EQTY LabDocs

Models and prompts

Model, with Reasoning and Thinking settings, and the System and Developer prompts.

View as Markdown

A definition names the models an agent uses, with the settings it calls them with, and the prompts it gives them.

Model

A model the agent uses, with the settings it’s called with.

Settings are an allowlist, so a client’s headers, keys or other options can’t end up in a definition. Leave a setting out (None) unless it’s set; library defaults that are set should appear as set.

Example:

python
ga.Model("openai", "gpt-4.1", temperature=0, max_tokens=1024)
ga.Model("anthropic", "claude-sonnet-5", thinking=ga.Thinking("enabled", 2048))

Attributes:

  • provider (str) – Who serves the model, such as "openai" or "anthropic".
  • model (str) – The model’s name, as the provider’s API takes it.

Reasoning

Reasoning settings, as OpenAI’s reasoning models take them.

Attributes:

  • effort (str | None) – How hard the model reasons, such as "medium".
  • summary (str | None) – Whether and how it summarises its reasoning.

Thinking

Extended-thinking settings, as Anthropic’s models take them.

Attributes:

  • type (str) – "enabled", "disabled" or "adaptive".
  • budget_tokens (int | None) – The most tokens the model may think with.

System

A system prompt: the agent’s instructions.

Example:

python
ga.System("You plan the steps needed to resolve a ticket.")
ga.System(Path("prompts/planner.md"))  # read now, as UTF-8

Developer

A developer prompt: instructions in the developer role.

Example:

python
ga.Developer(Path("prompts/house-style.md"))