Build a generate content request structure.

This function is exposed publicly to allow the streaming manager and other components to construct requests using the same validation and normalization logic as the main content generation functions.

Parameters

• contents - List of Content structs or strings
• opts - Options including generation_config, safety_settings, etc.

Returns

• Map representing the request structure ready for JSON encoding

@spec chat(chat_options()) :: {:ok, map()}

Start a chat session for multi-turn conversations.

All options from Gemini.options/0, plus:

• :history - Chat history as a list of Content structs

Parameters

• opts - Options including:
  • :model - Model name (default: from config)
  • :history - Chat history as list of Content structs
  • :generation_config - GenerationConfig struct
  • :safety_settings - List of SafetySetting structs
  • :system_instruction - System instruction

Examples

iex> {:ok, chat} = Gemini.Generate.chat()
iex> Gemini.Generate.send_message(chat, "Hello!")
{:ok, response, updated_chat}

@spec content(String.t() | [Gemini.Types.Content.t()], Gemini.options()) ::
  {:ok, Gemini.Types.Response.GenerateContentResponse.t()}
  | {:error, Gemini.Error.t()}

Generate content using a Gemini model.

See Gemini.options/0 for available options.

Parameters

• contents - List of Content structs or strings
• opts - Options including:
  • :model - Model name (default: from config)
  • :generation_config - GenerationConfig struct
  • :safety_settings - List of SafetySetting structs
  • :system_instruction - System instruction as Content or string
  • :tools - List of tool definitions
  • :tool_config - Tool configuration

Examples

iex> Gemini.Generate.content("Hello, world!")
{:ok, %GenerateContentResponse{candidates: [%Candidate{...}]}}

iex> contents = [Content.text("Explain quantum physics")]
iex> config = GenerationConfig.creative()
iex> Gemini.Generate.content(contents, generation_config: config)
{:ok, %GenerateContentResponse{...}}

@spec count_tokens(String.t() | [Gemini.Types.Content.t()], Gemini.options()) ::
  {:ok, Gemini.Types.Response.CountTokensResponse.t()}
  | {:error, Gemini.Error.t()}

Count tokens in the given content.

See Gemini.options/0 for available options.

Parameters

• contents - List of Content structs or strings
• opts - Options including:
  • :model - Model name (default: from config)

Examples

iex> Gemini.Generate.count_tokens("Hello, world!")
{:ok, %CountTokensResponse{total_tokens: 3}}

@spec send_message(map(), String.t() | Gemini.Types.Content.t()) ::
  {:ok, Gemini.Types.Response.GenerateContentResponse.t(), map()}
  | {:error, Gemini.Error.t()}

Send a message in a chat session.

Uses options from the chat session (see Gemini.Generate.chat_options/0).

Parameters

• chat - Chat session from chat/1
• message - Message content as string or Content struct

Returns

• {:ok, response, updated_chat} on success
• {:error, error} on failure

@spec stream_content(String.t() | [Gemini.Types.Content.t()], Gemini.options()) ::
  {:ok, [Gemini.Types.Response.GenerateContentResponse.t()]}
  | {:error, Gemini.Error.t()}

Generate content with streaming support.

See Gemini.options/0 for available options.

Returns a stream of partial responses as they become available.

Parameters

• contents - List of Content structs or strings
• opts - Same options as content/2

Examples

iex> Gemini.Generate.stream_content("Write a story")
{:ok, [%GenerateContentResponse{...}, ...]}

@spec text(String.t() | [Gemini.Types.Content.t()], Gemini.options()) ::
  {:ok, String.t()} | {:error, Gemini.Error.t()}

Generate content and return only the text from the first candidate.

See Gemini.options/0 for available options.

This is a convenience function for simple text generation.

Examples

iex> Gemini.Generate.text("What is the capital of France?")
{:ok, "The capital of France is Paris."}