Skip to main content

Public Tutor

HRIDA TUTOR answers visitors' questions about Hrida.ai on the public pages, without signing in: a chat button in the bottom-right corner of the landing, about and demo pages, and the Ask anything about Hrida AI section on the landing page. You set it up in Admin Panel › Settings › Public Tutor.

How it answers​

Each question goes to the model you choose, as the account that owns the API key you save. The model's system prompt, knowledge, tools and filters apply as in a normal chat, so the tutor answers from whatever you give that model. The question is sent with the last few messages of the conversation, so follow-up questions work. Visitors only see the answers: the model and the key are never sent to the browser. The chat button keeps the conversation while the visitor moves between the public pages, until the browser tab is closed.

Set it up​

  1. Choose the model that answers. To answer from the Hrida.ai documentation, use the HridaDoc Agent from Digital Employees: its setup syncs every page of the documentation site (from its sitemap.xml) into the agent's knowledge base. Run its setup again to pick up new pages. Any other model works too, for example one in Workspace › Models with a system prompt and your documentation attached as knowledge.
  2. Create an API key for the account it answers as. Use a dedicated user (for example tutor@yourcompany.com) rather than a person's account. Create the key in that account's Settings › Account › API keys. Creating keys needs Enable API Keys switched on in Admin Panel › Settings › General (or ENABLE_API_KEYS); the tutor itself works whether it stays on or not.
  3. Save the settings. In Admin Panel › Settings › Public Tutor, switch on Enable Public Tutor, choose the model, paste the API key and click Save. Saving lets every account read the chosen model, so the tutor's account can use it.
  4. Click Send a test message. It asks the tutor a question with the settings on screen, saved or not, and shows the answer or exactly why there is none.

The chat button and the landing page section appear while the tutor is on and its API key belongs to an account.

Limits​

  • A visitor can send 10 messages a minute (counted per IP address). Longer messages are cut at 2,000 characters.
  • Every answer uses the chosen model, and usage limits for the tutor's account apply. If the account reaches a hard limit, the tutor stops answering until the limit resets.

Troubleshooting​

When the tutor can't answer, visitors see Sorry, I could not get an answer right now and the browser console shows a 502 for /api/v1/public/tutor/chat. Click Send a test message in Admin Panel › Settings › Public Tutor to see the reason; the server log has it too (Public tutor could not answer: …).

Test resultWhat to do
The API key does not belong to any accountThe key was deleted or regenerated. Create a key and save it again.
The chat endpoint answered 400: Model not foundThe model was deleted or renamed, or the connection its base model uses is down. Choose another model, or fix the connection in Admin Panel › Settings › Connections.
The chat endpoint answered 403: …The tutor's account can't use the model. Save the settings again (this shares the model with every account), or check the model's access settings.
The chat endpoint answered 429: …The tutor's account reached a usage limit. Raise or remove the limit for that account.
The chat endpoint answered 400: 'NoneType' object is not callableYour Studio is older than this release: models with knowledge (such as the HridaDoc Agent), web search or tools failed for requests from outside a chat. Update Studio.
The model did not answer within 60 secondsThe model or its provider is slow or unreachable. Try it in a normal chat.
The chat button doesn't appearThe tutor is off, or no API key that belongs to an account is saved.
Earlier versions

Before this release, the tutor only worked while Enable API Keys was on, and not at all with a model that has knowledge, web search or tools ('NoneType' object is not callable). Both now work. The same fix applies to anyone calling /api/chat/completions with an API key.

Environment variables​

VariableDefault
ENABLE_PUBLIC_TUTORTrueTurns the tutor on or off.
PUBLIC_TUTOR_MODEL_IDhrida-tutorThe model that answers.
PUBLIC_TUTOR_API_KEY(empty)The API key of the account it answers as.

These three are ConfigVar variables: they set the first values, and after that the settings saved in the Admin Panel are used. See Public Tutor variables for HRIDAAI_INTERNAL_URL.

Hrida.ai is proprietary software of Zlabs Innovation. See the license for terms. © 2026 Zlabs Innovation.