From mixed research inputs to hypothesis-ready graph context.

Research agents lose time before the first useful question.

Papers, trial summaries, cohort exports, and biomarker notes arrive in different shapes from different sources. Without shared structure, PDF chunks, CSV rows, and summaries stay isolated, every agent reconstructs source context from scratch, literature evidence cannot be connected to cohort evidence, and each follow-up hypothesis vanishes at the end of the run instead of seeding the next one.

Graph intelligence on ingest

Incoming data becomes queryable graph context.

Research inputs land as PAPER records with nested PDF_CHUNK text, TOPIC tags, COHORT rows, and TRIAL summaries, all scoped to a workspace_id like oncology-01. RushDB types each field on write, auto-links PDF_CHUNK and TOPIC records to their parent PAPER, and lets suggested-relationship analysis surface connections between a paper chunk and a related COHORT or TRIAL record imported separately.

Suggested relationship analysis requires an LLM configured for the project. Suggestions stay in draft form until you approve them, so inferred domain meaning never mutates the graph silently. You can also add explicit relationships through the SDK or API.

Data model

One flexible graph for the workflow.

Start with the payload shape your product already produces. RushDB stores it as Records, infers typed properties, and keeps nested or approved domain relationships queryable.

{
  "workspaceId": "oncology-01",
  "PAPER": [{
    "paperId": "paper-egfr-2026",
    "title": "Resistance patterns in EGFR cohorts",
    "AUTHOR": [{ "name": "Dr. Lee" }],
    "PDF_CHUNK": [{ "chunkId": "p1-c4", "text": "Response varied after acquired resistance..." }]
  }],
  "TRIAL": [{
    "trialId": "trial-9",
    "COHORT": [{ "cohortId": "cohort-egfr", "sampleSize": 84, "BIOMARKER": [{ "name": "EGFR" }] }]
  }],
  "HYPOTHESIS": [{ "title": "Compare outcomes by resistance marker", "status": "draft" }]
}

Working example

Persist the evidence trail and the next hypothesis.

A literature scout searches paper chunks for EGFR resistance evidence, then the workflow keeps the connected paper, cohort, trial, and follow-up hypothesis available for the next pass.

PAPER paper-17
  TOPIC EGFR resistance
  PDF_CHUNK "Response varied after acquired resistance..."

COHORT cohort-4
TRIAL trial-9
HYPOTHESIS "Compare outcome differences by resistance marker"

{
  "labels": ["PDF_CHUNK"],
  "propertyName": "text",
  "query": "EGFR resistance evidence with outcome differences",
  "where": { "workspace_id": "oncology-01" }
}

{
  "paper": "paper-17",
  "topic": "EGFR resistance",
  "cohort": "cohort-4",
  "trial": "trial-9",
  "next_hypothesis": "Compare outcome differences by resistance marker"
}

Python SDK

Ingest once. Ground each research loop in live structure.

Keep document parsing and domain review in your application. RushDB stores the resulting records, exposes schema, and retrieves relevant evidence for the next step.

from rushdb import RushDB

db = RushDB('RUSHDB_API_KEY')

schema = db.ai.get_schema_markdown({'labels': ['PAPER', 'PDF_CHUNK', 'TRIAL', 'COHORT']}).data
evidence = db.ai.search({
    'labels': ['PDF_CHUNK'],
    'propertyName': 'text',
    'query': 'EGFR resistance evidence with outcome differences',
    'where': {'workspace_id': 'oncology-01'},
    'limit': 5,
})
cohorts = db.records.find({'labels': ['COHORT'], 'where': {'BIOMARKER': {'name': {'$contains': 'EGFR'}}}, 'limit': 10})

Implementation blueprint

Build the medical research-loop path.

Use this sequence to connect paper chunks, trial metadata, cohort rows, and follow-up hypotheses in a research workspace.

Relevant docs

Read the exact primitives behind this pattern.

These links point to the RushDB docs pages that map directly to this blueprint: ingestion, labels, properties, values, SearchQuery, relationships, semantic search, MCP, or deployment.

How it works

Build the smallest useful workflow first.

Know where it fits.

Questions developers ask.