I am Lyra

Harmony in Evidence

The medical research agent in the DNAi Healthcare Fleet. Lyra conducts systematic literature search, evidence synthesis, meta-analysis support, clinical-trial review, and research gap identification across a dedicated biomedical research corpus.

Try Lyra via API Get an API key

API-only access · Inherits Asha's Sacred Refusals · Citation-precise

5.1M
PubMed Abstracts
16.5M
OpenAlex Top-Cited
5.2M
PMC Full-Text Cited
13.1M
Structured Citations

What Lyra Does

Lyra is the research librarian who knows every paper and every limitation. She is built for the questions a methodical evidence reviewer asks.

📚

Systematic literature search

PICO-framed queries (Population, Intervention, Comparison, Outcome) across PubMed, OpenAlex, PMC full text, StatPearls, and clinical guidelines.

🔗

Evidence synthesis

Multi-collection retrieval combined into a coherent summary with quality grading and explicit confidence assessment.

📊

Meta-analysis support

Heterogeneity assessment, publication-bias flags, and statistical aggregation cues for downstream review software.

🧪

Clinical trial review

Trial-design evaluation, outcome interpretation, and risk-of-bias assessment using Oxford CEBM evidence levels.

🔍

Research gap identification

What the corpus does not contain. Where evidence is thin, where trials contradict, where the next study is needed.

Citation verification

Confirms PMIDs and DOIs exist and match the claim. Lyra cannot fabricate references; that is an architectural constraint inherited from Asha.

How Lyra Is Grounded

Lyra runs in platform mode with research-weighted boosts defined in agents/lyra/agent.yaml. The corpus is the same fleet-wide store of roughly 829 live collections and over 126 million vectors.

Collection Boost Scale What it carries
pubmed_abstracts 1.5x 5.1M abstracts Biomedical abstract corpus, the backbone of any review
openalex_top_cited 1.4x 16.5M works Citation-weighted top research across disciplines
pmc_fulltext_cited 1.4x 5.2M articles Open-access full text from PubMed Central
statpearls_clinical 1.3x 76K articles Peer-reviewed clinical reference
pkg2_cited_structured 1.3x 13.1M records Structured citation graph for forward and backward tracing
pkg2_keywords 1.2x 7M records Keyword-indexed research for precise term filtering
clinical_guidelines 1.2x varies Specialty-society practice guidelines

Sample Queries

Lyra is built for the questions a senior reviewer would frame. Below are representative shapes; cited evidence is returned with collection name and PMID when available.

Comparative effectiveness

"Compare GLP-1 receptor agonists and dual GIP/GLP-1 agonists for type 2 diabetes weight outcomes. Grade evidence."

Lyra retrieves head-to-head trials from pubmed_abstracts and pmc_fulltext_cited, scores study quality, and reports a confidence rating with explicit citations.

Systematic-review scoping

"Scope a systematic review on omega-3 supplementation for unipolar depression. Note heterogeneity."

Returns PICO frame, inclusion criteria, candidate trials from openalex_top_cited, and a heterogeneity flag where effect sizes diverge.

Trial-design critique

"Risk-of-bias assessment for a recent atrial-fibrillation pulsed-field ablation trial."

Pulls from statpearls_clinical and clinical_guidelines, applies CEBM grading, and flags blinding, randomization, and outcome-ascertainment concerns.

Research gap analysis

"What is the evidence gap on long-COVID neurocognitive sequelae in adolescents?"

Returns existing coverage with confidence ratings plus an explicit list of what the corpus does not answer. Useful for grant scoping and methods sections.

API Access

Lyra is available exclusively through the DNAi A2A protocol. There is no consumer chat surface; Lyra runs underneath Asha and is reachable to external researchers via API key.

Get started: sign up at dnai.systems/developers, provision an API key, and call POST /a2a/v1/message:send with agent_id: lyra. Free tier includes 50 queries per month; Developer, Pro, and Enterprise tiers scale to unlimited.

Standard endpoint

POST https://api.askasha.org/a2a/v1/message:send with bearer token. Specify agent_id: lyra in the message payload.

Task lifecycle

Async tasks accessible via GET /a2a/v1/tasks/{id}. Cancel in-flight work with POST /a2a/v1/tasks/{id}:cancel.

Discovery card

Lyra's agent card lives at /.well-known/agent-card.json with epistemic metadata, capabilities, and rate limits.

Frequently Asked Questions

Can I chat with Lyra directly?
No. Lyra has no standalone consumer surface. She is reached through the DNAi A2A API or spawned by Asha as a research subagent. For patient-facing answers, use Asha at app.askasha.org.
Will Lyra fabricate a citation if she cannot find one?
No. Citation fabrication is an inherited Sacred Refusal. Where the corpus has no answer, Lyra returns "evidence insufficient" with an explicit gap rationale.
What evidence grading does Lyra use?
Oxford CEBM levels 1 through 5, supplemented by GRADE qualifiers where the source explicitly provides them. Heterogeneity, blinding, and outcome-ascertainment risks are flagged inline.
How current is Lyra's literature?
The corpus includes a dedicated post-cutoff index ingesting recent PubMed papers, refreshed regularly. Each cited record carries a temporal anchor so you can see when the source was indexed.
Can Lyra prescribe or diagnose?
No. Lyra inherits all seven of Asha's Sacred Refusals. She synthesizes evidence; clinical decisions remain with a qualified clinician.
Is Lyra appropriate for grant-writing?
Yes. Research gap analysis and citation-graph traversal are designed for the methods sections of grants and protocols. Confirm every cited PMID against PubMed before submission as a separate review step.

Wire Lyra into your research workflow

Get an API key in minutes. Free tier available, pay only when you scale.

Try Lyra via API