semantica is an open-source developer library that provides a semantic/context layer under LLMs and agent frameworks, turning enterprise data into a governed Context Graph and knowledge graph. It combines multi-source ingestion and graph construction with deterministic reasoning engines and decision provenance.
Project overview
Its reasoning engines, knowledge graph construction, and provenance are fully deterministic and require no LLM, addressing the common problem that embedding-based agent decisions cannot be traced or justified in regulated domains.
Project type
AI Agent · RAG · Evaluation & Observability
Use cases
Knowledge Q&A · Search & Research · Automation
Deployment
Refer to project documentation
License
MIT
Best for
AI engineers and agent-framework developers who need structured, explainable context and decision provenance rather than embedding-only retrieval.
Data teams working with enterprise sources who want to build a governed knowledge graph for LLM and agent applications.
Key capabilities
Multi-source ingestion, parsing, entity-aware chunking, NER/relation/event extraction, conflict detection and deduplication that builds a knowledge graph from enterprise data.
Forward chaining, Rete, Datalog and SPARQL inference with explainable paths, requiring no LLM.
Limitations and risks
Python 3.11.5–3.13 is required; Python 3.10 installs only the last 0.7.x release with no further updates.
The library is used through a Python library interface and a CLI, so coding is required.
Getting started
Install with pip (pip install semantica), instantiate ContextGraph(advanced_analytics=True) in Python, and run semantica doctor to verify the installation. Setup is easy: a single pip install of a lightweight core with optional extras.
Evidence and sources
README: ```bash pip install semantica ```
README: Most AI agents run on embeddings, not meaning: similarity scores with no structure, no relationships, and no way to explain why a result came back.
README: Semantica is the semantic/context layer underneath your LLM, vector store, and agent framework
README: **Knowledge Pipeline:** Multi-source ingestion, entity-aware chunking, NER/relation/event extraction, and graph construction, with semantic dedup and provenance-preserving merges…
README: In Semantica, a decision is not a log line. It is a first-class graph node with a full lifecycle. In regulated domains, every AI decision must be traceable to a source and defensi…