Semantica · 2026-08-20 · major
Semantica v0.6.6 — security release hardens ingestion, adds CrewAI support
Semantica v0.6.6 fixes a privately disclosed batch of vulnerabilities across backup, export, outbound requests and triplet-store backends, and hardens ingestion against SSRF. The same release adds CrewAI integration and graph retraction.
A strongly recommended security upgrade for the MIT-licensed knowledge-graph store behind a lot of agent memory.
Key specs
| GitHub stars | 9,825 |
|---|
Quick facts
| Project | Semantica — graph-native context store for AI agents |
|---|---|
| Version | v0.6.6 |
| Released | 20 August 2026 |
| Type | Security release — upgrade strongly recommended |
| License | MIT |
| Install | pip install semantica==0.6.6 |
| Previous release | v0.6.5 on 11 August 2026 |
What is it?
Version 0.6.6 of Semantica, the graph-native context and memory store for AI agents, closes a privately disclosed batch of vulnerabilities. The maintainers mark it a security release and state that "upgrading is strongly recommended". It follows v0.6.5, itself a security release, by nine days.
How does it work?
The fixes cluster around anything that touches outside data. Ingestion picked up guards for tarball path traversal, SQL injection, DNS-rebinding TOCTOU races, stored XSS, SPARQL injection, header injection and credential leakage across HTTP redirects, plus an SSRF guard applied to feed requests and a hardened GitPython clone surface in the repository ingestor. Backup and restore, database export and the triplet-store backends were patched alongside them.
Why does it matter?
An agent memory store is the one component that deliberately swallows untrusted content — web pages, feeds, repositories, uploaded files — so an injection bug there reaches everything downstream. Beyond the patches, the release adds a CrewAI integration, ContextGraph retraction and purge, a declared RDF vocabulary with deterministic IRIs, and timezone-aware timestamps across export and provenance.
Who is it for?
teams running Semantica as agent memory or a RAG knowledge graph
Frequently asked questions
- Should I upgrade to Semantica v0.6.6 right away?
- The Semantica maintainers label v0.6.6 a security release and write that "upgrading is strongly recommended". It closes a privately disclosed batch of vulnerabilities rather than a single bug, and it arrived nine days after v0.6.5, which was also a security release. Anyone running Semantica on untrusted input should treat it as a priority upgrade.
- Which parts of Semantica were vulnerable?
- The v0.6.6 fixes span backup and restore, database export, outbound requests and the triplet-store backends. The ingestion path got the widest hardening: tarball path traversal, SQL injection, DNS-rebinding time-of-check-to-time-of-use races, stored XSS, SPARQL injection, header injection and credential leakage across redirects. An SSRF guard now also covers feed ingestion.
- What did Semantica v0.6.6 add besides the security fixes?
- Semantica v0.6.6 ships first-class CrewAI integration, retraction and purge operations on ContextGraph so facts can be pulled back out of a graph, a declared Semantica RDF vocabulary with deterministic entity and relationship IRIs, and timezone-aware timestamps throughout the export and provenance modules.
- Can agents now be told to forget something stored in Semantica?
- Yes, that is what the new ContextGraph retraction and purge operations in v0.6.6 are for. Retraction removes a fact from the active graph while the provenance record shows it was withdrawn; purge deletes it outright. Both matter for agent memory that has ingested data a user later asks to have deleted.
Try it
pip install --upgrade semantica==0.6.6