# `Jido.AI.Actions.Reasoning.Infer`
[🔗](https://github.com/agentjido/jido_ai/blob/v2.4.0/lib/jido_ai/actions/reasoning/infer.ex#L1)

A Jido.Action for drawing logical inferences from given premises.

This action uses ReqLLM with a specialized system prompt for logical reasoning,
helping to draw valid conclusions from given information.

## Parameters

* `model` (optional) - Model alias (e.g., `:reasoning`) or direct spec
* `premises` (required) - The given facts/information as premises
* `question` (required) - What to infer from the premises
* `context` (optional) - Additional background information
* `max_tokens` (optional) - Maximum tokens to generate (default: `2048`)
* `temperature` (optional) - Sampling temperature (default: `0.3`)
* `timeout` (optional) - Request timeout in milliseconds

## Examples

    # Basic inference
    {:ok, result} = Jido.Exec.run(Jido.AI.Actions.Reasoning.Infer, %{
      premises: "All cats are mammals. Fluffy is a cat.",
      question: "Is Fluffy a mammal?"
    })

    # With context
    {:ok, result} = Jido.Exec.run(Jido.AI.Actions.Reasoning.Infer, %{
      premises: "If it rains, the ground gets wet. The ground is wet.",
      question: "Can we conclude that it rained?",
      context: "Consider that sprinklers can also make the ground wet."
    })

# `__action_metadata__`

Returns the Action metadata. Alias for to_json/0.

# `category`

Returns the category of the Action.

# `description`

Returns the description of the Action.

# `name`

Returns the name of the Action.

# `on_after_run`

Lifecycle hook called after Action execution.

# `on_after_validate_output`

Lifecycle hook called after output validation.

# `on_after_validate_params`

Lifecycle hook called after parameter validation.

# `on_before_validate_output`

Lifecycle hook called before output validation.

# `on_before_validate_params`

Lifecycle hook called before parameter validation.

# `on_error`

Lifecycle hook called when an error occurs.

# `output_schema`

Returns the output schema of the Action.

# `run`

Executes the infer action.

## Returns

* `{:ok, result}` - Successful response with `result`, `reasoning`, `confidence`, `model`, and `usage` keys
* `{:error, reason}` - Error from ReqLLM or validation

## Result Format

    %{
      result: "The inferred conclusion",
      reasoning: "Step-by-step reasoning chain",
      confidence: 0.9,
      model: "anthropic:claude-sonnet-4-20250514",
      usage: %{...}
    }

# `schema`

Returns the input schema of the Action.

# `tags`

Returns the tags associated with the Action.

# `to_json`

Returns the Action metadata as a JSON-serializable map.

# `to_tool`

Converts the Action to an LLM-compatible tool format.

# `validate_output`

```elixir
@spec validate_output(map()) :: {:ok, map()} | {:error, String.t()}
```

Validates the output result for the Action.

## Examples

    iex> defmodule ExampleAction do
    ...>   use Jido.Action,
    ...>     name: "example_action",
    ...>     output_schema: [
    ...>       result: [type: :string, required: true]
    ...>     ]
    ...> end
    ...> ExampleAction.validate_output(%{result: "test", extra: "ignored"})
    {:ok, %{result: "test", extra: "ignored"}}

    iex> ExampleAction.validate_output(%{extra: "ignored"})
    {:error, "Invalid output for Action: Required key :result not found"}

# `validate_params`

```elixir
@spec validate_params(map()) :: {:ok, map()} | {:error, String.t()}
```

Validates the input parameters for the Action.

## Examples

    iex> defmodule ExampleAction do
    ...>   use Jido.Action,
    ...>     name: "example_action",
    ...>     schema: [
    ...>       input: [type: :string, required: true]
    ...>     ]
    ...> end
    ...> ExampleAction.validate_params(%{input: "test"})
    {:ok, %{input: "test"}}

    iex> ExampleAction.validate_params(%{})
    {:error, "Invalid parameters for Action: Required key :input not found"}

# `vsn`

Returns the version of the Action.

---

*Consult [api-reference.md](api-reference.md) for complete listing*