NexusFlow

English

简体中文

English

简体中文

Appearance

FAQ ​

Connection Issues ​

NexusFlow keeps showing "disconnected" after launch? ​

NexusFlow communicates with UE via a Named Pipe and supports any startup order — whether you launch NexusFlow or the UE Editor first, both sides will automatically discover each other and establish a connection.

If the connection still fails, check the following:

  • Version mismatch: The NexusFlow desktop app and UE plugin must be compatible versions. Make sure both are updated to the latest release
  • Is the NexusFlow plugin enabled in your UE project? (Edit > Plugins)
  • Could a firewall or security software be blocking communication?
  • Is another NexusFlow instance already running?

Version matching

If you recently updated the NexusFlow desktop app, update the UE plugin to the corresponding version as well. Version mismatches are a common cause of connection failures.

Do I need to restart NexusFlow after restarting UE or switching projects? ​

No. NexusFlow has a built-in auto-reconnect mechanism:

  • When UE shuts down or switches projects, the connection drops temporarily
  • NexusFlow automatically waits for UE to reconnect
  • The connection typically restores within seconds after UE finishes starting

Just restart UE — NexusFlow handles the reconnection automatically.

AI Models ​

Model validation failed after adding it? ​

Model validation sends a test request to the AI provider. Common reasons for failure:

CauseFix
Incorrect or expired API KeyDouble-check the key; confirm it's still valid
Cannot reach the API endpointCheck your network; consider whether a proxy is needed
Wrong model identifierVerify the exact spelling, e.g., gpt-5.4 not gpt54
Custom Provider misconfiguredConfirm the API URL and communication protocol match

AI response cuts off or there's no response at all? ​

If the AI stops mid-reply or never responds, possible causes include:

  • Insufficient account balance — Check your provider's billing page
  • Rate limiting — Too many requests in a short period; wait and try again
  • Unstable network — AI responses stream in real time; network hiccups can interrupt the flow
  • Max Tokens set too low — Increase the Max Tokens value (at least 4096 is recommended)

Retrying

After an interruption, simply resend your message or type "continue" to prompt the AI to pick up where it left off.

Is Max Tokens required for Anthropic (Claude) models? ​

Yes. The Anthropic Messages API requires a Max Tokens parameter — this is part of Anthropic's specification.

  • Always set Max Tokens when adding an Anthropic model
  • Recommended value: 4096 or higher
  • Omitting it will cause API calls to fail

Required field

When using the Anthropic Messages API protocol — including Custom Providers with this protocol selected — Max Tokens is mandatory. See Settings - AI Models for details.

The AI replies in the wrong language? ​

The AI response language can be configured independently of the interface language:

  • Go to Settings > Agent Configuration > Response Language
  • The default is "Follow interface language"
  • If the UI is in English but you want Chinese replies, manually select "简体中文" here

Which AI providers are supported? Can I use local models? ​

NexusFlow includes built-in support for:

  • OpenAI — GPT series
  • Anthropic — Claude series
  • Google Gemini — Gemini series
  • DeepSeek — DeepSeek series
  • Zhipu GLM — GLM series
  • MiniMax — MiniMax series
  • Qwen — Qwen series
  • Kimi — Kimi series

Through Custom Provider, you can connect any service compatible with the OpenAI / Anthropic / Gemini API, including locally hosted models:

Just enter the API URL and select the matching communication protocol in the Custom Provider form.

Blueprint Operations ​

Can I stop the AI while it's modifying a Blueprint? ​

Text generation can be stopped; tool operations cannot.

  • ✅ While the AI is writing a text reply → click the stop button to interrupt
  • ❌ While the AI is executing a tool operation (e.g., modifying a Blueprint) → you'll see a message: "Tool is executing, cannot interrupt"

You'll need to wait for the current tool operation to finish before taking further action.

UE Quick Menu doesn't respond when clicked? ​

The Quick Menu depends on a live connection between NexusFlow and UE. If nothing happens:

  1. Confirm NexusFlow is running
  2. Check whether the status bar at the bottom of the sidebar shows a 🟢 green connection indicator
  3. If disconnected, see Cannot connect to UE for troubleshooting steps

Note about the Level Editor Toolbar button: When NexusFlow is not running, clicking this button shows a notification with download links instead of opening the sidebar. This is by design — it helps new users who installed the UE plugin but haven't installed the desktop app yet.

Prerequisite

All Quick Menu AI operations (right-click menus and Blueprint toolbar buttons) require NexusFlow to be connected to UE.

The AI says it can't find a Blueprint node? ​

The AI needs up-to-date Blueprint context before it can operate. Verify the following:

  1. The target Blueprint is open — double-click the Blueprint asset in UE
  2. UE is connected — the status bar should show a green indicator

Context syncs automatically — you don't need to tell the AI which Blueprint you're editing. If the AI still can't find the node, try reopening the Blueprint or describing the target node more explicitly in the chat.

Settings & Configuration ​

Is my API Key safe? Where is it stored? ​

API Keys are encrypted and stored locally:

  • Encrypted file location: %APPDATA%/cn.nexusflow.desktop/secrets.enc
  • The settings file settings.json does not contain plaintext API Keys
  • All keys are stored locally — nothing is uploaded to any server

Where are the config files? How do I reset settings? ​

NexusFlow stores its configuration at:

%APPDATA%/cn.nexusflow.desktop/
├── settings.json    # Settings (model configs, appearance, hotkeys, etc.)
├── secrets.enc      # Encrypted API Keys
└── logs/            # Log files

To reset everything: Delete the entire %APPDATA%/cn.nexusflow.desktop/ directory and restart NexusFlow. It will restore default settings and re-launch the first-run setup wizard.

Caution

Resetting will erase all AI model configurations and API Keys. You'll need to set them up again.

Global hotkey changes don't take effect? ​

Global hotkeys must be registered with the operating system after changes. If they're not working:

  1. Check for conflicts with other applications' hotkeys
  2. Try a different key combination
  3. Restart NexusFlow to apply the new hotkeys

TIP

If hotkeys stop working entirely, go to Settings > Hotkeys and click "Reset to Default."

Performance & Compatibility ​

Operations on large Blueprints are slow? ​

When working with large Blueprints (many nodes), NexusFlow needs to:

  1. Read the Blueprint's full data
  2. Convert and send it to the AI
  3. Wait for the AI to analyze and generate instructions
  4. Execute the operations in UE

Each step takes time, and larger Blueprints naturally take longer. This is expected behavior during the Beta phase; performance will continue to improve in future releases.

Which UE versions are supported? ​

NexusFlow supports UE 5.6 and UE 5.7.

  • ✅ UE 5.7: Full support
  • ✅ UE 5.6: Full support
  • ❌ UE 5.5 and earlier: Not supported (requires ONNX Runtime 1.20+ for RAG embeddings, available in UE 5.6+)
  • ❌ UE 4.x: Not supported

Engine Version

NexusFlow requires UE 5.6 or 5.7. Earlier versions lack the ONNX Runtime version needed for the RAG embedding model. If ONNX Runtime is unavailable, RAG search will fall back to FTS5 full-text search with reduced accuracy.

Is macOS supported? ​

Currently, only Windows 10/11 is supported.

macOS support is on the roadmap, but OS-level restrictions (App Sandbox limitations on inter-process communication) require additional work to resolve.

More Help ​

If the answers above didn't resolve your issue: