Model
Choose the language model and tune the temperature, Top P and token limit.
Model
The Model tab decides which language model the agent uses, and its generation parameters.
Reach it through the Settings icon in the editor's side navigation, Model tab.
Model provider
| Option | What it is |
|---|---|
| Roberty's built-in LLM | The default. Uses the model the platform manages, with nothing else to configure |
| An external connection (coming soon) | Connects to an external provider (OpenAI, Gemini or Anthropic) through a connection configured in Roberty Studio |
Temperature
Controls how random the answers are. Range: 0 to 2.
| Value | What it does |
|---|---|
| Near 0 | More deterministic, more predictable answers |
| Near 2 | More creative, more varied answers |
Parameters for an external connection only
The fields below only appear with An external connection selected.
Connection type
The provider: OpenAI, Gemini or Anthropic.
Connection
The connection you configured in Roberty Studio for that provider.
Model
The specific model on that connection.
OpenAI: gpt-4o, gpt-4o-mini, gpt-4-turbo, gpt-3.5-turbo
Gemini: gemini-1.5-pro, gemini-1.5-flash, gemini-2.0-flash
Anthropic: claude-3-5-sonnet, claude-3-5-haiku, claude-3-opus
Top P
Nucleus sampling. Controls diversity by limiting the choice to the most likely tokens. Range: 0 to 1.
Maximum tokens
Caps the length of the generated answers. Range: 1 to 32,000.