Package {agenticr}

AI-Powered R Console Assistant

Description

An AI-powered assistant for the R console. Type natural language or incorrect R code directly in the console, and agenticr routes it to an LLM agent for processing.

Author(s)

Maintainer: Lanfeng Pan panlanfeng@gmail.com

Examples

agentic_providers()
agentic_skills()

Start an AgenticR session

Description

Start an interactive AI-powered R console session. Type R code for direct execution, natural language for LLM processing, or slash commands for meta-operations. Press Ctrl+C twice to exit.

Usage

agentic(auto = TRUE, ...)

Arguments

Details

Valid R code executes directly with zero LLM overhead. Natural language queries are routed to the configured LLM provider, which generates and executes R code in the current session.

The ; prefix forces natural language mode for any input (e.g. ;hist(mtcars[[\"mpg\"]]) sends it to the LLM instead of executing directly).

Value

Invisible. Called for its side effects.

Examples

## Not run: agentic()

Chat with the agent

Description

Send a single natural language query to the LLM agent and get a response. Useful for programmatic use or one-off queries without entering the interactive REPL.

Usage

agentic_chat(query)

Arguments

Value

Invisible. Agent output is printed to the console.

Examples

## Not run: agentic_chat("what is the mean of mpg?")

Read and manage configuration

Description

Read, update, and persist agenticr configuration. Switching providers auto-loads the provider's defaults (model, base URL, context window).

Usage

agentic_config(..., save = FALSE)

Arguments

Details

reasoning_effort controls DeepSeek's thinking depth. Valid values: "minimal", "low", "medium", "high". Default is "medium".

When switching provider (e.g. provider = "openai"), the preset model and base URL are auto-loaded. Individual fields can be overridden.

Value

Invisibly returns the updated config list.

Examples

cfg <- agentic_config()
cat("Default provider:", cfg$provider, "\n")

Disable error interceptor

Description

See package documentation for details.

Usage

agentic_disable()

Value

No return value, called for side effects.

Examples

agentic_enable()
agentic_disable()

Enable error interceptor

Description

See package documentation for details.

Usage

agentic_enable(auto_process = TRUE)

Arguments

Value

No return value, called for side effects.

Examples

agentic_enable()
# Normal R code runs directly without involving the LLM
mean(1:10)
agentic_disable()

Install a skill from a URL

Description

Downloads a SKILL.md file from a URL and installs it to ⁠agenticr data directory⁠. Skills are opt-in prompt templates that modify agent behavior.

Usage

agentic_install_skill(url, name = NULL)

Arguments

Value

Invisible.

Examples

## Not run: agentic_install_skill("https://raw.githubusercontent.com/user/skill/main/SKILL.md")

List connected MCP servers

Description

See package documentation for details.

Usage

agentic_mcp()

Value

No return value, called for side effects.

Examples

agentic_mcp()

Add an MCP server

Description

Connect to an external Model Context Protocol (MCP) server via JSON-RPC over stdio. Tools from the server are merged into the agent's tool list.

Usage

agentic_mcp_add(name, command, args = character(0), env = list(), save = FALSE)

Arguments

Value

Invisible.

Examples

## Not run: agentic_mcp_add("search", "npx", args = c("-y", "mcp-server"))

Process a single natural language query

Description

Internal function that processes a query through the agent loop. This is what the error interceptor calls when it detects natural language input.

Usage

agentic_process(query)

Arguments

Value

Invisible. Results are printed to the console.

Examples

## Not run: agentic_process("what is 2+2?")

List available provider presets

Description

See package documentation for details.

Usage

agentic_providers()

Value

No return value, called for side effects.

Examples

agentic_providers()

Resume a previous agentic session

Description

Resume a previous agentic session from its turns.jsonl log file. Restores conversation history so the agent has full context of the prior session.

Usage

agentic_resume(session_id, ...)

Arguments

Value

Invisible. Starts the REPL with prior conversation restored.

Examples

## Not run: 
agentic_sessions()
agentic_resume("20250515_120000_a1b2c3d4")

## End(Not run)

List available sessions for resuming

Description

See package documentation for details.

Usage

agentic_sessions()

Value

No return value, called for side effects.

Examples

agentic_sessions()

Interactive setup wizard

Description

Walks the user through configuring an API key and provider interactively. Creates ⁠agenticr config file⁠ if it does not exist.

Usage

agentic_setup()

Details

The wizard prompts for API key, provider, and optional model override. Configuration is saved to ⁠agenticr config file⁠ and can be edited manually later.

Value

Invisible.

Examples

## Not run: agentic_setup()

List installed skills

Description

See package documentation for details.

Usage

agentic_skills()

Value

No return value, called for side effects.

Examples

agentic_skills()

Print agenticr configuration

Description

Pretty-prints the current agenticr configuration using cli formatting. Shows provider, model, temperature, max tokens, and API key status (masked).

Usage

## S3 method for class 'agenticr_config'
print(x, ...)

Arguments

Value

Invisibly returns x.

Examples

## Not run: agentic_config()