What you can do with DerridAI

Import documents, plain text, images, audio, and selected online sources. PDF extraction can use OCR, while audio records can retain timed passages and speaker information. Each source keeps the locators that make sense for its medium.

Configure a build, monitor extraction and enrichment, and start reviewing available records while background work continues. Publication checks surface unresolved problems before you publish.

Define fields, accepted values, and evidence rules with a metadata schema. Inspect proposed values alongside their passages, then confirm, correct, or reject them. The record keeps both the origin of the proposal and the review decision.

Distinguish document author, speaker, position holder, and stance where the schema calls for them. Reviewed precedents can guide later suggestions, and optional Document Intelligence can add entities, coreference, and quotation context as advisory annotations.

Browse works and records with filters, then move between bibliographic metadata, source locations, and annotations. Semantic maps help you explore connections between concepts, people, and passages.

Search by words or meaning. Filter by the fields in your corpus, and rebuild the search index when needed. The index is a derived view of your records.

Ask a question over a selected corpus or supply passages yourself. Retrieval can combine keyword and semantic search, diversify results, remove duplicates, and rerank candidates before the model receives the evidence packet.

Inspect the passages and citations behind an answer. Use claim support links where available to check the proposition against the record revision, including who holds the position and whether the answer preserves the source’s qualifiers.

Use Pipeline Studio to inspect versioned workflows, stage inputs, traces, and timing. Compare variants or use fixed cases to see how a change in retrieval or model choice affects the result.

Assess response quality alongside latency. Review citation accuracy, attribution, evidence fidelity, and interpretive usefulness against the source passages.

Export an immutable snapshot of selected works as a portable research site. Readers can browse records, filter metadata, annotate locally, and search. Exports can include browser-ready vectors or let readers build a semantic index on their device.

Run DerridAI on infrastructure you control, use a local model or a compatible endpoint, and manage access through roles. Choose what to publish according to the rights and access rules for your sources.