v2.0.0 — Memgraph migration & cloud-native deployment

ComptoxAI v2.0.0 — Memgraph migration & cloud-native deployment

This is a major release that re-platforms ComptoxAI's knowledge graph from Neo4j
to Memgraph, introduces reproducible cloud infrastructure, and substantially
expands the REST API. The graph backend change is breaking: deployments must now
point at Memgraph (with the Bolt proxy / Memgraph Lab) rather than Neo4j.

⚠️ Breaking changes

• The graph database backend is now Memgraph instead of Neo4j. API
  connections, queries, and user-facing tooling (Memgraph Lab in place of the
  Neo4j browser) have all moved accordingly.

Highlights

Graph database: Neo4j → Memgraph

• Migrated the API, CI, and ISTA mapping configuration to run against Memgraph.
• Updated all user-facing references and documentation from Neo4j to Memgraph.
• Surfaced Memgraph Lab from the browse page for interactive graph exploration.

Cloud infrastructure & deployment

• Added Terraform for provisioning the hosting environment.
• Added Docker Compose infrastructure for the full stack.
• Build comptox-api and bolt-proxy as multi-architecture container images.
• Pointed the apex site at the new api / lab subdomains; bumped Caddy
  (2.10 → 2.11) for the Go 1.25 toolchain and the caddy-l4 dependency.

REST API

• New /stats endpoint plus convenience and cross-entity chemical endpoints.
• New query builder for constructing graph queries.
• Fixed path search so adjacent nodes correctly return a path (BFS lower bound).
• Fixed a parallel-transaction bug in /stats and raised the cache TTL to 7 days.

Chemical featurization

• Added the chemical featurizer with support for 3D conformer-based features and
  pretrained featurizers from Hugging Face and DGL.
• Added a QSAR scripting workflow.

Website, docs & build

• Fixed /data.html to load the current Vite bundle; converted the
  DatasetBuilder from CommonJS require() to ESM import.
• Several Sphinx documentation template fixes.
• CI now runs reliably against the Memgraph stack.

Version 1.2.0

This version includes a major update of many source databases and greatly expanding the set of chemicals represented in the knowledge graph, among several data model changes and performance improvements.

We have begun to simplify interaction with graph databases, and fixed many bugs in connecting to the new version (both locally and remotely) of the knowledge graph. Importantly, we have deprecated the use of a config file, with all options being provided to methods. New sane defaults are also included in method parameters to cover the most common use cases.

The compatible version of Neo4j is 5.7.0.

Version 1.1.2

• SMILES strings are added to chemical nodes in the knowledge graph
• YouTube video tutorials page is now included in the User Guide
• Various content tweaks and fixes on the website

Version 1.1

We've made a lot of improvements to ComptoxAI since v1.0! Some of the big ones include:

• New data browsing and information retrieval tools, available via the website
• Significantly reduced repository size (due to removing large data files)
• Improved documentation (although there are still some gaps - please contribute via pull requests if interested)
• Expanded the scope of the knowledge base
• Cleaned up the knowledge base by removing extraneous ontology nodes and unimportant metadata
• Simplified the Python package in many ways. Most of the functionality at this point is contained in comptox_ai.db.GraphDB, and new features will be coming soon.
• Many more small changes.

Keep an eye on the repository for new features, fixes, and other goodies - we'll be keeping our users up to date with more frequent patch releases from this point onward!

Version 1.0

This version marks the first major release of ComptoxAI that is ready for use in scientific applications.

Key features of v1.0 include:

• Interactive data browser
• Public read-only Neo4j instance containing the graph database
• Fully revised database build process; inclusion of assays and other new datatypes
• Expansion to over 840,000 nodes and 1.3M edges in the database
• Documented web API with new routes
• Improved static website content
• Summary of different interfaces for accessing data