omtx 0.3.4

Python SDK for Om

OMTX Python SDK

Lightweight helper for the OM Gateway. Queue diligence jobs, poll for results, and unlock/stream private data sets.

Installation

pip install omtx

Quick start

from omtx import OMTXClient

client = OMTXClient()  # picks up OMTX_API_KEY

# 1) Generate claims (returns 202 + job_id)
claims_job = client.diligence_generate_claims(
    target="BRAF",
    prompt="Summarize known inhibitors"
)

# 2) Check status immediately (returns metadata + payload when ready)
job_record = client.job_status(claims_job["job_id"])
print(job_record["status"])

# 3) Optionally wait for completion (defaults to a 1 hour timeout)
claims_result = client.wait_for_job(
    claims_job["job_id"],
    result_endpoint="/v2/jobs/generateClaims/{job_id}",
)
print("total claims:", claims_result["total_claims"])

# 4) Resynthesize report using a gene_key, auto-waiting for completion
report = client.diligence_synthesize_report(
    gene_key="acad8",
    wait=True,
)
print(report["sections"][0]["title"])

# 5) Deep research with automatic polling
deep = client.diligence_deep_research(
    query="CRISPR applications in cancer therapy",
    wait=True,
)
print(deep["final_report"][:400])

Setup

Get an API key

1. Sign up at https://omtx.ai
2. Generate an API key from the dashboard

Provide the API key

export OMTX_API_KEY="your-api-key"

The SDK defaults to the hosted gateway at https://api-gateway-129153908223.us-central1.run.app. If you need a different deployment, pass base_url explicitly (or set OMTX_BASE_URL).

Or pass both API key and base URL when constructing the client:

from omtx import OMTXClient

client = OMTXClient(
    api_key="your-api-key",
    base_url="https://api-gateway-129153908223.us-central1.run.app",
)

Usage examples

Error handling and context manager

from omtx import OMTXClient, InsufficientCreditsError, OMTXError

with OMTXClient() as client:
    try:
        report = client.diligence_synthesize_report(gene_key="acad8", wait=True)
    except InsufficientCreditsError:
        print("Add credits before running resynthesis jobs.")
    except OMTXError as exc:
        print(f"Gateway call failed: {exc}")

Selective data access

from omtx import OMTXClient

client = OMTXClient()

# 1) Unlock a dataset (consumes one Access Credit)
client.access_unlock(protein_uuid="aa11bb22", gene_name="KRAS")

# 2) Stream the private dataset
stream = client.data_access_selective_stream(
    dataset="selectivity",
    protein_uuid="aa11bb22",
    limit=100_000,
    fmt="csv",
)

print("Rows:", stream.headers.get("X-Row-Count"))
with open("kras_selective.csv", "wb") as fh:
    for chunk in stream.iter_bytes():
        fh.write(chunk)
stream.close()

Gene key discovery

from omtx import OMTXClient

client = OMTXClient()
gene_keys = client.diligence_list_gene_keys(min_true=5)
print(gene_keys["items"][:5])

Retries & idempotency

from omtx import OMTXClient

client = OMTXClient()

# Every POST call receives a timestamp-based idempotency key automatically.
# Provide `idempotency_key=` when you need deterministic retries.
claims = client.diligence_generate_claims(
    target="BRAF",
    prompt="Summarize inhibitors",
    idempotency_key="braf-inhibitors-run-1",
)

Available helper methods

• diligence_generate_claims(target, prompt, wait=False)
• diligence_synthesize_report(gene_key, wait=False)
• diligence_deep_research(query, wait=False, …)
• diligence_list_gene_keys(min_true=1, …)
• jobs_history(...), job_status(job_id), wait_for_job(job_id, …)
• wait_for_job uses a 3,600 second default timeout (set timeout=None to wait indefinitely)
• access_unlock(protein_uuid, gene_name=None), list_access_unlocks()
• data_access_selective_stream(...), data_access_points_stream(...)
• data_access_selective_stats(...), data_access_points_stats(...)
• credits(), health()

Requirements

• Python 3.9 or higher
• An OMTX API key

Support

• Email: [e-mail]
• Docs: https://docs.omtx.ai
• Issues: https://github.com/omtx/python-sdk/issues

License

MIT License – see LICENSE for details.