About EvidenceLens

Biomedical research is fractured across PubMed, preprint servers, clinical trial registries, regulatory agencies, and funding databases — each with its own search interface, query language, and access model. EvidenceLens was built to close that gap: it is a free, public, agentic search engine that indexes documents from all these sources into a single corpus and lets you search across them as one body of evidence.

The goal is to make the full landscape of biomedical evidence accessible without a paywall or institutional subscription. Whether you are a clinician doing a literature review, a researcher checking whether a trial has been published, or a patient trying to understand the evidence behind a treatment, EvidenceLens gives you one place to look — not six.

Every document is processed through an automated ingestion pipeline that fetches, normalizes, and embeds it into a vector index. When you search, results are scored by a hybrid ranker that combines BM25 keyword relevance, dense vector similarity, citation graph influence, and recency. The system surfaces what is most relevant rather than what happens to rank well in a single signal. You can also examine industry payments to physicians via a standalone CMS Open Payments lookup to surface financial relationships that may bear on the evidence.

EvidenceLens is open source (MIT) and entirely self-hosted on a single Docker Compose stack. NATS JetStream queues documents through the pipeline; Meilisearch handles full-text indexing; Milvus manages vector embeddings; and Neo4j stores the citation graph that feeds the ranker. The frontend streams results over WebSocket so you see hits as they are scored, and an agentic layer lets you ask questions against the corpus in natural language via a BYOK (bring-your-own-key) proxy that supports Anthropic, OpenAI-compatible, and Ollama backends. A WebLLM option runs a small model entirely in-browser for users who prefer not to send queries to any server.

It is not medical advice. EvidenceLens is a research tool. Always verify findings against primary sources and consult a clinician for any health decision.