DerridAI

Research applicability

Local-first scholarly corpus environment documenting source ingestion, researcher review, provenance-aware metadata and evidence-grounded retrieval-augmented generation.

Getting started

Follow the README clean-checkout configuration and model setup, start Docker Compose, create the initial account and review a small corpus before querying its evidence. These are documentation-based editorial adoption suggestions, not an exercised workflow.

Maintenance

Maintenance and runtime behavior not independently audited.

License

README identifies GNU Affero General Public License v3.0; obligations and dependency licenses not audited.

Workflow & human oversight

Corpus setup, build, review and publication plus research querying form a defined researcher-supervised workflow; provenance architecture is documented, not independently tested here.

Workflow level is not an evidence score; higher is not better evidence.

Approach

Docker application separates sources, records, revisions, assertions, evidence and generated claims; derived search indexes support retrieval while linked evidence supports auditing.

Limitations

No installation, source-code security audit, retrieval/citation accuracy evaluation or scholarly outcomes reproduced. Local-first is not a blanket offline guarantee: imports and remote providers may use networks. Generated claims and automatically enriched metadata still require source checks. The retrieved README was head/tail truncated; reviewed introductory, architecture, setup and license sections support this bounded listing, not an exhaustive feature audit.

Reproducibility

Primary repository documentation checked live. No installation, software execution, security audit, or outcome replication performed.

Adoption requirements

Git, Docker with Compose, an LLM endpoint and configured generation/embedding models; optional local Ollama or other supported endpoint.