Unusual questions
Odd and edge-case questions about the Model API.
- Can I call the Model API from a spreadsheet or no-code tool?If the tool can send an HTTPS POST with custom headers, yes. Keep the key out of shared sheets.FAQ
- Can I create one API key per customer?An account can hold up to 25 keys, so per-customer keys only work for a few customers. Track customers in your own logs.FAQ
- Can I use the Model API in a game or chat bot (like a Discord bot)?Yes, from your server. Your account is responsible for players' or members' text, so moderate input.FAQ
- Can my mobile app use SuperCosmos?Yes, through your own server: the app calls your backend, which calls the Model API with the key.FAQ
- Does X-Clatter-Account have to be uppercase?No. Account ID matching is case-insensitive, so cl-7k2m9qxa works like CL-7K2M9QXA.FAQ
- If I retry a request, am I charged twice?Each completed request is billed from its own usage, so a retry that completes is billed again.FAQ
- If I send the same prompt and seed in parallel, will I get identical answers?Often similar, but not guaranteed: seed gives best-effort repeatability only.FAQ
- What happens if I send more than 64 tools?The limit is 64 function tools per request. Choose the relevant tools for each request in your code.FAQ
- What happens if I send more than 800 messages?The request is refused with 413 request_too_large. Trim or summarise older turns.FAQ
- What happens if I send temperature 3 or top_p 0?Out-of-range values are refused as invalid requests. temperature is 0 to 2; top_p is above 0 up to 1.FAQ
- What happens if my request is over 3 MB?413 request_too_large. Send less text, or split the work across requests.FAQ
- What happens to my API keys if I delete my Clatter Labs account?Your keys stop working with your account. See the account help for what deletion removes.FAQ
- What if a tool message has the wrong tool_call_id?The tool result must answer a tool call from the previous assistant message; mismatched ids make the request invalid.FAQ
- What if I call /v1/embeddings or /v1/responses?You get 404 not_supported. The Model API offers only GET /v1/models and POST /v1/chat/completions.FAQ
- What if I forget Content-Type: application/json?Send the JSON body with Content-Type: application/json. Without it the body may not be read as JSON.FAQ
- What if I set max_tokens higher than the model allows?max_tokens is capped per model: 32,768 output tokens on Flash and Max, 49,152 on Heavy.FAQ
- What if I use a colleague's key with my own Account ID?It fails with 401 account_mismatch. A key only works with the Account ID that owns it.FAQ
- What if I use the same key on several servers?It works, but all servers share that key's per-minute limit and log trail. Separate keys are easier to manage.FAQ
- What if my users paste logs or messages that contain swearing?Quoted text (code, logs, pasted errors, quotes) doesn't count as abuse. Only insults at the assistant do.FAQ
- Will building a profanity filter with SuperCosmos get me a warning?No. Asking about a word, translating it or building a filter for it doesn't count under the Abuse policy.FAQ
Still need help?
Tried the steps above and still stuck? Email us. We usually reply within the first hour, and always within 24 hours. A full support team is with you.
Infrastructure questions, or permanent deletion of your data
infrastructure@clatterlabs.xyzFor deleting your whole account or the data of one service.