Graph workflows — what advanced features do you need?

[AmitoVrito]

Our graph engine supports: parallel execution, conditional routing, cycles, checkpointing, Mermaid export.

What's next? Help us prioritize:

• Human-in-the-loop (interrupt, get approval, resume)
• Subgraphs (nested graph execution)
• Token-level streaming from LLM nodes
• Typed state with reducers (no more dict.update merge conflicts)
• SSE/WebSocket streaming for web apps
• Graph versioning and migration
• Execution visualization (replay with state at each step)

Which of these would unblock your use case?