Package index • dsprrr

Signatures & Inputs

Declare typed schemas for LLM operations

Signature() signature(): Create a Signature for LLM Operations
input(): Create an input specification for a Signature
input_string() input_number() input_boolean() input_integer() input_enum() input_array() input_object(): Create typed input helpers for common cases
signature_to_json_schema(): Convert a Signature to JSON Schema
dsp_trainset(): Create Training Data for DSPrrr

Signature Transforms

Add reasoning, chain-of-thought, and other prompt-level behaviors

with_instructions() append_instructions(): Signature Transforms for Advanced Reasoning Modules
with_reasoning(): Add Chain-of-Thought Reasoning to a Signature
without_reasoning(): Remove Chain-of-Thought from a Signature
has_reasoning(): Check if a Signature has Chain-of-Thought
chain_of_thought(): Create a Chain-of-Thought Module

Modules

Create, run, and inspect LLM modules

module(): Create an LLM Module
flex(): Optimize a Module's Implementation with Flex
run(): Execute an LLM Module
run_dataset(): Execute Module on Data
concurrency_control(): Control Batch Concurrency
predict(<Module>) predict(<PredictModule>): Predict Method for Modules (tidymodels-style)
dsp(): Declarative Structured Prediction
as_module(): Create a Module from a Chat
module_fn(): Callable Module
module_graph() named_modules() named_parameters() map_modules() replace_module() freeze_modules() is_module_frozen() set_module_lm() module_children(): Traverse and Transform Module Graphs
get_last_trace(): Get the Last DSP Trace

Pipelines

Chain modules into multi-step workflows

pipeline(): Pipeline Module for Sequential Module Composition
`%>>%`: Pipe Operator for Module Composition
step(): Create a Pipeline Step with Mappings
map_inputs(): Map Inputs for Pipeline Steps
with_inputs(): Inject Static Inputs for Pipeline Steps
select_outputs(): Select Outputs for Pipeline Steps

Reasoning & Ensemble Modules

Combine and refine outputs for improved accuracy

module-wrapper: Wrapper Modules for Advanced Reasoning Patterns
best_of_n(): Create a BestOfN Wrapper Module
refine(): Create a Refine Wrapper Module
as_reward_fn(): Convert a Metric to a Reward Function
module-multichain: MultiChainComparison Module
multi_chain_comparison(): Create a MultiChainComparison Module
module-ensemble: Ensemble Module for Combining Multiple Modules
ensemble(): Create an Ensemble Module
reduce_majority(): Majority Vote Reducer
reduce_weighted_vote(): Weighted Vote Reducer
reduce_first(): First Successful Output Reducer
reduce_best_by_metric(): Best by Metric Reducer

Agentic Modules

Tool use, code execution, RAG, and recursive language models

module-react: ReAct Module
rag_module(): Create a RAG Module
r_code_runner(): Create an R Code Runner
r-code-runner: R Code Execution Backend
mcp_repl_runner(): Posit mcp-repl Code Runner
program_of_thought(): Create a Program of Thought Module
module-program-of-thought: Program of Thought Module
code_act(): Create a CodeAct Module
module-codeact: CodeAct Module
rlm_module(): Create a Recursive Language Model (RLM) Module
rlm(): Run a Recursive Language Model in one call
module-rlm: Recursive Language Model (RLM) Module

Assertions

Validate outputs and retry with automatic backtracking

Assertion() AssertionSet() assert_output() suggest_output() assertion_set(): Assertions for Output Validation
with_assertions(): Wrap a Module with Assertions
assertion-helpers: Assertion Helper Functions
assert_length(): Assert Output Length
assert_contains(): Assert Output Contains Substring
assert_not_contains(): Assert Output Does Not Contain Substring
assert_matches(): Assert Output Matches Pattern
assert_not_matches(): Assert Output Does Not Match Pattern
assert_one_of(): Assert Output is One Of
assert_custom(): Assert Custom Condition
assert_not_empty(): Assert Output is Not Empty
assert_range(): Assert Numeric Value in Range

Evaluation & Metrics

Score module outputs and measure performance

evaluate(): Evaluate a DSPrrr module
evaluate_dsp(): Evaluate a Compiled Module
metric_contains(): Create a Contains Metric
metric_custom(): Create a Custom Metric
metric_exact_match(): Create an Exact Match Metric
metric_f1(): Create an F1 Score Metric
metric_field_match(): Create a Field Equality Metric
metric_threshold(): Create a Threshold Metric
metric_with_feedback(): Create a Metric with Textual Feedback
metric_with_trace(): Create a Trace-Aware Metric
metric_model_graded_qa() metric_model_graded_fact() metric_detect_match() metric_detect_includes() metric_detect_pattern(): Pre-built Vitals-backed Metrics

Teleprompters

Automated prompt optimization strategies

Teleprompter(): Teleprompter Base Class
LabeledFewShot(): LabeledFewShot Teleprompter
GridSearchTeleprompter: GridSearchTeleprompter
BootstrapFewShot(): BootstrapFewShot Teleprompter
BootstrapFewShotWithRandomSearch(): BootstrapFewShotWithRandomSearch Teleprompter
KNNFewShot(): KNNFewShot Teleprompter
BetterTogether(): BetterTogether Teleprompter
Omni() print(<Omni>): Omni Teleprompter
AutoResearch() print(<AutoResearch>): AutoResearch Teleprompter
MetaHarness() print(<MetaHarness>): Meta-Harness Teleprompter
SIMBA(): SIMBA Teleprompter
GEPA(): GEPA Teleprompter
MIPROv2(): MIPROv2 Teleprompter
COPRO(): COPRO Teleprompter
teleprompter-ensemble: Ensemble Teleprompter
Ensemble(): Ensemble Teleprompter
ensemble_from_programs(): Compile Programs into an Ensemble

Optimization

Compile modules and tune parameters

optimize_grid(): Grid Search Optimisation
compile(): Compile S7 Generic and Methods
compile_module(): Compile a DSPrrr Program
module_parameters(): Suggest tidymodels parameters for a module
module_trials(): Summarise optimisation trials for a module
module_metrics(): Summarise optimisation metrics per trial

Optimization Results

Inspect trials, extract best configurations, and export

optimizer-accessors: Optimizer Convenience Functions
best_params(): Extract Best Parameters from a Module
best_demos(): Extract Best Demos from a Compiled Module
module_demos_as_tibble(): Convert module demos to a tibble
apply_best_config(): Apply Best Configuration from One Module to Another
top_trials(): Get Top Performing Trials
config_diff(): Compare Module Configuration Before and After Optimization
export_module_code(): Export Module Configuration as R Code
optimization_summary(): Get Optimization Summary
print(<dsprrr_optimization_summary>): Print method for optimization summary

Optimizer Infrastructure

Low-level building blocks for custom optimizers

OptimizerControl(): Optimizer Control Parameters
optimizer_control(): Create Optimizer Control
eval_program(): Evaluate a Program on a Dataset
sample_dataset(): Sample from a Dataset Deterministically
split_dataset(): Split Dataset into Train and Validation Sets
Trial(): Trial Record
TrialLog: Trial Log
create_trial(): Create a Trial Record
complete_trial(): Complete a Trial
write_trials_jsonl(): Write Trials to JSONL File
read_trials_jsonl(): Read Trials from JSONL File
load_trial_log(): Load Trial Log from Directory

Configuration

Set up LLM providers, caching, and global options

dsp_configure(): Configure dsprrr Default Settings
dsprrr_sitrep(): dsprrr Situation Report
default-chat: Default Chat Configuration
get_default_chat(): Get the Default Chat
set_default_chat(): Set the Default Chat
clear_default_chat(): Clear Cached Default Chat
with_lm(): Execute Code with a Scoped LM Override
local_lm(): Set Local LM Override
configure_cache(): Configure dsprrr Cache
clear_cache(): Clear dsprrr Cache
cache_stats(): Get Cache Statistics

Async & Streaming

Non-blocking execution and streaming responses

async: Asynchronous Module Operations
run_async(): Run a module asynchronously
stream_async(): Stream module output asynchronously
run_stream(): Run a Module with Streaming Listeners and Status Events
stream_listener(): Create a Stream Listener for a Module Output Field

Debugging & Traces

Inspect prompts, execution history, and traces

prompt-visibility: Prompt Visibility and Inspection
get_last_prompt(): Get the Last Prompt
inspect_history(): Inspect LLM Call History
clear_prompt_history(): Clear Prompt History
export_traces(): Export Module Traces
summarize_traces(): Summarize Module Traces
clear_traces(): Clear Module Traces

Results & Printing

Extract outputs, costs, and display formatted results

accessors: Accessor Functions for DSPrrr Results
get_output(): Get output from a result
get_metadata(): Get metadata from a result
get_tokens(): Get token counts from a result
get_cost(): Get cost from a result
session_cost(): Session Cost Summary
print(<dsprrr_evaluation>): Print method for dsprrr_evaluation
print(<dsprrr_batch_result>): Print method for dsprrr_batch_result
print(<dsprrr_cost_summary>): Print method for dsprrr_cost_summary
print(<EvalResult>): Print method for EvalResult
print(<Trial>): Print method for Trial
print(<BootstrapFewShot>): Print method for BootstrapFewShot
print(<BootstrapFewShotWithRandomSearch>): Print method for BootstrapFewShotWithRandomSearch
print(<GEPA>): Print method for GEPA
AutoResearch() print(<AutoResearch>): AutoResearch Teleprompter
MetaHarness() print(<MetaHarness>): Meta-Harness Teleprompter
print(<SIMBA>): Print method for SIMBA
print(<COPRO>): Print method for COPRO

Production & Deployment

Pin configurations, validate workflows, and deploy to production

program_artifact() save_program() load_program(): Persist Complete dsprrr Programs
orchestration: Orchestration Helpers for Production Workflows
pin_module_config(): Pin a Module Configuration
restore_module_config(): Restore a Module from Pinned Configuration
pin_trace(): Pin Module Traces
pin_vitals_log(): Pin Vitals Evaluation Log
use_dsprrr_template(): Use dsprrr Workflow Templates
validate_workflow(): Validate Workflow Configuration

Vitals Integration

Bridge to the vitals evaluation framework

as_vitals_solver(): Convert a dsprrr module into a vitals solver
as_vitals_task(): Create a vitals Task from a dsprrr module
as_vitals_cost(): Convert dsprrr cost data to vitals format
as_vitals_samples(): Convert dsprrr traces to vitals samples format
as_dsprrr_traces(): Convert vitals samples to dsprrr traces format
summarize_traces_df(): Summarize a traces data frame
as_dsprrr_metric(): Adapt a vitals scorer for use as a dsprrr metric
metric_model_graded_qa() metric_model_graded_fact() metric_detect_match() metric_detect_includes() metric_detect_pattern(): Pre-built Vitals-backed Metrics

ellmer & RAG

Tool registration with ellmer and retrieval with ragnar

as_ellmer_tool(): Convert a DSPrrr Module to an ellmer Tool
register_dsprrr_tool(): Register a DSPrrr Module as a Tool in a Chat
ragnar_tool(): Create a ragnar Search Tool for ReAct Modules
create_search_tool(): Create a Semantic Search Tool from Documents
print(<ragnar_tool>): Print method for ragnar_tool

tidymodels Integration

Use LLM modules as parsnip model engines with dials parameters

llm_predict(): LLM Prediction Model Specification
fit_llm_predict(): Fit LLM Predict Model
predict_llm_class(): Predict Class Labels with LLM
predict_llm_numeric(): Predict Numeric Values with LLM
register_dsprrr_engine(): Register dsprrr Engine with parsnip
temperature(): Temperature Parameter for dials
top_p(): Top-p Parameter for dials
reasoning_effort(): Reasoning Effort Parameter for dials

Demo & Interactive Tools

Interactive demonstrations and exploration tools

run_demo(): Run the Interactive RLM Demo

Utilities

Model detection, provider defaults, and helper functions

is_reasoning_model(): Check if a model is a reasoning model
provider_defaults(): Get default parameters for a provider
eval_vignette(): Determine if vignettes should be evaluated
has_ellmer_credentials(): Check for ellmer credentials