ontocast 0.4.3

Agentic ontology and knowledge graph co-generation

OntoCast

Agentic ontology-assisted framework for semantic triple extraction

---

Overview

OntoCast is a framework for extracting semantic triples (creating a Knowledge Graph) from documents using an agentic, ontology-driven approach. It combines ontology management, natural language processing, and knowledge graph serialization to turn unstructured text into structured, queryable data.

---

Key Features

• Ontology-Guided Extraction: Ensures semantic consistency and co-evolves ontologies
• Entity Disambiguation: Resolves references across document chunks (embedding + symbolic alignment)
• Parallel Processing: Per-unit ontology and facts loops with configurable worker concurrency
• Multi-Format Support: Handles text, JSON, PDF, and Markdown
• Semantic Chunking: Splits text based on semantic similarity
• RDF Output: Produces standardized RDF/Turtle (optional JSON-LD LLM wire format)
• RDF 1.2 Provenance: Quoted-triple / reification support; optional strip_provenance on API output
• Triple Store Integration: Apache Fuseki (production) or in-memory pyoxigraph (default)
• Ontology Context Modes: Catalog selection, Qdrant vector retrieval, or fixed catalog ontology
• Tenancy: Partition Fuseki datasets and Qdrant collections by tenant and project
• REST API: Document processing, ontology catalog management, and graph-matching endpoints
• Hierarchical Configuration: Type-safe configuration system with environment variable support
• Automatic LLM Caching: Built-in response caching for improved performance and cost reduction
• GraphUpdate Operations: Token-efficient SPARQL-based updates instead of full graph regeneration
• Budget Tracking: Comprehensive tracking of LLM usage and triple generation metrics
• Ontology Versioning: Automatic semantic versioning with hash-based lineage tracking

---

Applications

OntoCast can be used for:

• Knowledge Graph Construction: Build domain-specific or general-purpose knowledge graphs from documents
• Semantic Search: Power search and retrieval with structured triples
• GraphRAG: Enable retrieval-augmented generation over knowledge graphs (e.g., with LLMs)
• Ontology Management: Automate ontology creation, validation, and refinement
• Data Integration: Unify data from diverse sources into a semantic graph

---

Installation

uv add ontocast[doc-processing] 
# or
pip install ontocast

Optional features: document processing (PDFs, PPT, OCR, semantic chunking), vector mode for ontology retrieval:

uv add "ontocast[doc-processing,lancedb]"
# or
pip install "ontocast[doc-processing,lancedb]"

---

Quick Start

1. Configuration

Copy the example file and edit as needed:

cp .env.example .env

Minimal settings:

# LLM Configuration
LLM_PROVIDER=openai
LLM_API_KEY=your-api-key-here
LLM_MODEL_NAME=gpt-4o-mini
LLM_TEMPERATURE=0.0

# Server Configuration
PORT=8999
BASE_RECURSION_LIMIT=1000
ESTIMATED_CHUNKS=30
MAX_VISITS=1
RENDER_MODE=ontology_and_facts
ONTOLOGY_MAX_TRIPLES=50000

# Path Configuration
ONTOCAST_WORKING_DIRECTORY=/path/to/working
ONTOCAST_ONTOLOGY_DIRECTORY=/path/to/ontologies
ONTOCAST_CACHE_DIR=/path/to/cache

# Optional: Triple Store Configuration (Fuseki for production persistence)
FUSEKI_URI=http://localhost:3030
FUSEKI_AUTH=admin/admin
# Datasets default to ontocast--test--facts / ontocast--test--ontologies when unset

# Optional: Web search grounding (search-later mode)
WEB_SEARCH_ENABLED=false
WEB_SEARCH_PROVIDER=duckduckgo
WEB_SEARCH_TOP_K=3

See Configuration Guide and .env.example for the full surface (embeddings, Qdrant, aggregation, etc.).

2. Start Server

# Backend automatically detected from .env configuration
ontocast --env-path .env

# Process a specific file (batch mode)
ontocast --env-path .env --input-path ./document.pdf

# Process only the first N chunks (for testing)
ontocast --env-path .env --head-chunks 5

Paths and triple-store credentials are configured via .env — not CLI overrides.

3. Process Documents

curl -X POST http://localhost:8999/process -F "file=@document.pdf"

JSON body example:

curl -X POST http://localhost:8999/process \
  -H "Content-Type: application/json" \
  -d '{"text": "Your document text here"}'

4. API Endpoints

The OntoCast server exposes a REST API. Common endpoints:

• POST /process — full document pipeline (JSON or multipart file upload)
• POST /process_unit — single-unit pipeline (useful for debugging prompts)
• POST /ontologies — upload catalog ontologies (Turtle)
• POST /flush — flush/clean triple store data
• GET /health, GET /info — health and service metadata
• POST /match/entities (and related match routes) — entity alignment and evaluation

curl -X POST http://localhost:8999/process -F "file=@document.pdf"

curl -X POST http://localhost:8999/flush

curl -X POST "http://localhost:8999/flush?dataset=my_dataset"

For Fuseki, optional tenant/project query parameters target a specific partition. Without them, the active server scope is flushed.

Full reference: API Endpoints.

---

Workflow

The extraction pipeline converts documents to Markdown, chunks them, runs parallel per-unit ontology and facts loops, then merges and serializes results.

Landscape layout: docs/assets/graph.lr.png. Per-unit loops: ontology_loop, facts_loop — Workflow guide.

Regenerate diagrams after graph changes: uv run plot-graph

---

LLM Caching

OntoCast includes automatic LLM response caching to improve performance and reduce API costs. Caching is enabled by default and requires no configuration.

Cache Locations

• Tests: .test_cache/llm/ in the current working directory
• Windows: %USERPROFILE%\AppData\Local\ontocast\llm\
• Unix/Linux: ~/.cache/ontocast/llm/ (or $XDG_CACHE_HOME/ontocast/llm/)

Benefits

• Faster Execution: Repeated queries return cached responses instantly
• Cost Reduction: Identical requests don't hit the LLM API
• Offline Capability: Tests can run without API access if responses are cached
• Transparent: No configuration required — works automatically

Details: LLM Caching

---

Configuration System

OntoCast uses a hierarchical configuration system built on Pydantic BaseSettings:

Environment Variables

Variable	Description	Default	Required
LLM_API_KEY	API key for LLM provider	-	Yes (OpenAI)
LLM_PROVIDER	LLM provider (openai, ollama)	openai	No
LLM_MODEL_NAME	Model name	gpt-4o-mini	No
LLM_TEMPERATURE	Temperature setting	0.0	No
ONTOCAST_WORKING_DIRECTORY	Working directory path	-	No
ONTOCAST_ONTOLOGY_DIRECTORY	Ontology seed files	-	No
PORT	Server port	8999	No
MAX_VISITS	Maximum render/critic visits per unit loop	1	No
BASE_RECURSION_LIMIT	Base recursion limit for workflow	1000	No
RENDER_MODE	ontology, facts, or ontology_and_facts	ontology_and_facts	No
ONTOLOGY_MAX_TRIPLES	Maximum triples in ontology graph	50000	No
ONTOCAST_CACHE_DIR	Custom cache directory	Platform default	No
WEB_SEARCH_ENABLED	Optional web grounding (search-later)	false	No

See Configuration Guide for chunking, Qdrant, embeddings, aggregation, and web-search variables.

Triple Store Configuration

# Fuseki (production persistence)
FUSEKI_URI=http://localhost:3030
FUSEKI_AUTH=admin/admin

When Fuseki is not configured, OntoCast uses an in-memory pyoxigraph backend automatically. See Triple Store Setup.

CLI Parameters

ontocast --env-path .env
ontocast --env-path .env --input-path ./document.pdf
ontocast --env-path .env --head-chunks 5
ontocast --env-path .env --tenant acme --project reports

---

Triple Store Setup

OntoCast uses a unified triple-store interface with two backends:

1. Apache Fuseki — persistent RDF with SPARQL (production)
2. In-Memory (pyoxigraph) — zero-config default when Fuseki is not configured

Quick Setup with Docker (Fuseki)

cd docker/fuseki
cp .env.example .env
docker compose --env-file .env fuseki up -d

See Triple Store Setup for detailed instructions.

---

Documentation

• Quick Start Guide — Get started quickly
• Workflow — Pipeline and diagrams
• Configuration System — Environment variables
• API Endpoints — REST reference
• Tenancy — Multi-tenant stores
• Ontology Context — Catalog vs vector retrieval
• Triple Store Setup — Fuseki and in-memory backends
• LLM Caching — Automatic response caching
• User Guide — Core concepts
• API Reference — Python API (MkDocs)

Build docs locally: uv run mkdocs build

---

Highlights

Ontology Management

• Automatic Versioning: Semantic version increment based on change analysis (MAJOR/MINOR/PATCH)
• Hash-Based Lineage: Git-style versioning with parent hashes for tracking ontology evolution
• Multiple Version Storage: Versions stored as separate named graphs in Fuseki triple stores
• Timestamp Tracking: updated_at field tracks when ontology was last modified

GraphUpdate System

• Token Efficiency: LLM outputs structured SPARQL operations (insert/delete) instead of full TTL graphs
• Incremental Updates: Only changes are generated, dramatically reducing token usage
• Structured Operations: TripleOp operations with explicit prefix declarations for precise updates

Budget Tracking

• LLM Statistics: Tracks API calls, characters sent/received for cost monitoring
• Triple Metrics: Tracks ontology and facts triples generated per operation
• Summary Reports: Budget summaries logged at end of processing

See CHANGELOG.md for release-by-release notes.

---

Examples

Basic Usage

from ontocast.config import Config
from ontocast.toolbox import ToolBox

config = Config()
tools = ToolBox(config)

# Process documents via tools / workflow graph

Server Usage

ontocast --env-path .env --input-path ./document.pdf --head-chunks 10

---

Running tests

Tests load settings from the process environment. To run the suite with variables from a project .env file:

bash -c 'set -a; source .env; set +a; uv run pytest test'

Integration tests (for example Qdrant) read optional variables such as QDRANT_URI and QDRANT_API_KEY from that environment; they skip automatically when the service is unreachable.

---

Contributing

We welcome contributions! Please see our Contributing Guide for details.

---

License

This project is licensed under the Apache License 2.0 — see the LICENSE file for details.

---

Support

• Documentation: docs/
• Issues: GitHub Issues
• Discussions: GitHub Discussions