FourthWayQ: What exactly are credits here ?
1) what do AI credits get me? I am slightly confused. Can you explain it a bit further? Like one credit gets me one message or one conversation or what ? could you please explain that?
2) so if suppose I'm on tier 4 then how does the combination of your credits and my API keys work? Is it that first your 3,000 AI credits will be used, which are provided by you, and once that is exhausted and time is still remaining in the month, then the credits will start consuming from my own API keys? Is it something like that or is it that I can use only one of these? Could you please elaborate?
3) Finally do you have integration with Deep Seek, I mean not through Open Router but directly through the platform Deep Seek or through Open AI compatible endpoints?
4) RAG if not optimised well can use a lot of tokens per message. How do you handle it ?
Akshay_ChatbotBuilder
Apr 28, 2026A: Hi There,
Thanks for reaching out to us with your questions and concerns. Here are our responses to your questions:
Q1. What do AI credits mean?
A1. In simple terms, 1 AI credit = 1 message. So every time your bot responds to a user, it consumes 1 credit. It’s not per conversation, it’s per message/response generated by the bot. Additionally, when you train your bot (for example, using a PDF, website URL, or other data sources), there is a one-time credit consumption per file/source. This happens only during training and is not recurring, so once your bot is trained, you won’t be charged again unless you retrain or add new data.
Q2. How do platform credits + your API keys (BYOK) work?
A2. Even if you add your own API key (BYOK), AI credits are still used on the platform. You can think of AI credits as the way usage is measured within ChatbotBuilder, they are consumed when:
A. Your bot sends messages
B. You train your bot (one time per source)
So:
A. If you use our models/credits, everything runs on our infrastructure, and credits are deducted
B. If you use your own API key, the model cost is billed by your provider, but platform credits are still consumed for running the bot on ChatbotBuilder
This ensures you continue to get full platform functionality regardless of the model/provider you choose.
Q3. DeepSeek integration?
A3. At the moment, DeepSeek is supported via OpenRouter-compatible endpoints / BYOK setup.
Q4. RAG and token usage
A4. You’re absolutely right, poorly optimized RAG can increase token usage significantly. We handle this in a few ways:
A. Retrieving only the most relevant chunks instead of sending the entire dataset
B. Applying context size limits and smart trimming
C. Using chunking strategies to keep data structured and efficient
D. Avoiding redundant context being sent repeatedly across turns
E. Optimizing prompt construction so only necessary information is passed
F. Prioritize semantic search quality, so fewer but more relevant chunks are retrieved
In case you have any further questions or concerns, you can reach out to us at [email protected], and we will be happy to assist you further.
Regards,
Akshay