Package 'rgemini'

Create an Array Response Template

Description

This function creates a template for an array response consisting of JSON objects. Each element (object) in the array is expected to have the provided fields.

Usage

arr_resp(...)

Arguments

...	Named elements that define the fields for each JSON object in the array.

Value

A list with class 'arr_resp' that serves as a template for an array response.

Examples

# Create an array response template with two fields.
arr_template <- arr_resp(city = "Paris", country = "France")
str(arr_template)

---

Cache a Gemini context

Description

Caches a Gemini context to the API.

Usage

gemini_cache_context(
  context,
  verbose = TRUE,
  api_key = getOption("gemini_api_key")
)

Arguments

context	A 'gemini_context' object.
verbose	Logical; if 'TRUE', prints additional logging information.
api_key	The API key for authentication (default is taken from 'getOption("gemini_api_key")').

Value

The updated 'gemini_context' object with caching details.

---

Extract content from a more detailed run_gemini response

Description

Extract content from a more detailed run_gemini response

Usage

gemini_content(res_df)

---

Create a Gemini context

Description

Initializes a Gemini context object with caching options.

Usage

gemini_context(
  prompt = NULL,
  model = "gemini-1.5-flash-001",
  media = NULL,
  ttl_sec = 10 * 60,
  role = "user",
  do_cache = TRUE,
  api_key = getOption("gemini_api_key")
)

Arguments

prompt	The input prompt for the model.
model	The model to use (default: '"gemini-1.5-flash-001"').
media	Optional media input.
ttl_sec	Time-to-live for cache in seconds (default: '60*60').
role	The role of the request (default: '"user"').
do_cache	Logical; if 'TRUE', attempts to cache the context.
api_key	The API key for authentication (default is taken from 'getOption("gemini_api_key")').

Value

A 'gemini_context' object.

---

Get remaining context cache expiration time

Description

Computes the remaining seconds until a cached context expires.

Usage

gemini_context_cache_expire_sec(context)

Arguments

context

A 'gemini_context' object.

Value

The remaining seconds until expiration, or 'NA' if not cached.

---

Delete all cached contexts

Description

Deletes all cached contexts from the Gemini API.

Usage

gemini_delete_all_context_caches(api_key = getOption("gemini_api_key"))

Arguments

api_key

The API key for authentication (default is taken from 'getOption("gemini_api_key")').

Value

The number of caches deleted.

---

Delete a cached context

Description

Deletes a cached context from the Gemini API.

Usage

gemini_delete_context_cache(
  context = NULL,
  cache_name = context$cache_name,
  api_key = getOption("gemini_api_key"),
  verbose = FALSE
)

Arguments

context	An optional context object containing the cache name.
cache_name	The name of the cache to delete (default is extracted from 'context').
api_key	The API key for authentication (default is taken from 'getOption("gemini_api_key")').
verbose	Logical; if TRUE, prints messages about the deletion process.

Value

Returns 'TRUE' if deletion is successful, 'FALSE' otherwise. If 'context' is provided, returns the updated context.

---

List all cached contexts

Description

Retrieves a list of all cached contexts from the Gemini API.

Usage

gemini_list_context_caches(api_key = getOption("gemini_api_key"))

Arguments

api_key

The API key for authentication (default is taken from 'getOption("gemini_api_key")').

Value

A data frame with cached context details, or 'NULL' if no caches are found.

---

Upload a document or image to Gemini API.

Description

Returned object or a list of such objects can then be passed to gemini_run. Media should be available for around an hour.

Usage

gemini_media_upload(
  file_path,
  mime_type = NULL,
  display_name = NULL,
  api_key = getOption("gemini_api_key"),
  verbose = FALSE
)

---

Update cached context expiration

Description

Updates the TTL of a cached Gemini context.

Usage

gemini_update_context_cache(
  context,
  ttl_sec = context$ttl_sec,
  only_if_worked = TRUE,
  only_if_expired = TRUE,
  expired_sec_margin = 10,
  verbose = TRUE,
  api_key = getOption("gemini_api_key")
)

Arguments

context	A 'gemini_context' object.
ttl_sec	New TTL value (default: existing TTL).
only_if_worked	Logical; if 'TRUE', only updates if caching previously worked.
only_if_expired	Logical; if 'TRUE', only updates if cache is expired.
expired_sec_margin	Margin in seconds before cache is considered expired (default: '10').
api_key	The API key for authentication (default is taken from 'getOption("gemini_api_key")').

Value

The updated 'gemini_context' object.

---

Infer the JSON Response Type

Description

This function inspects the provided value and returns a string representing the JSON type. It distinguishes between objects, arrays, and primitive types.

Usage

infer_response_type(x)

Arguments

x	The value to inspect.

Value

A character string: one of "null", "object", "array", "string", "integer", "number", or "boolean".

---

Create an Object Response Template

Description

This function creates a template for a single JSON object response. The response is expected to be a JSON object with the provided fields.

Usage

obj_resp(...)

Arguments

...	Named elements that define the fields for the JSON object.

Value

A list with class 'obj_resp' that serves as a template for an object response.

Examples

# Create an object response template with three fields.
obj_template <- obj_resp(city = "Paris", country = "France", population = 5.2)
str(obj_template)

---

Generate a JSON Schema from a Response Template

Description

This function recursively generates a JSON Schema based on a response template. The template must be created using either obj_resp (for a single object response) or arr_resp (for an array response). The function infers the JSON Schema by inspecting the data types of the provided fields without including the example values.

Usage

response_schema(example)

Arguments

example

A response template created by obj_resp or arr_resp.

Value

A list representing the inferred JSON Schema for the provided response template.

---

Generate Content with Gemini API

Description

Sends a text prompt (and optionally one or more media objects) to the Gemini API to generate content.

Usage

run_gemini(
  prompt,
  model = "gemini-2.0-flash",
  media = NULL,
  json_mode = !is.null(response_schema),
  response_schema = NULL,
  temperature = 0.1,
  context = NULL,
  detailed_results = FALSE,
  make_content = TRUE,
  add_prompt = FALSE,
  verbose = FALSE,
  api_key = getOption("gemini_api_key")
)

Arguments

prompt	A character string containing the text prompt to be sent to the Gemini API.
model	A character string specifying the Gemini model to use. Defaults to '"gemini-2.0-flash"'.
media	Either a single media object or a list of media objects to be included in the prompt. Call gemini_media_upload to upload the document or media.
json_mode	Logical. If TRUE, expects the response in JSON format. Defaults to !is.null(response_schema).
response_schema	Optional structured output schema for the response. Defaults to NULL. Even if you want well structures JSON output, you may not always want to provide a response schema. I experienced several cases in which the actual output becomes worse if a schema is provided.
temperature	A numeric value controlling the randomness of the output. Defaults to 0.1.
context	A context object generated by gemini_context. Allows to cache long prompt prefixes or uploaded documents.
detailed_results	Logical. If TRUE, returns detailed results of different retriveal steps. Contains also an element has_error indicating if an error occured and an err_msg. Use this if you want error handling or debugging. If FALSE just returns finally parsed content and throws an error if somewhere in the retrieval process and error occurred.
make_content	Only relevant if detailed_results=TRUE. If set FALSE don't attempt last step to generate main content. If json_mode = TRUE the last content genaration step is a source of error with models that don't reliable generate valid json.
add_prompt	Only relevant if detailed_results=TRUE. Then Logical. If TRUE, includes the prompt in the returned res_df. Defaults to FALSE.
verbose	Logical. If TRUE, prints debugging and request information. Defaults to FALSE.
api_key	A character string containing your Gemini API key. Defaults to the value obtained from getOption("gemini_api_key").

Details

The function builds a JSON payload that includes the provided text prompt as well as any media objects supplied via the media parameter. Each media object is appended to the parts of the request under the file_data key. The payload is then sent via a POST request to the Gemini API endpoint for content generation.

Value

A list containing the Gemini API response. If as_data_frame is TRUE, the response is converted to a data frame.

See Also

POST, toJSON

Examples

## Not run: 
# Example using only a text prompt:
result <- run_gemini(prompt = "Tell a joke", verbose = TRUE)


## End(Not run)

---

Set your Gemini API

Description

Set your Gemini API

Usage

set_gemini_api_key(key = NULL, file = NULL)

---