Troubleshooting and FAQ
Frequently asked questions
Short answers to the questions people ask most about SentientOne: models, keys, knowledge, the widget, the API, credits and data.
Reference3 min readUpdated 23 October 2026
Can't find your question? Search the help centre, or email support@sentientone.ai.
The platform
What is SentientOne?
An AI platform for building agents: you give an agent a prompt, a model, your knowledge and your tools, then ship it through a streaming REST API, a website widget or a workflow, and trace every request. See What is SentientOne.
Which models can I use?
Models from OpenAI, Anthropic, Google Gemini and Groq, chosen per agent. See Supported models.
Do I need my own model provider key?
Yes. Every agent runs on your own provider key, saved encrypted. Your provider bills you for tokens at their prices; SentientOne doesn't mark them up.
Is there a self-hosted version?
Not as a download. For custom hosting on your own infrastructure, email support@sentientone.ai.
Can my team share an account?
Accounts are for one person; there are no team invites or roles.
Knowledge
What can I add as knowledge?
PDFs up to 50 MB in the app (plain-text files too through the API), websites (up to 200 pages per crawl), snippets up to 300 characters, and Q&A pairs.
How does search work?
Each question runs a pgvector search and a Postgres full-text search, merges them with reciprocal rank fusion, reranks with Cohere rerank-v3.5 and gives the top 5 passages to the model. See How retrieval works.
Can I see which sources an answer used?
Yes: the stream sends a sources event before the answer, and the Playground shows them beside each reply.
API and widget
Is there a non-streaming API?
Chat is streaming only. Read the events and join the delta text. Workflow runs can return JSON with "stream": false.
How do I continue a conversation?
Send the conversation_id from the previous reply. History is kept for you. See Conversations.
Can I put an agent on my website without exposing my API key?
Yes. The widget uses a publishable pk-so- key that only reaches one agent, from the websites you allow. Secret keys are refused on the widget.
Can I style the widget?
Choose a style (bubble, side panel or popup), an accent colour, the name in its header and its welcome line. See Widget options.
Does the widget work on WordPress, Shopify or Webflow?
Yes, anywhere you can add a script tag. See Install on your platform.
What's the rate limit?
60 requests a minute per agent for new agents, shared across channels; the widget adds 120 a minute per key and 12 per visitor.
Credits and billing
What does one credit buy?
One agent reply from any channel. Knowledge costs 5 credits per PDF or website page and 1 per snippet or Q&A pair.
Do credits expire?
No.
What happens when I run out?
Requests get 402 OUT_OF_CREDITS and the widget says the assistant is unavailable. Buy credits and everything resumes. You get emails before it happens.
Is there a subscription?
No. Prepaid packs from US$10 for 1,000 credits, and optional auto top-up.
Data and security
How are my provider keys protected?
They're encrypted with AES-256-GCM before they're stored and never sent back to the browser in full.
How long are traces kept?
90 days. Emails, API keys and phone numbers are redacted before they're stored.
How do I debug a wrong answer?
Open its trace: what was retrieved, each model round, every tool call, tokens and timing. See Debug a bad answer.
Still stuck?
Email us with what you tried and any request id. A person replies.