geminiGrounding4R: Gemini API Google Search Grounding Request (v1/v1beta,...

In chatAI4R: Chat-Based Interactive Artificial Intelligence for R

Gemini API Google Search Grounding Request (v1/v1beta, 2025-07)

Description

A thin R wrapper for Google Gemini API with Google Search grounding enabled. Grounding improves factuality by incorporating web search results.

Usage

geminiGrounding4R(
  mode,
  contents,
  model = "gemini-2.5-flash",
  api_version = "v1beta",
  store_history = FALSE,
  dynamic_threshold = 0.7,
  api_key = Sys.getenv("GoogleGemini_API_KEY"),
  max_tokens = 2048,
  debug = FALSE,
  enable_grounding = TRUE,
  ...
)

Arguments

mode	One of '"text"', '"stream_text"', '"chat"', '"stream_chat"'.
contents	Character vector (single-turn) or list of message objects (chat modes). See Examples.
model	Gemini model ID. Default '"gemini-2.5-flash"'.
api_version	API version for the Generative Language API ('"v1"' or '"v1beta"'). Default '"v1beta"'.
store_history	Logical. If TRUE, chat history is persisted to the 'chat_history' env-var (JSON).
dynamic_threshold	Numeric [0,1] for dynamic retrieval threshold (default: 0.7). Only used for Gemini 1.5 models with 'googleSearchRetrieval'; ignored for newer models that use 'google_search'.
api_key	Your Google Gemini API key (default: 'Sys.getenv("GoogleGemini_API_KEY")').
max_tokens	Maximum output tokens. Default 2048.
debug	Logical. If TRUE, prints request details for debugging.
enable_grounding	Logical. If TRUE, enables Google Search grounding (default).
...	Additional 'httr::POST' options (timeouts etc.).

Value

For non-stream modes, a parsed list. For stream modes, a list with 'full_text' and 'chunks'.

Author(s)

Satoshi Kume (revised 2025-07-01)

Examples

## Not run: 
  # Synchronous text generation with grounding:
  result <- geminiGrounding4R(
    mode = "text",
    contents = "What is the current Google stock price?",
    store_history = FALSE,
    debug = TRUE,  # Enable debug to see request details
    api_key = Sys.getenv("GoogleGemini_API_KEY")
  )
  print(result)

  # Basic text generation without grounding (for troubleshooting):
  basic_result <- geminiGrounding4R(
    mode = "text",
    contents = "Hello, how are you?",
    enable_grounding = FALSE,
    debug = TRUE,
    api_key = Sys.getenv("GoogleGemini_API_KEY")
  )
  print(basic_result)

  # Chat mode with history storage:
  chat_history <- list(
    list(role = "user", text = "Hello"),
    list(role = "model", text = "Hi there! How can I help you?")
  )
  chat_result <- geminiGrounding4R(
    mode = "chat",
    contents = chat_history,
    store_history = TRUE,
    dynamic_threshold = 0.7,
    api_key = Sys.getenv("GoogleGemini_API_KEY")
  )
  print(chat_result)

  # Streaming text generation:
  stream_result <- geminiGrounding4R(
    mode = "stream_text",
    contents = "Tell me a story about a magic backpack.",
    store_history = FALSE,
    dynamic_threshold = 0.7,
    api_key = Sys.getenv("GoogleGemini_API_KEY")
  )
  print(stream_result$full_text)

## End(Not run)

chatAI4R documentation built on Jan. 10, 2026, 5:07 p.m.