Connector · AI models
OpenAI, in every client’s workflows.
Ask a model, with the client’s own OpenAI account.
- Triggers
- 0start a workflow
- Actions
- 1steps in a workflow
Do it in OpenAI
A step with its own fields: values from the run go in by name, and what OpenAI answers is there for the next step.
- Ask a modelOne prompt, one reply: the output’s Reply.
Connect the client’s OpenAI
On the client’s page in Thrum: Add credential, type OpenAI API key.
platform.openai.com → API keys → Create new secret key, in the project that bills this client.
The key is stored encrypted and bound to that client: another client’s workflow cannot use it, and no template or pack copies it.
Watched like everything else
When OpenAI stops answering or refuses the client’s key, you hear about it, and your client does not.
- A failing request
- Tried 3 times, 5 seconds apart and doubling, only for errors a retry can fix
- A service that asks for a pause
- Given it, up to an hour
- Requests at once, per account
- 2, per worker; the rest wait their turn
- Timeout
- 15 seconds, unless the step sets its own
- Private and internal addresses
- Refused, on every redirect too
- The client’s key
- Encrypted, bound to the client, and kept out of every run record
- A step that keeps failing
- Marks the workflow Degraded and alerts you, not the client
More in AI models
OpenAI and its logo are trademarks of their owner, used here only to say what Thrum works with; no endorsement is implied.
Thrum is cooking.
Get early access and you’ll be among the first in. Early members get a launch perk.