Open-source accessibility scanning and supported remediation for higher education.
Aelira helps institutions find accessibility issues in course materials and apply supported changes, with recorded outcomes and human review. Our public core covers documents, LMS course content, web pages and media.
The core is in beta and under active validation. Some files need manual work, and some cannot produce a downloadable remediated artifact. Automated scores describe scanner findings, not accessibility conformance. Review any saved candidate before publishing it.
aelira-core contains the FastAPI engine, React dashboard and TypeScript command-line client.
- Documents: PDF, Word, PowerPoint, Excel and LaTeX scanning, supported remediation and review artifacts.
- LMS: course discovery, scanning and provider-specific remediation workflows.
- Web: browser-based accessibility detection and generated code fixes.
- Media: transcription, captions and related review outputs.
See the quickstart and document capability guide for the current scope.
The self-hosted core is free, with no paid feature tier. The engine and dashboard are AGPL-3.0; the CLI is MIT-licensed. Hosted infrastructure and support are available through aelira.ai.
Integration maturity varies. Canvas is integration verified, with institution-specific user-journey acceptance still needed. Brightspace is beta; Blackboard and Moodle are experimental. Moodle uses a REST API connector. Review the current LMS status table before relying on a connector. Google Drive and Microsoft 365 connectors are also available, with permissions and operation limits described in the core documentation.
Severity is computed from scanner rule and impact data rather than generated by a language model. Supported remediation changes differ by format and source structure; ambiguous or unsupported cases remain unresolved.
You can select a cloud AI provider or configure local inference through Ollama. Institution-controlled AI processing depends on the actual provider, fallback, storage, logging and network settings. Local inference alone does not establish a privacy or compliance guarantee. See the self-hosting guide and LMS AI policy guide.
We aim to evaluate how AI assists practitioners with accessibility remediation, including source fidelity, errors and the time spent checking outputs. Planned comparisons include generalist agents, specialised accessibility workflows and further-trained open-weight models.
Our longer-term goal is an accessibility specialist model. This is a research direction, with public evaluation tools, rights-cleared material and findings as intended outputs.
Founded and technically maintained by RD (Reg) Crampton. Contributions, evaluation collaborators and accessibility practitioner feedback are welcome.
- Contributing guide
- Maintainers and governance
- Support routes
- Developer onboarding
- Discussions and setup questions
- Report a bug
- Request a feature
- Security reporting policy
Website · LinkedIn · X · Self-hosting guide