goalflow

name	goalflow
description	Route goalflow (wanmol/goal-flow) work — a LangGraph framework that combines workflow graphs with agent loops — into exactly one mode: fit check, transpiling a Dify DSL export into runnable LangGraph Python, authoring workflow nodes and edges, building an `agent_kit` loop with middleware and a harness, wiring the serving layer (data adapters, SSE streaming, HITL, Redis/MySQL, API-key registration), or running the pre-publish security gate. Use when the user wants Dify's visual design without Dify's runtime, a graph node that hosts an agent loop, an OpenAI-compatible wire protocol over their own workflows, or prompt-injected `SKILL.md` capabilities. Triggers on: goalflow, goal-flow, dify to langgraph, dify transpiler, dify DSL export, BaseWorkflow, agent_kit, AgentBaseNode, DataAdapter, chunk processor, HITL interrupt, dify2langgraph. Route plain graph-API questions to `langgraph-fundamentals` and `langgraph-workflow`.
allowed-tools	Bash Read Write Edit Glob Grep
compatibility	Python 3.12, plus Redis (cache, conversation variables, stop flags) and MySQL (durable messages, HITL reviews, and the LangGraph checkpointer). `src/` layout with two packages: `goalflow` and the vendored `agent_kit`. MIT. Defaults lean toward Qwen/DashScope and Alibaba Cloud OSS because the project was extracted from an internal production system.
metadata	{"tags":"goalflow, goal-flow, langgraph, dify, dify2langgraph, transpiler, agent-loop, agent-kit, middleware, hitl, sse-streaming, data-adapter, llmops, python","platforms":"Claude, ChatGPT, Gemini, Codex","version":"1.0","source":"https://github.com/wanmol/goal-flow"}

goalflow — Graph-Orchestrated Agent Loop

goalflow gives you two ways to build LLM applications on LangGraph and lets you combine them: visual-first (design in Dify, transpile the exported DSL into a LangGraph Python file you own) and code-first (ReAct/Deep agent loops via the vendored agent_kit). A graph node can host an agent loop, and an agent can call sub-workflows as tools.

Two things shape almost every answer about this project:

1. It is extracted from an internal production system. The generalizable core is real; the defaults are opinionated (Qwen/DashScope, Alibaba OSS).
2. Its own docs open with a CAUTION: live credentials remain in git history. That is a blocker before any public push, not a footnote.

When to use this skill

• Converting a Dify flow into version-controlled LangGraph code without staying on Dify's runtime
• Choosing or authoring workflow nodes, edges, branch routing, or iteration/loop subgraphs
• Building an agent loop with agent_kit — graph builder, middleware chain, harness/model router, executable skills
• Wiring the serving layer: data adapters, SSE streaming, HITL interrupts, Redis/MySQL, and workflow registration
• Authoring SKILL.md capabilities that get LLM-matched and injected into prompts
• Auditing the repo before publishing it or deploying it anywhere real

When not to use this skill

• Plain LangGraph graph/state/checkpointer API questions → langgraph-fundamentals, langgraph-workflow, langgraph-persistence
• Generic LangChain chains, retrievers, or RAG design → langchain-fundamentals, langchain-rag
• Generic HITL patterns not tied to this engine → langgraph-human-in-the-loop
• deepagents used directly, outside agent_kit's wrapper → deepagents
• Authoring skills for this repo's agent catalog rather than goalflow's runtime → skill-standardization, write-a-skill
• LLM tracing/eval platform selection → langsmith, opik

Instructions

Step 1: Capture the intake packet

Four facts decide the mode: