Ironclad/rivet

Features

• AI Agent Orchestrators - Provides a visual node-based interface for designing complex prompt chains and coordinating AI agent behaviors.
• AI Agent Orchestration Frameworks - Serves as a runtime for managing autonomous agent loops and tool discovery within durable logic pipelines.
• Node-Based Execution Graphs - Processes logic as a directed graph where functional nodes process data and edges define the information flow.
• Node-Based Logic Composition - Uses a node-based visual editor to map data flow and logic between language models, vector databases, and APIs.
• Agentic RAG Development - Offers a visual system for building intelligent, self-correcting retrieval augmented generation pipelines.
• Agentic Execution Loops - Orchestrates multi-turn conversational workflows by piping results back into the chain to create feedback loops.
• Input Definitions - Allows the definition of arguments and default values passed into a logic graph from external code.
• Conditional Execution Flows - Implements conditional logic that can prevent downstream nodes in a prompt chain from executing based on specific predicates.
• External Service Integrations - Connects to external model providers, vector databases, and transcription services to manage data within agent environments.
• Application Logic Integrations - Allows running predefined AI workflows within a TypeScript environment to exchange data between graph logic and application code.
• Prompt Chaining - Implements sequential chaining where the output of one processing node serves as the input for the next step.
• Nested Execution Flows - Enables the organization of prompt chains into modular graphs that can be called by other graphs for hierarchical execution.
• Retrieval Augmented Generation - Integrates vector database searches with language models to provide grounded context for AI responses.
• Large Language Model Integration - Integrates external large language models into prompt chains and agent workflows to process text and image inputs.
• LLM Application Development - Provides a TypeScript library to embed visual AI logic graphs and prompt execution into JavaScript applications.
• LLM TypeScript SDKs - Ships a TypeScript library for embedding visual AI graphs and executing prompt logic within applications.
• Prompt Chaining Frameworks - Orchestrates sequences of nodes to send inputs to language models and retrieve responses in a multi-step chain.
• Prompt Iteration Workflows - Provides an interface for iteratively adjusting and testing prompt templates to improve the quality of AI outputs.
• Visual LLM Pipeline Designers - Provides a graphical tool for connecting language models and data sources into structured conversational pipelines.
• Prompt Engineering Tools - Provides an interface for testing, evaluating, and iteratively refining prompt templates and model outputs.
• AI Application Debugging - Ships tools for monitoring real-time execution and replaying recorded interactions to troubleshoot agent logic.
• Prompt Execution Tracing - Provides the ability to monitor and trace the internal execution flow of LLM prompt chains in real-time.
• Agent Workflow Monitoring - Provides real-time tracking and inspection of AI agent execution states and input data.
• Vector Nearest Neighbor Search - Implements vector nearest neighbor search to support retrieval-augmented generation pipelines.
• Deterministic Record and Replay Frameworks - Captures the full sequence of node inputs and outputs to enable deterministic replay and debugging of agent logic.
• Agentic Loop Controllers - Implements a controller that runs iterative cycles of nodes to update agent state until specific conditions are met.
• Visual Prompt Designers - Provides a graphical interface for refining and testing prompt templates with real-time input and output previews.
• Agent Node Data Passing - Implements a system for passing primitives, media, and vectors between different nodes in an AI orchestration graph.
• AI Workflow Routing - Directs AI execution flow by matching extracted data against predefined patterns or conditional logic.
• JavaScript Logic Runtimes - Enables the execution of visual prompt chains and agent logic within any JavaScript or server-side environment.
• Loop Control Flow - Includes a controller node to manage iterative execution cycles and termination conditions for complex AI tasks.
• Fallback Value Logic - Provides mechanisms to intercept excluded signals and provide fallback values to ensure robust agent execution.
• Parallel Execution - Allows a single node to be executed multiple times in parallel by mapping across an input array.
• Visual Runtime Tracking - Features real-time visual highlighting of active nodes and graphs to identify running processes during execution.
• Graph Execution Recording - Captures the full state and sequence of node outputs during a run for later replay and debugging.
• Reusable Graph Components - Allows the encapsulation of node collections into standalone graphs that can be embedded as single nodes in other graphs.
• Subgraph Encapsulation - Groups collections of nodes into reusable components with explicitly defined input and output ports.
• Agent Server Hosting - Allows serving a visual agent project via a local server or container for external application integration.
• Loop Termination Controls - Manages iterative execution cycles by routing data back to a controller node until a break condition is met.
• Audio Transcription - Converts audio files into text transcripts by connecting to external speech-to-text services.
• Output Definitions - Provides a system to specify the exact return values a logic graph provides to calling code or parent graphs.
• Speech Reasoning Processors - Analyzes speech transcripts using large language models to perform reasoning and structured data extraction.
• LLM Performance Evaluators - Runs predefined test cases through workflows to measure performance against evaluation benchmarks.
• Prompt Engineering Workflows - Includes specialized interfaces for iteratively refining prompt templates and performing A/B testing on model outputs.
• Remote Graph Execution - Provides a system to run local logic graphs on a remote server for deployment-free testing.
• Conversation Debugging - Visualizes the history and real-time progress of chat interactions to troubleshoot AI agent behavior.
• Prompt Variant Experiments - Provides specialized nodes for A/B testing different prompt versions to enhance agent response quality.
• Data Joins - Provides capabilities to combine parallel data arrays into single values through concatenation or custom logic.
• Remote Debugging - Allows developers to connect to and inspect the state of AI graphs running on remote production servers.
• Prompt Chain Debugging - Ships a tool to observe real-time execution of LLM workflows in running applications to identify issues.
• Live App Instance Inspections - Allows connecting to a running graph instance to view its state and troubleshoot execution flow.
• Runtime Process Controls - Provides tools for pausing or aborting active local processes to freeze state for examination.
• Regex-Based Output Parsing - Ships utilities to extract structured data from AI model responses using regular expressions and JSON.
• Graph State Synchronizers - Synchronizes the real-time state of a visual execution graph between a remote server and a local editor for control and inspection.
• AI Model Output Validators - Compares actual prompt chain outputs against expected results using a secondary validator graph.
• Version Control Workflows - Uses YAML serialization of logic graphs to enable standard version control workflows and collaborative code reviews.
• Node Type Definitions - Supports the definition of custom processing blocks with specific inputs, outputs, and execution logic.
• Graph-Based Logic Triggering - Triggers the execution of predefined logic graphs from project files to activate AI agents within an application.
• Nested Parallel Workflow Loops - Supports wrapping parallel processing logic within subgraphs to create multi-level nested concurrent workflows.
• Logic Graph Serializers - Stores visual graph configurations and node states as YAML files to enable version control and portability.
• Extensible Plugin Architectures - Implements a framework that supports modular extension and custom logic injection via plugin definitions.
• Agent Conversation Streaming - Streams live input and output for active chat nodes to track agent execution in real time.
• Plugin-Based Extensibility - Allows adding new processing blocks and menu actions through an external plugin initialization function and definition.
• Input-Driven Parallelization - Triggers multiple concurrent node instances by mapping a list of input values across a single node definition.
• Agent State Debugging - Provides tools for inspecting the runtime internal state and execution profiles of agents to identify logic issues.
• Agent Execution Replays - Provides visualizations of AI agent prompt and tool call sequences to step through recorded interactions.
• Agent State Inspections - Provides read-only access to the internal runtime state of AI agents to verify generation results.
• Agent Logic Testing - Verifies agent logic by executing repeatable test suites and validation graphs.
• Automated Test Suites - Provides the ability to execute sets of automated test cases to verify agent chain reliability.
• Remote Graph Execution Control - Enables pausing or aborting AI logic graphs running on remote servers through a visual interface.
• Model Evaluation - Measures the quality and accuracy of language model outputs through a set of predefined prompts.
• Node Output Inspection - Visualizes and analyzes data produced by individual nodes in a logic graph, including loop iterations.
• Agent Testing Suites - Ships automated testing frameworks specifically designed for validating agentic workflows and prompt chain behavior.
• Stochastic Pass-Rate Testing - Implements stochastic pass-rate testing by executing test cases multiple times to account for model randomness.
• Agent Output Verifications - Verifies agent logic by comparing graph outputs against expected results using a custom validator.
• Application Frameworks - IDE for creating complex AI agents and prompt chaining.
• Developer Utilities - Visual programming environment for AI agents.

Star history