturing, retail, energy, government, insurance, logistics, pharma
6 database backends	SQLite (default), Postgres, MySQL, MSSQL, Oracle, DB2
5 LLM adapters	Anthropic, OpenAI, Vertex AI, Ollama, OCI
2 vector backends	FAISS, Chroma (scaffolded: Qdrant, pgvector)
5 deploy targets	Docker · Compose · Fly.io · Render · Cloud Run

Two-line install

DockerPyPI

docker run --rm -p 5051:5051 ghcr.io/synaptixs/ontomesh:latest
open http://localhost:5051

pip install 'ontoforge[wizard]'
ontoforge-wizard

Trust signals

• Apache-2.0 — full source, no proprietary core.
• Cosign-signed images — keyless OIDC, verifiable trust chain to synaptixs/ontomesh.
• SBOM + SLSA provenance — every release ships SPDX SBOM and SLSA attestation as OCI artefacts.
• 297+ tests running on every PR.
• Public roadmap + changelog — no surprises.

Pick a path

•   Just want to run it?

  ---

  A single docker run brings the wizard up locally. No accounts, no credentials.

  → Quickstart

•   Want to understand it?

  ---

  Read how the pipeline phases compose, what SHACL gives you, and why we mine logs.

  → Concepts

•   Deploying to production?

  ---

  Pick a managed runtime, wire up Postgres + Redis, harden the image.

  → Deployment

•   Integrating it?

  ---

  SDK, REST endpoints, MCP tools, JSON-LD artifacts — all the surfaces.

  → Reference

What's new in 3.8.0

• Reasoning Search — ask your database in plain English and get a cited, ontology-grounded answer with a live SPARQL subgraph. See Reasoning Search. (Flag-gated: ONTOFORGE_SEARCH=1.)
• First PyPI release under the ontoforge name (pip install ontoforge); GHCR image path unchanged.
• Prometheus ontomesh_search_* metrics + a bundled Grafana dashboard (deploy/monitoring/).

What's new in 3.7.0

• Apache-2.0 licence + NOTICE + SECURITY.md + Contributor Covenant CoC
• Public image at ghcr.io/synaptixs/ontomesh:3.7.0, cosign-signed, SBOM'd, SLSA-attested
• gunicorn (gthread) replaces the Flask dev server; multi-worker safe via Redis SSE bus
• Differentiated /live + /ready probes; structured JSON logs with X-Request-Id
• Prometheus /metrics endpoint with cardinality-bounded labels

→ Full changelog