REST APIhttp://localhost:8000

API Reference

Integrate PetriDish's knowledge graph reasoning into your research pipeline. No authentication required during beta.

POST/reasonNatural language → multi-hop graph reasoning → synthesised answer

Request body

{
  "question": "What genes does Metformin target and what diseases are they associated with?",
  "max_hops": 3,
  "india_context": true
}

Response

{
  "answer": "Metformin targets GPD1, PRKAB1, and ETFDH...",
  "paths": [{ "nodes": [...], "edges": [...], "confidence": "HIGH" }],
  "cypher_steps": [{ "step": "...", "cypher": "MATCH ..." }]
}
POST/cypherExecute a read-only Cypher query directly against the Neo4j graph

Request body

{
  "cypher": "MATCH (d:Drug)-[:TARGETS]->(g:Gene)-[:ASSOCIATED_WITH]->(dis:Disease) WHERE toLower(d.name) CONTAINS 'metformin' RETURN d.name, g.name, dis.name LIMIT 10"
}

Response

{
  "results": [
    { "d.name": "Metformin", "g.name": "GPD1", "dis.name": "hypertriglyceridemia" },
    ...
  ]
}
GET/healthCheck API health, Neo4j connectivity, and active LLM provider

No body

Response

{
  "status": "ok",
  "neo4j": "connected",
  "node_count": 90408,
  "llm_provider": "groq",
  "llm_model": "llama-3.3-70b-versatile"
}
GET/statsFull graph statistics — node counts by label, edge counts by type

No body

Response

{
  "nodes": [{ "label": "Gene", "count": 26973 }, ...],
  "edges": [{ "type": "SYNERGISTIC_WITH", "count": 2672628 }, ...]
}

Graph Schema

Node labels:
  Drug           — 7,951 nodes (DrugBank)
  Disease        — 11,783 nodes (OMIM, MONDO)
  Gene           — 26,973 nodes (UniProt, NCBI)
  Pathway        — 2,516 nodes (Reactome, KEGG)
  Phenotype      — 6,465 nodes (HPO)
  Phytochemical  — 16 nodes (IMPPAT 2.0)
  Anatomy        — 4,861 nodes (Uberon)
  BiologicalProcess — 19,755 nodes (GO)
  Variant        — 14 nodes (IndiGen/PharmGKB)

Key relationships:
  Drug -[:TARGETS]-> Gene                  (16,258 edges)
  Drug -[:INDICATED_FOR]-> Disease         (4,153 edges)
  Drug -[:CONTRAINDICATED_FOR]-> Disease   (12,400 edges)
  Drug -[:SYNERGISTIC_WITH]-> Drug         (2,672,628 edges)
  Gene -[:ASSOCIATED_WITH]-> Disease       (46,205 edges)
  Gene -[:EXPRESSED_IN]-> Anatomy          (223,632 edges)
  Gene -[:PROTEIN_PROTEIN_INTERACTION]-> Gene (640,708 edges)
  Phytochemical -[:HAS_TRADITIONAL_USE]-> Disease (39 edges)
  Variant -[:AFFECTS_RESPONSE]-> Drug      (52 edges)

All nodes: id (string), name (string), source (string)

Code Examples

import requests

# Drug repurposing query
response = requests.post("http://localhost:8000/reason", json={
    "question": "Which drugs could be repurposed for MDR-Tuberculosis?",
    "max_hops": 3,
    "india_context": True
})
data = response.json()

print(data["answer"])
for path in data["paths"]:
    nodes = " → ".join(n["name"] for n in path["nodes"])
    print(f"[{path['confidence']}] {nodes}")
None (beta)
Auth required
Groq free tier
Rate limit
4.3M edges · 90K nodes
Graph size