Async and polling

Some operations run a model and take a while: gcai chat create and gcai playbooks run. The CLI handles the waiting for you.

Wait by default

chat create and playbooks run submit the job, poll it to completion, and print the finished result. You do not need to run a separate poll loop. If the job ends in an error, the command exits non-zero with the error message.

# Blocks until the answer is ready, then prints it
gcai chat create "Draft a mutual NDA" --playbook-ids <id>

Two flags control the wait:

Flag	Effect
`--async`	Return the pending job immediately instead of waiting. You get a `job_id` to poll yourself.
`--timeout <seconds>`	Bound the client-side wait before giving up (default 300).

Polling a job yourself

If you used --async (or want to check on a job later), poll it with jobs get:

gcai chat create "…" --async          # returns { job_id, status: "pending", … }
gcai jobs get <job_id> --wait 30       # server long-polls up to 30s, then returns

gcai jobs get keeps its own --wait <seconds> for a single server-side long-poll. Re-run it until status is succeeded, failed, or canceled.

File processing

Uploads process after they land (text extraction and embeddings). This is not a job envelope: poll the file itself until it is ready.

gcai files upload ./contract.pdf       # returns the file with status: extracting
gcai files get <file_id>               # re-run until status: ready

REST

MCP

CLI

Async and polling

Wait by default

Polling a job yourself

File processing

​Wait by default

​Polling a job yourself

​File processing

Wait by default

Polling a job yourself

File processing