incheck

The official Python SDK for the InCheck AI platform. Two operating modes, one client, typed end-to-end:

• EMS mode — ask the model an EMS protocol question. No setup required beyond an API key.
• Unified mode — onboard one or more documents into a Pod (one Pod per org_id), then chat against that Pod and the answers are grounded in your content.

Status: 0.0.2 — chat (EMS + unified, sync/async/streaming) and document onboarding (Pods, presigned uploads, processing jobs).

Install

pip install incheck
# or
uv add incheck

Requires Python 3.10+.

Authenticate

Generate an API key from the InCheck admin (/admin/api-keys — Teams plan + admin role) and set the env var:

export INCHECK_API_KEY="incheck_prod_..."

Your key is bound to one organization. InCheck derives a namespace from your org's subdomain — every org_id you use must start with <namespace>_. Cross-namespace requests are rejected at the gateway.

EMS mode — no setup, just chat

from incheck import Client

with Client() as client:
    reply = client.chat.send(
        "Adult dose of atropine for symptomatic bradycardia?",
        scope="ALS",
        state="Massachusetts",
    )
    print(reply.content)

That's it. The model answers from general EMS knowledge under the given scope/state. No org_id needed.

Unified mode — chat with your documents

Onboard documents once (a Pod), then chat against that Pod. A Pod is identified by your org_id; you can hold multiple files in the same Pod and query across all of them.

from incheck import Client

with Client() as client:
    namespace = client.documents.list_orgs().filtered_by
    org_id = f"{namespace}_dispatch"           # your Pod

    # 1. Onboard documents — initiate → upload → complete → poll, in one call.
    status = client.documents.upload(
        org_id,
        files=["./dispatch_sop.pdf", "./policies.docx"],
    )
    print("processed:", status.progress.processed_pages, "pages")

    # 2. Chat against the Pod
    reply = client.chat.send(
        "What's our hazmat escalation policy?",
        org_id=org_id,                          # ← turns on unified mode
        user_id="alice@hospital.org",
    )
    print(reply.content)

You can add or update files in the Pod later with documents.upload(...) again, or scope the change with initiate_update / complete_update.

How does the document processing work? That's our IP. The contract you see — upload, wait for the job to complete, query — is the whole public surface. If you need deeper guarantees about extraction accuracy, retention, or custom pipelines, talk to us.

Streaming

Both modes support streaming:

for chunk in client.chat.stream("Summarize the SOP.", org_id="acme_dispatch"):
    if chunk.content:
        print(chunk.content, end="", flush=True)

Async

The SDK ships an async client with the same method names:

import asyncio
from incheck import AsyncClient

async def main():
    async with AsyncClient() as client:
        # EMS
        r = await client.chat.send("Adult dose of epinephrine for anaphylaxis?")
        print(r.content)

        # Unified
        await client.documents.upload("acme_dispatch", ["./sop.pdf"])
        r = await client.chat.send("Summarize.", org_id="acme_dispatch")
        print(r.content)

asyncio.run(main())

Environments

Pick one in code or via env var. Production is the default.

Client(environment="staging")          # explicit
# or: INCHECK_ENVIRONMENT=staging      # via env

Endpoints covered

Documents

Chat

AsyncClient exposes the same methods under the same names.

Errors

Every non-2xx response becomes a typed exception:

from incheck import (
    Client,
    AuthenticationError,
    PermissionError,
    ValidationError,
    JobFailedError,
    JobTimeoutError,
    RateLimitError,
)

with Client() as client:
    try:
        client.documents.upload("royal_dispatch", ["./sop.pdf"])
    except PermissionError as e:
        print("namespace mismatch:", e)
    except ValidationError as e:
        print("bad request:", e)
    except JobFailedError as e:
        print("processing failed:", e.job_id, e.status)
    except JobTimeoutError as e:
        print("still pending:", e.last_status)
    except RateLimitError as e:
        print(f"slow down; retry after {e.retry_after}s")
    except AuthenticationError:
        print("check your API key")

Hierarchy: IncheckError → AuthenticationError, PermissionError, NotFoundError, ValidationError, RateLimitError, APIError, APIConnectionError, JobFailedError, JobTimeoutError.

Configuration

You can also pass api_key=, environment=, and base_url= explicitly to either client.

Breaking changes since 0.0.1

chat.send and chat.stream no longer take org_id as the first positional argument. The new signature is:

client.chat.send(content, *, org_id=None, ...)

Callers that previously did chat.send(org_id, content) should switch to chat.send(content, org_id=org_id). Omitting org_id is the new EMS-mode shape.

License

MIT.