Finyuus

Product Introduction

1. Definition: Finyuus is a code-first, durable AI workflow orchestration platform. It is a technical platform that provides a domain-specific language (DSL) and a runtime environment for building, executing, and governing complex AI agentic systems.
2. Core Value Proposition: Finyuus exists to decouple AI logic from application code, enabling teams to manage AI behavior, governance, and observability as an independent, production-grade system. It solves the problem of brittle, embedded AI logic by providing durable execution, version control, and centralized governance for AI workflows.

Main Features

1. Indentation-Based DSL: The platform introduces a proprietary, indentation-based domain-specific language for defining AI workflows. This DSL provides first-class syntax for agents, tools, guards, human approvals, and nested workflows. Unlike visual builders, workflows are stored as plain text, enabling Git-based version control, code review, and diffing.
2. Durable Temporal Runtime: Finyuus executes workflows on Temporal, an open-source workflow orchestration engine. This provides enterprise-grade features including automatic retries, reliable cancellation, long-running execution (hours/days), replayability for debugging, and robust child workflow orchestration, ensuring AI processes are stateful and fault-tolerant.
3. Integrated Governance & Observability: The platform bakes governance into the workflow language with native guard and approval statements. It integrates with Langfuse for detailed tracing of LLM calls, token usage, and costs. All runs are versioned, observable via a central dashboard, and auditable, providing a complete chain of custody for AI decisions.

Problems Solved

1. Pain Point: AI logic becomes tightly coupled and scattered across application codebases, making systems difficult to test, modify, and audit. Prototype-grade prompt chains lack production durability and operational controls.
2. Target Audience: Engineering teams building production AI applications, ML/AI engineers, platform teams needing to standardize AI orchestration, and compliance officers requiring audit trails for AI-driven processes.
3. Use Cases: Automating complex customer support escalations with human-in-the-loop approvals, running multi-step financial analysis with data validation guards, orchestrating content moderation pipelines, and building durable RAG (Retrieval-Augmented Generation) agent systems with structured outputs.

Unique Advantages

1. Differentiation: Unlike visual low-code AI builders (e.g., Zapier, n8n), Finyuus offers a developer-centric, code-first approach optimized for review and reuse. Unlike simple prompt chaining libraries, it provides a full-stack, durable runtime with built-in governance, unlike frameworks that require assembling disparate observability and orchestration tools manually.
2. Key Innovation: The synthesis of a purpose-built AI DSL with the industrial-grade durability of Temporal. This combination allows developers to define complex, governed AI logic in a readable, versionable format while relying on a battle-tested engine for state management, making AI workflows as reliable as traditional backend services.

Frequently Asked Questions (FAQ)

1. What is the difference between Finyuus and LangChain? LangChain is a framework for composing LLM components primarily within application code. Finyuus is a full platform that externalizes workflow definitions into a dedicated DSL and provides a separate runtime with durable execution, a versioned object registry, a dashboard, and built-in governance features like guards and approvals.
2. Can I use Finyuus for non-AI workflows? While optimized for AI and LLM-centric tasks, its Temporal foundation and DSL support HTTP calls, conditionals, and data operations, making it suitable for any workflow requiring durability, human approval steps, and complex branching. However, its primary value is in orchestrating agents, tools, and model calls.
3. How does Finyuus handle versioning and deployments? All core entities—agents, tools, guards, and pipelines—are versioned within the platform. Workflows (pipelines) are defined in text files, enabling integration with standard Git workflows for peer review and CI/CD. You can pin specific versions of agents or tools within a pipeline definition.
4. What kind of observability does Finyuus provide? Every workflow run generates a detailed trace in Langfuse, capturing LLM calls, tool executions, token usage, and costs. The Finyuus dashboard provides a run history, visual workflow execution graphs, guard/approval outcomes, and the ability to replay past runs for debugging.
5. Is Finyuus open source? The repository (mariusndini/Finyuus) on GitHub is publicly accessible, allowing developers to view the source code, run it locally, and potentially contribute. The licensing and commercial distribution model should be verified from the official repository documentation.