For Developers

POST /api/chat streams Server-Sent Events. Send an OpenAI-style message array; receive token deltas and tool-call results (PharmGKB, KEGG, cBioPortal, GDC, AlphaFold) as they resolve.

curl -N -X POST https://genrox.aafiyet.com/api/chat \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"CYP2D6 *4/*4 on codeine"}]}'

POST /api/public/v1/chat is the authenticated, API-key-based equivalent of the chat API for external integrations. Pass your key as a Bearer token; the same tool-augmented streaming response is returned.

API access is part of aafinomics Pro. API calls share your Pro allowance with the website: 50 questions per rolling 5 hours, then pay-as-you-go credits. Each response carries X-Quota-Used and X-Quota-Limit headers; a 429 means the window and credits are used up, and a 403 means the key's account is not on Pro.

curl -N -X POST https://genrox.aafiyet.com/api/public/v1/chat \
  -H "Authorization: Bearer grx_..." \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"What is CYP2D6 *4?"}]}'

Python SDK (requests, stdlib-friendly):

import requests

API_KEY = "grx_..."

resp = requests.post(
    "https://genrox.aafiyet.com/api/public/v1/chat",
    headers={"Authorization": f"Bearer {API_KEY}"},
    json={"messages": [{"role": "user", "content": "What is CYP2D6 *4?"}]},
    stream=True,
)

for chunk in resp.iter_content(chunk_size=None, decode_unicode=True):
    if chunk:
        print(chunk, end="")

For scripted, bulk queries against a CSV/TSV of prompts, download the dependency-free batch CLI:

Create and manage API keys for the public REST API. Keys require an active Pro plan.

Run generated code on infrastructure you control. The runner long-polls POST /api/ngs-submit to claim jobs and posts results back, authenticated with a node key. POST /api/compute-ping validates connectivity from the Compute setup dialog.

# Poll for and execute compute jobs
python runner.py --endpoint https://your-host/api/ngs-submit \
  --key $GENROX_NODE_KEY

hpc/hpc_evo2_worker.py is a persistent daemon that loads Evo 2 in bfloat16 with FlashAttention, pins weights in VRAM for the allocation window, and long-polls the orchestrator for scoring jobs — outbound only, so firewalled compute nodes need no inbound access. Launch it with hpc/launch_slurm_worker.sh (1 node, A100/H100 80GB, 16 CPUs, 64GB RAM, 8h wall time) via SLURM with graceful SIGTERM drain.

backend/main.py is a FastAPI + Redis reference orchestrator: a cached star-allele fast path, a reliable job queue with lease reaping, authenticated claim/complete/heartbeat endpoints, and a WebSocket status stream (Queued → Assigned → Scoring in VRAM → Complete).