7x24hrs What you need
- A website set up with us, on a paid plan (Basic, Standard, Elite or Enterprise). Sign up with your e-mail and enter your website's address if you have not yet.
- An API key: in your dashboard, open Chat apps & API, give the key a name (for example “AstrBot”) and press Create API key. Copy it at once: it is shown only that one time.
- AstrBot (or another tool) running and connected to your chat app. AstrBot's own documentation explains how to install it and connect Discord, Telegram, Slack, WeChat or QQ.
These are the three values every tool asks for:
API Base URL: https://7x24hrs.com/v1
API Key: sk-7x24-…
Model: 7x24hrs-assistantSet it up in AstrBot
- Open the AstrBot dashboard and go to Providers. On the Chat Completion tab, press + Add next to Provider Sources.
- In New model provider, choose OpenAI Compatible.
- Give it an ID you will recognise, such as
7x24hrs. Paste your key into API Key, and set API Base URL tohttps://7x24hrs.com/v1. - Press Save And Fetch Models. Under Available Models, press + next to
7x24hrs-assistant. In the window that opens, leave only Text ticked under Model capabilities (untick Image, Audio and Tool use: the assistant answers text) and press Save. - Make it the model your bot uses: in Config, choose it as the default chat model (or choose it for one conversation with the model switch in Chat).
- Ask a question in AstrBot's Chat page, or in your connected chat app. The answer comes from your website's data.
Leave AstrBot's own persona and system prompt empty, or keep them short: your assistant follows your rules and character from the 7x24hrs dashboard and does not take instructions from the chat app's system prompt.
Dify, LobeChat and other tools
Any tool with an “OpenAI-compatible” or “custom OpenAI” provider works the same way: Base URL https://7x24hrs.com/v1, your API key, and the model 7x24hrs-assistant.
- Dify: in Settings → Model Provider, add OpenAI-API-compatible, enter the model name, your key and the Base URL as the API endpoint, then use the model in your app.
- LobeChat: in the settings, under language models, turn on the OpenAI provider, enter your key and set the API proxy address to the Base URL, then add
7x24hrs-assistantto the model list. - Your own code: call the API with any OpenAI client library, setting the base URL and key.
To try the API from a terminal:
curl https://7x24hrs.com/v1/chat/completions \
-H "Authorization: Bearer sk-7x24-…" \
-H "Content-Type: application/json" \
-d '{"model": "7x24hrs-assistant", "user": "visitor-42",
"messages": [{"role": "user", "content": "Do you ship to Canada?"}]}'With "stream": true the answer arrives as server-sent events, the same way as OpenAI's:
curl -N https://7x24hrs.com/v1/chat/completions \
-H "Authorization: Bearer sk-7x24-…" \
-H "Content-Type: application/json" \
-d '{"model": "7x24hrs-assistant", "stream": true,
"messages": [{"role": "user", "content": "Do you ship to Canada?"}]}'How answers in chat apps work
| Chat button on your website | Chat app through the API | |
|---|---|---|
| Answers from | Your website's data, standard answers and character | The same |
| When it is not sure | Says so and passes the visitor to you | Says so and gives your contact address in the answer |
| You are notified | By e-mail, Slack, Discord or Telegram | The same, marked “via API / AstrBot” |
| Answering the visitor yourself | From your chat app or dashboard | Not possible from here: reply by e-mail |
| Conversations counted | One per visitor and conversation | The same, from the same monthly allowance |
Which plans include chat apps?
Every paid plan: Basic $19 a month for 90 conversations, Standard $59 for 275, Elite $199 for 1,000 and Enterprise $1,099 for 6,000. Conversations through the API come out of the same monthly allowance as the chat button. The free plan does not include chat apps: keys made on it are paused until you choose a paid plan. See pricing.
What counts as a conversation?
One visitor chatting, however many questions they ask, exactly as with the chat button. Tools that send a user field with each request let us tell visitors apart; otherwise a conversation is recognised by its first message.
Can a system prompt in AstrBot change what the assistant says?
No. System prompts sent by the chat app are not followed. The assistant keeps your rules: it answers only from your website's data, says when it does not know, and promises nothing your website does not say. Its character is the one you set in the dashboard.
Are pictures, voice messages or tool calls supported?
Not yet. Text is answered; pictures and other attachments are left out of a message, and a message with no text gets a clear error. Tool definitions sent by the client are ignored.
What if a key leaks?
Revoke it in Chat apps & API: it stops working at once. Then create a new one and paste it into your tool. Keys are stored only as a fingerprint, so we cannot show a key again either.