Literature Review Agent Release Notes

Release date: 06-04-2026

What’s New

This is the initial public release of the Literature Review Agent. The full feature set is available from day one.

5-step guided UI — Search, Data Points & Criteria, Run & Monitor, Select Literatures, and Literature QA.
Multi-source search — DeepLens medical knowledge bases, PubMed (NCBI), Semantic Scholar, Europe PMC, and OpenAlex, all queryable from a single screen.
Keyword, semantic, and document-ID search modes with MeSH query expansion for PubMed and DeepLens sources.
Rich result filters — date, article type, language, journal, author, open-access, impact factor, journal quartile, h-index, and citation count.

Plain-language data points — define what to extract using a label and a natural-language prompt; no fixed schema required.
Inclusion/exclusion criteria — per-article verdicts with supporting explanations for every rule.
AI-assisted criteria generation — describe your review in one sentence and get a drafted set of data points and rules to edit.
Evidence-backed extraction — every extracted value is paired with the exact source sentence.

Distributed extraction with live progress monitoring (total / processed / succeeded / failed / % complete).
Delta processing — add new articles to an existing run without reprocessing completed ones.
Copy into a new run — clone a run’s full configuration for reuse or experimentation.
Run history sidebar — all previous runs accessible with a single click.
CSV export — full result set with metadata, extracted values, evidence sentences, and verdicts.

Select Literatures — curate a focused article set from extraction results.
Streaming Q&A — ask questions grounded in selected articles; answers stream in real time with source citations.

ZIP upload — ingest .pdf, .txt, and .docx files from a local archive.
S3 ingestion — point to an S3 prefix and the worker fetches and indexes everything automatically.

PREVIOUSSupport