Package: HydraR 0.2.4

Chi Nam Ignatius Pang

HydraR: Stateful Agentic Orchestration for Scientific Reproducibility

A high-performance framework for orchestrating complex agentic workflows in R, specifically designed for scientific reproducibility and auditability. HydraR provides a robust, state-managed engine for Directed Acyclic Graphs (DAGs) and iterative state machines, prioritizing CLI-native LLM interactions (e.g., Gemini, Claude, Copilot) and hardened state persistence (DuckDB/SQLite). It features isolated execution via Git Worktrees for parallel file-modifying tasks, autonomous quality control through integrated auditors, and human-readable visualization using Mermaid.js. Ideal for creating auditable, resumable research assistants and complex bioinformatics pipelines.

Authors:Chi Nam Ignatius Pang [aut, cre], Aidan Tay [aut], APAF Agentic Workflow [ctb]

HydraR_0.2.4.tar.gz
HydraR_0.2.4.zip(r-4.7-any)HydraR_0.2.4.zip(r-4.6-any)HydraR_0.2.4.zip(r-4.5-any)
HydraR_0.2.4.tgz(r-4.6-any)HydraR_0.2.4.tgz(r-4.5-any)
HydraR_0.2.4.tar.gz(r-4.7-any)HydraR_0.2.4.tar.gz(r-4.6-any)
HydraR_0.2.4.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
DESCRIPTION |NEWS
card.svg |card.png
HydraR/json (API)

# Install 'HydraR' in R:
install.packages('HydraR', repos = c('https://ropensci-staging.r-universe.dev', 'https://cloud.r-project.org'))

Reviews:rOpenSci Software Review #()

Bug tracker:https://github.com/apaf-bioinformatics/hydrar/issues

Pkgdown/docs site:https://apaf-bioinformatics.github.io

On CRAN:

Conda:

5.21 score 1 stars 109 scripts 60 exports 31 dependencies

Last updated from:ceb3c632cd. Checks:10 OK. Indexed: no.

TargetResultTimeFilesSyslog
linux-develOK198
pkgdown docsOK202
source / vignettesOK254
linux-releaseOK199
macos-releaseOK152
macos-oldrelOK167
windows-develOK179
windows-releaseOK185
windows-oldrelOK173
wasm-releaseOK143

Exports:AgentBashNodeAgentDAGAgentDriverAgentLLMNodeAgentLogicNodeAgentMapNodeAgentNodeAgentObserverNodeAgentPythonNodeAgentRouterNodeAgentStateAgentToolAnthropicAPIDriverAnthropicCLIDriverauto_node_factoryCheckpointercleanup_jules_branchesConflictResolverCopilotCLIDrivercreate_merge_harmonizerDriverRegistryDuckDBMessageLogDuckDBSaverextract_r_code_advancedformat_toolsetGeminiAPIDriverGeminiCLIDriverGeminiImageDriverget_agent_rolesget_default_driverget_driver_registryget_logicget_roleget_role_promptinit_bot_historyis_named_listJSONLMessageLoglist_logicload_workflowMemoryMessageLogMemorySavermermaid_to_dagMessageLogOllamaDriverOpenAIAPIDriverOpenAICodexCLIDriverparse_mermaidRDSSaverreducer_appendreducer_merge_listregister_logicregister_rolerender_workflow_fileresolve_default_driverRestrictedStateset_default_driverspawn_dagvalidate_workflow_filevalidate_workflow_fullWorktreeManager

Dependencies:askpassbase64encclicodetoolscpp11curlDBIdigestfurrrfutureglobalsgluehttr2igraphjsonlitelatticelifecyclelistenvmagrittrMatrixopensslparallellypkgconfigpurrrR6rlangsysuuidvctrswithryaml

HydraR: The Complete Guide
Rendered frommanual.Rmdusingknitr::rmarkdown

Last update: 2026-04-09
Started: 2026-03-29

Agentic Travel Planning: Sydney to Hong Kong
Introduction | Setup | Loading the Workflow (Declarative YAML) | Instantiating the DAG | Visualizing the Workflow | Conclusion
Rendered fromhong_kong_travel.Rmdusingknitr::rmarkdown

Last update: 2026-04-07
Started: 2026-03-28

State Persistence & Restart: Resilient Workflows
Persistence Concepts | Declarative Pipeline | Running the Scenario | 1. Initial Run (The Pause) | 2. The Restart (The Success) | Summary
Rendered fromstate_persistence.Rmdusingknitr::rmarkdown

Last update: 2026-04-07
Started: 2026-04-03

Parallel Benchmarking with Git Worktrees
Overview | 🏗️ Step 1: Initialize an Isolated Repository | 🧠 Step 2: Declarative Workflow Loading | 🧘 Summary
Rendered fromsorting_benchmark.Rmdusingknitr::rmarkdown

Last update: 2026-04-05
Started: 2026-03-29

Creating Custom Agent Drivers
1. The AgentDriver Base Class | 2. Testing with a Mock Driver | 3. Integration with YAML Workflows | Step 1: Define Your Workflow (extending_hydrar.yml) | Step 2: Use a Node Factory to Resolve the Driver | 4. Why Use Custom Drivers?
Rendered fromextending_hydrar.Rmdusingknitr::rmarkdown

Last update: 2026-04-03
Started: 2026-04-03

Reproducible Agentic Pipelines with targets
The "Agent-as-Target" Pattern | Example: Literature Synthesis Pipeline | Benefits of Integration | 1. Caching Expensive LLM Calls | 2. Dependency Tracking | 3. Resilience and Resumption | Conclusion
Rendered fromtargets_integration.Rmdusingknitr::rmarkdown

Last update: 2026-04-03
Started: 2026-03-29

Readme and manuals

Help Manual

Help pageTopics
Create an LLM Agent Nodeadd_llm_node
Create an R Logic Nodeadd_logic_node
Bash Execution NodeAgentBashNode
Agent Graph R6 ClassAgentDAG
Agent Driver R6 ClassAgentDriver
Agent LLM Node R6 ClassAgentLLMNode
Agent Logic Node R6 ClassAgentLogicNode
Agent Map Node R6 ClassAgentMapNode
Agent Node Base ClassAgentNode
Agent Observer Node R6 ClassAgentObserverNode
Python Execution NodeAgentPythonNode
Agent Router Node R6 ClassAgentRouterNode
Agent State R6 ClassAgentState
Agent Tool R6 ClassAgentTool
Anthropic API DriverAnthropicAPIDriver
Anthropic CLI DriverAnthropicCLIDriver
Automatic Node Factory for Mermaid-as-Sourceauto_node_factory
Checkpointer InterfaceCheckpointer
Cleanup Stale GitHub Branchescleanup_jules_branches
Git Conflict Resolver R6 ClassConflictResolver
Copilot CLI Driver R6 ClassCopilotCLIDriver
Create a Merge Harmonizer Nodecreate_merge_harmonizer
Add an LLM Agent Node directly to a DAGdag_add_llm_node
Add an R Logic Node directly to a DAGdag_add_logic_node
Create an Agent Graphdag_create
Driver Registry R6 ClassDriverRegistry
DuckDB Message Log R6 ClassDuckDBMessageLog
DuckDBSaver CheckpointerDuckDBSaver
Extract R Code from LLM Responseextract_r_code_advanced
Format Toolset for Promptformat_toolset
Gemini API Driver R6 ClassGeminiAPIDriver
Gemini CLI Driver R6 ClassGeminiCLIDriver
Gemini Image API Driver R6 ClassGeminiImageDriver
List Registered Rolesget_agent_roles
Get the Default Agent Driverget_default_driver
Global Driver Registry Accessorget_driver_registry
Get Logic Functionget_logic
Get an LLM Role (System Prompt)get_role
Get a Role-specific System Promptget_role_prompt
Initialize Bot History DuckDBinit_bot_history
Check if an object is a named listis_named_list
JSONL Message Log R6 ClassJSONLMessageLog
List Registered Componentslist_logic
Load Multi-Agent Workflow from Fileload_workflow
Memory Message Log R6 ClassMemoryMessageLog
MemorySaver CheckpointerMemorySaver
Create AgentDAG from Mermaidmermaid_to_dag
Message Log Base R6 ClassMessageLog
Ollama Driver R6 ClassOllamaDriver
OpenAI API DriverOpenAIAPIDriver
OpenAI Codex CLI Driver R6 ClassOpenAICodexCLIDriver
Parse Mermaid Flowchart Syntaxparse_mermaid
RDS File CheckpointerRDSSaver
Built-in Reducer: Appendreducer_append
Built-in Reducer: Merge Listreducer_merge_list
Register Logic Functionregister_logic
Register an LLM Role (System Prompt)register_role
Render Workflow Diagram from Filerender_workflow_file
<!- APAF Bioinformatics | factory.R | Approved | 2026-03-30 -> Resolve a Default Driver from Shorthand IDresolve_default_driver
Restricted State R6 ClassRestrictedState
Set the Default Agent Driverset_default_driver
Spawn an AgentDAG from a Workflow Objectspawn_dag
Standard Node Factory for Mermaidstandard_node_factory
Validate Workflow File Syntax and Consistencyvalidate_workflow_file
Validate HydraR Workflow Integrationvalidate_workflow_full
Git Worktree Manager R6 ClassWorktreeManager