Skip to contents

Pass R Output to AI for Interpretation

Source: R/pass.R
pass.Rd

Pipes any R output (test results, model summaries, tables) to an AI model for scientific interpretation. Perfect for getting publication-ready interpretations of statistical results.

Usage

pass(
  x,
  prompt = NULL,
  append_prompt = NULL,
  action = c("write", "interpret", "explain", "summarize", "critique", "suggest"),
  style = c("scientific", "simple", "detailed", "brief"),
  output = c("text", "markdown", "md", "latex", "html"),
  provider = c("openai", "anthropic", "gemini", "openrouter"),
  model = NULL,
  base_url = NULL,
  api_key = NULL,
  context = NULL,
  system_message = NULL,
  auto_local = TRUE,
  copy = FALSE,
  quiet = FALSE
)

Arguments

x

Any R object to interpret (test result, model, data frame, etc.)

prompt

Custom prompt to use. If NULL, builds from action and style. Note: This REPLACES the default prompt entirely.

append_prompt

Additional instructions to ADD to the default prompt. Unlike `prompt`, this appends to (not replaces) the auto-generated prompt.

action

What to do with the output:

  • `"write"` (default): Write publication-ready text (methods/results)

  • `"interpret"`: Interpret the statistical results

  • `"explain"`: Explain what the analysis does and means

  • `"summarize"`: Brief summary of key findings

  • `"critique"`: Critical evaluation with limitations

  • `"suggest"`: Suggest follow-up analyses

style

Interpretation style:

  • `"scientific"` (default): APA-style for academic papers

  • `"simple"`: Plain language, no jargon

  • `"detailed"`: Comprehensive with assumptions, limitations, caveats

  • `"brief"`: Just the key takeaway

output

Output format:

  • `"text"` (default): Plain text

  • `"markdown"` or `"md"`: Markdown formatted

  • `"latex"`: LaTeX formatted for papers

  • `"html"`: HTML formatted

provider

AI provider: `"openai"` (default), `"anthropic"`, `"gemini"`, or `"openrouter"`.

model

Model to use. Defaults: `"gpt-4.1-nano"` (OpenAI), `"claude-sonnet-4-20250514"` (Anthropic), `"gemini-2.5-flash"` (Gemini), `"anthropic/claude-sonnet-4"` (OpenRouter).

base_url

Custom API base URL for OpenAI-compatible servers (e.g., LM Studio, Ollama, vLLM). Example: `"http://127.0.0.1:1234"` for LM Studio. When set, uses OpenAI-compatible format regardless of provider setting.

api_key

API key. If NULL, checks environment variables (`OPENAI_API_KEY`, `ANTHROPIC_API_KEY`, `GEMINI_API_KEY`, or `OPENROUTER_API_KEY`), then prompts interactively. For local servers like LM Studio, use `api_key = "none"` or any string.

context

Optional context about your study (e.g., "This is a study on student learning outcomes with N=500 participants")

system_message

Optional custom instructions for the AI (e.g., "Focus on clinical implications", "Be more concise", "Emphasize effect sizes")

auto_local

Logical. Automatically detect and use local AI servers (LM Studio on port 1234, Ollama on port 11434)? Default: TRUE. Set to FALSE to force using cloud providers.

copy

Logical. Copy result to clipboard? Default: FALSE

quiet

Logical. Suppress messages? Default: FALSE

Value

Character string with the AI interpretation (invisibly). Also prints the interpretation.

Details

On first use, you'll be prompted to enter your API key. The key is stored in your R environment for the session. To persist it, add to your .Renviron: “` ANTHROPIC_API_KEY=your-key-here # or OPENAI_API_KEY=your-key-here “`

Examples

if (FALSE) { # \dontrun{
# Basic usage - pipe test results
t.test(mpg ~ am, data = mtcars) |> pass()

# With context
cor.test(mtcars$mpg, mtcars$hp) |>
  pass(context = "Studying fuel efficiency in 1974 automobiles")

# Different actions
lm(mpg ~ wt + hp, data = mtcars) |> summary() |> pass(action = "write")
chisq.test(mtcars$cyl, mtcars$am) |> pass(action = "explain", style = "simple")

# Get LaTeX output for paper
aov(mpg ~ factor(cyl), data = mtcars) |> summary() |>
  pass(action = "write", output = "latex")

# Critique an analysis
lm(mpg ~ ., data = mtcars) |> summary() |> pass(action = "critique")

# Custom prompt (REPLACES default - specific request to the AI)
my_results |> pass(prompt = "Focus only on the interaction effects")

# Add to default prompt (keeps "write methods/results" + your addition)
my_results |> pass(append_prompt = "Also mention limitations of the sample size")
my_results |> pass(action = "write", append_prompt = "Include a brief discussion section")

# Custom context (about your study)
t.test(score ~ group, data = mydata) |>
  pass(context = "RCT comparing drug vs placebo, N=200 patients with diabetes")

# Custom system message (instructions for the AI)
my_results |> pass(system_message = "Focus on clinical implications and effect sizes")

# Combine all customizations
lm(outcome ~ treatment * age, data = mydata) |> summary() |>
  pass(
    action = "write",
    context = "Phase 3 clinical trial for hypertension medication",
    system_message = "Emphasize clinical significance over statistical significance",
    prompt = "Pay special attention to the treatment-age interaction"
  )
} # }