Package index • LLMR

Configure, call, and transform data

llm_config(): Create an LLM configuration (provider-agnostic)
call_llm(): Call an LLM (chat/completions or embeddings) with optional multimodal input
llm_fn(): Apply an LLM prompt over vectors/data frames
llm_mutate(): Mutate a data frame with LLM output
finish_reason() tokens() is_truncated() as.character(<llmr_response>) print(<llmr_response>): LLMR Response Object
llm_api_key_env(): Declare an API key sourced from an environment variable
print(<llm_config>) format(<llm_config>): Print an LLM configuration with the API key masked

build_factorial_experiments(): Build Factorial Experiment Design
call_llm_par(): Parallel LLM Processing with Tibble-Based Experiments (Core Engine)
call_llm_broadcast(): Parallel API calls: Fixed Config, Multiple Messages
call_llm_sweep(): Parallel API calls: Parameter Sweep - Vary One Parameter, Fixed Message
call_llm_compare(): Parallel API calls: Multiple Configs, Fixed Message
llm_judge(): LLM-as-a-Judge Evaluation
setup_llm_parallel(): Setup Parallel Environment for LLM Processing
reset_llm_parallel(): Reset Parallel Environment

Local call logs, repeated measurements, usage summaries, failures, and study reports.

llm_log_enable() llm_log_disable() llm_log_status() llm_log_active(): Record every LLM call in a local audit log
llm_log_read(): Read an LLMR audit log into records and a manifest
llm_replicate(): Run the same prompt several times per row
llm_agreement(): Agreement across replicated LLM annotations
llm_usage(): Summarize token usage and outcomes of an LLM run
llm_failures(): List the rows of an LLM run that failed or were truncated
report(): Draft a methods-section report from an LLMR-family result object
llm_methods_text(): Deprecated methods-paragraph helper

enable_structured_output(): Enable Structured Output (Provider-Agnostic)
disable_structured_output(): Disable Structured Output (clean provider toggles)
llm_fn_structured(): Vectorized structured-output LLM
llm_mutate_structured(): Data-frame mutate with structured output
call_llm_par_structured(): Parallel experiments with structured parsing
llm_parse_structured(): Parse structured output emitted by an LLM
llm_parse_structured_col(): Parse structured fields from a column into typed vectors
llm_validate_structured_col(): Validate structured JSON objects against a JSON Schema (locally)
llm_fn_tags(): Vectorized LLM with tag extraction
llm_mutate_tags(): Data-frame mutate with XML-like tag output
call_llm_par_tags(): Parallel experiments with tag parsing
llm_parse_tags(): Parse XML-like tags emitted by an LLM
llm_parse_tags_col(): Parse XML-like tag fields from a column

chat_session() as.data.frame(<llm_chat_session>) summary(<llm_chat_session>) head(<llm_chat_session>) tail(<llm_chat_session>) print(<llm_chat_session>): Chat Session Object and Methods
llm_tool(): Define a tool the model may call
llm_tool_signature(): A stable signature for a tool's identity
call_llm_tools(): Call an LLM with tools and run the tool loop
tool_calls(): Extract tool calls from a response
call_llm_stream(): Stream a chat completion token by token
llm_logprobs(): Extract token log-probabilities from a response

A bundled example persona dataset and functions for inspecting and validating persona data frames shared across the LLMR family.

anes_2024_personas: Example participant profiles derived from ANES 2024
llm_persona_split(): Split one persona into labeled demographics and labeled answers
llm_persona_overview(): A compact overview of a persona frame for display
llm_persona_demographic_fields(): Which columns of a persona frame are demographics
llm_validate_persona_frame(): Check that a data frame follows the persona contract
transcript_as_messages(): Build a role-flipped message array from a multi-speaker transcript
ensure_alternating_messages(): Make a message array provider-safe (alternating, user-leading)

Reconstruct calls from logs, inspect request records, and use common diagnostics and reset methods for package results.

llm_request_hash(): Stable request hash for an LLM call
llm_request_from_log(): Rebuild a callable request from a logged record
llm_response_record(): Flatten one LLM response to a provenance row
llm_hash(): Content hash for research artifacts
diagnostics(): Machine-readable diagnostics for an LLMR-family result object
reset(): Reset a stateful object to its initial position