|
<< Click to Display Table of Contents >> Navigation: 3. Script Language > AI - Artificial Intelligence Commands > AIG. - Google AI > AIG. - AI Google Gemini Integration |
MiniRobotLanguage (MRL)
4. Configuration (Generation Parameters)
Fine-Tuning the AI's Output Behavior
Overview
This section controls the "Personality" and "Creativity" of the AI. By adjusting these parameters, you determine whether the AI acts like a strict, logical calculator or a creative storyteller. These settings affect every subsequent call to AIG.Ask or AIG.ChatCompletion.
Visualizing the Parameters
┌──────────────────┐
│ AI ENGINE │
└────────┬─────────┘
│
[Temperature] ───────┼─────── [Top P]
(Randomness) │ (Word Choice Pool)
0.0 = Precise │ 0.1 = Strict
1.0 = Creative │ 1.0 = Diverse
│
[MaxToken] ──────────┼─────── [Stop Sequences]
(Length Limit) │ (Forced Brake)
e.g., 100 words │ e.g., "END"
▼
┌──────────────────┐
│ FINAL RESPONSE │
└──────────────────┘
Detailed Command List
•AIG.Set Temperature / Get - Controls the randomness of the output.0.0 - 0.3: Analytical, precise, deterministic. Best for code and data extraction.
•0.7 - 1.0: Balanced, conversational.
•1.2 - 2.0: Highly creative, random, can be hallucinatory.
•AIG.Set TopP / Get - Nucleus Sampling. Changes how many possible "next words" the AI considers.Lower values restrict the vocabulary to the most likely words (more predictable).
•AIG.Set MaxToken / Get - Limits the length of the response. Useful for controlling costs and ensuring brevity.
•AIG.SetResponseFormat / Get - Switch between "text" (default) and "json_object". Use JSON mode when you need the AI to output machine-readable data for your robot to parse.
•AIG.SetStop / Get - Defines up to 5 sequences (e.g., "END") that will force the AI to stop generating immediately if encountered.
•AIG.SetSafetySettings / Get - Configures the content filters (Hate speech, Harassment, etc.). Defaults to "BLOCK_NONE" to allow unrestricted processing.