alikon-art/DeterminFlow

A production-oriented AI workflow runtime for building, validating, recovering, and shipping complex AI workflows as dependable services. 面向生产的 AI 工作流运行时：快速开发、验证和恢复复杂 AI 工作流，并将其稳定交付为服务。

What it solves

DeterminFlow is designed to solve the instability, high cost, and maintenance difficulty associated with using single-agent frameworks for complex, predefined AI workflows. Unlike single agents that must remember entire histories and manage all tools, DeterminFlow breaks processes into discrete, versioned nodes to ensure reliable, predictable, and cost-effective execution.

How it works

The framework organizes LLMs, scripts, APIs, and human approvals into a structured workflow. It uses a "node-based" approach where each Agent Node has its own isolated local context and a restricted set of tools. The system freezes the workflow definition and inputs at task start, manages data flow and control flow via a runtime, and saves checkpoints for every attempt and output. This allows the system to recover from failures at the specific node where it failed rather than restarting the entire process.

Who it’s for

Developers and businesses building production-grade AI services where the process is already clearly defined and requires high reliability, auditability, and token efficiency (e.g., the provided example of AI novel production).

Highlights

  • Node-Level Isolation: Reduces token consumption by ensuring agents only read necessary local context instead of the full history.
  • Reliable Execution: Supports automatic and manual retries, skipping, and breakpoint recovery from failure nodes.
  • Structured Output: Includes JSON output detection, parsing, and automatic repair/retry mechanisms.
  • Visual Orchestration: Features a visual Workflow Editor supporting variables, conditions, parallel execution, loops, and sub-processes.
  • Plugin System: Allows packaging workflows, prompts, and scripts into plugins for easy delivery as APIs or business services.

Related

  • Project
    OpenDCAI/DataFlow

    DataFlow is an open‑source Python framework that lets you build, run, and share low‑code pipelines for generating, cleaning, and evaluating LLM training data. It provides a library of reusable operators, a visual WebUI, an AI‑assistant for auto‑creating pipelines, and a Ray‑based distributed execution layer.

  • Project
    PentesterFlow/agent

    PentesterFlow is an open‑source terminal assistant that lets a human analyst guide an LLM‑driven agent through penetration‑testing tasks. It integrates real security tools (curl, shell, Burp, file utilities), supports many LLM backends, stores evidence and findings as markdown, and retains project‑level memory across sessions while requiring explicit approval for risky actions.

  • Project
    devcodex-labs/devcodex

    DevCodex is a Node‑JS CLI that orchestrates AI‑coding assistants (Copilot, Claude Code, etc.) with an intent‑driven, project‑aware workflow. It builds a structured project Profile, loads domain‑specific Skills on demand, persists task state across sessions, and requires evidence before marking work complete. Designed for developers and teams who need reproducible, auditable AI‑generated code changes.

  • Project
    ai-sdlc-framework/ai-sdlc

    AI‑SDLC is an open‑source decision‑engine framework that orchestrates AI‑assisted software development. It front‑loads human decisions via a Definition‑of‑Ready gate, then automatically runs a 13‑step pipeline (code generation, three independent LLM reviewers, DSSE attestation, PR creation) using an autonomous orchestrator. The system includes a terminal UI, declarative governance resources, and SDKs for TypeScript, Python, and Go. Licensed under Apache 2.0, it targets enterprises and platform teams that need reproducible, governed AI‑driven code pipelines.

  • Project
    voocel/ainovel-cli

    ainovel‑cli is a command‑line/TUI tool that orchestrates multiple LLM agents (Architect, Writer, Editor) via a deterministic engine to automatically write long‑form novels (500+ chapters). It features rolling‑window planning, hierarchical summarisation, checkpoint/recovery, multi‑model support, import/export, and configurable rules, all runnable headlessly or in Docker.