How are AI responses generated?

Q: How are AI responses generated?

You do not need to understand this to use Questsmith.This is a technical explanation to help with Model Settings, AI experimentation, and advanced troubleshooting.Step 1: The interface sends your message to our server and waits for confirmation.Step 2: The server compiles your context (AI Instructions, Plot Components, triggered Story Cards, Memories, and recent story text including your latest input).Step 3: A tokenizer converts text into numerical tokens the model can process.Step 4: The model runs token inference in a large neural network, predicts likely next tokens, and samples output repeatedly until Output Length is reached.Step 5: Output tokens are converted back to readable text, then responses are processed and ranked; the best one is returned to your story and app.

Question

Accepted Answer

You do not need to understand this to use Questsmith.

This is a technical explanation to help with Model Settings, AI experimentation, and advanced troubleshooting.

Step 1: The interface sends your message to our server and waits for confirmation.

Step 2: The server compiles your context (AI Instructions, Plot Components, triggered Story Cards, Memories, and recent story text including your latest input).

Step 3: A tokenizer converts text into numerical tokens the model can process.

Step 4: The model runs token inference in a large neural network, predicts likely next tokens, and samples output repeatedly until Output Length is reached.

Step 5: Output tokens are converted back to readable text, then responses are processed and ranked; the best one is returned to your story and app.