maxmodel (Python)

The AI gateway with verified output. One API for every model, with a verified mode that returns answers grounded in your sources: verbatim citations, hallucinations stripped, checked by deterministic code (not an LLM judge).

pip install maxmodel

from maxmodel import MaxModel

mx = MaxModel(api_key="sk-...")  # your maxmodel.com gateway key

out = mx.verified.create(
    model="gpt-5.5-pro",
    messages=[{"role": "user", "content": "What is our refund policy?"}],
    sources=[{"id": "refunds.md", "text": "Full refunds within 30 days. No refunds after 30 days."}],
)

print(out.text)                 # grounded answer with inline [refunds.md] markers
for c in out.citations:
    print(f'✓ "{c.quote}" — {c.source} {c.range}')   # verbatim, traced to your source
for u in out.unsupported:
    print(f'⚠ dropped ({u.reason}): "{u.claim}"')     # hallucinations caught, not shipped

Why it's different

The grounding check is deterministic code: every cited quote must appear character-for-character in a source you supplied, or the claim is dropped. No "LLM judging an LLM".

API

MaxModel(api_key, *, base_url=..., timeout=60.0)

• base_url defaults to https://api.maxmodel.com; override for self-host/testing.

mx.verified.create(...) -> VerifiedResult

VerifiedResult (dataclass): id, model, text, citations, unsupported, grounded, usage.

mx.chat.completions.create(**params) -> dict

OpenAI-compatible passthrough.

Errors

Typed and catchable: MaxModelError (base) with AuthError (401), QuotaError (402), RateLimitError (429), NoSourcesError (400), ExtractionError (502), GatewayError (5xx). Each carries .type and .status.

from maxmodel import RateLimitError, NoSourcesError, MaxModelError

try:
    out = mx.verified.create(...)
except NoSourcesError:
    ...   # you forgot sources, or set allow_ungrounded=True
except RateLimitError:
    ...   # back off and retry
except MaxModelError as e:
    print(e.type, e.status)

Zero runtime dependencies (stdlib only). Python 3.8+. Docs: https://docs.maxmodel.com