QwenLM Behavior

This page covers the toggles and options that control how IntenseRP interacts with QwenLM (chat.qwen.ai).

Modes (model IDs)

In IntenseRP Next v2, the model you select in SillyTavern is mostly a mode selector, not a true model picker.

For QwenLM, these model IDs map to simple behavior presets:

Real Qwen model selection (web UI)

IntenseRP can also switch QwenLM's real model picker in the web UI:

Settings -> Provider Behavior -> QwenLM -> Model

The list is intentionally "what Qwen shows in the dropdown". If your selected model is missing (UI rollout / region / UI change), IntenseRP logs a warning and keeps going.

Thinking

QwenLM exposes three thinking modes in the web UI: Auto, Thinking, and Fast.

IntenseRP keeps it simple:

• Enable Thinking -> selects Thinking
• Thinking disabled -> selects Fast

Enable Thinking

Settings -> Provider Behavior -> QwenLM -> Enable Thinking

Send Thinking

When enabled, IntenseRP includes QwenLM's thinking summaries in the response, wrapped in <think> tags.

Settings -> Provider Behavior -> QwenLM -> Send Thinking

Search

QwenLM Web search can be toggled via IntenseRP.

Qwen streams tool/search payloads into the same response stream. For stability, IntenseRP strips these tool payloads, so search results are not sent to the client.

Settings -> Provider Behavior -> QwenLM -> Enable Search

Tools

QwenLM's Tools switch lives in the same + menu as Search and Upload.

IntenseRP keeps Tools off by default. If Enable Tools is turned on, the driver will try to flip Qwen's Tools switch on before sending. If Qwen shows the switch as disabled, IntenseRP stops trying for that request and leaves it off.

Settings -> Provider Behavior -> QwenLM -> Enable Tools

Count Tokens

QwenLM reports token usage during the response stream. When enabled, IntenseRP captures those values and returns them in the OpenAI-style usage fields (prompt_tokens, completion_tokens, total_tokens). This is enabled by default.

Settings -> Provider Behavior -> QwenLM -> Count Tokens

File Upload Mode

Instead of typing your message into QwenLM's chat box, IntenseRP can upload it as a text file attachment. This is useful for very long prompts that might hit input limits.

Settings -> Provider Behavior -> QwenLM -> Send As Text File

Text File Message

Optional text that IntenseRP pastes into QwenLM alongside the uploaded file.

Leave it empty to send a file-only message.

Settings -> Provider Behavior -> QwenLM -> Text File Message

File Upload Timeout

After uploading, the send button can take a moment to become available. This setting controls how long IntenseRP waits (in seconds) before giving up.

Settings -> Provider Behavior -> QwenLM -> File Upload Timeout

Message Send Timeout

QwenLM doesn't render the send button until there's something to send (typed text or an uploaded file). This timeout controls how long IntenseRP waits for the send button to appear in normal (non-file) mode.

Settings -> Provider Behavior -> QwenLM -> Message Send Timeout (s)

Qwen Quirks & Timing

QwenLM's web UI can sometimes show that it's sending before the actual completion request appears on the network. When that happens too slowly, IntenseRP may think the click was swallowed even though Qwen is still waking up.

Completion Request Timeout

How long (in seconds) IntenseRP waits after clicking Send or Regenerate for QwenLM's completion request to appear.

Settings -> Provider Behavior -> QwenLM -> Quirks -> Completion Request Timeout (s)

If you're seeing QwenLM: completion request not observed, try raising this. This is different from the stream timeout: it only covers the gap before Qwen's backend request starts.

First Chunk Timeout

How long (in seconds) IntenseRP waits for QwenLM's response stream to produce its first chunk after the completion request has started.

Settings -> Provider Behavior -> QwenLM -> Quirks -> First Chunk Timeout (s)

If you're seeing timed out waiting for intercepted first chunk errors on QwenLM, this is the setting to raise.

Provider guardrails (recommended)

QwenLM has a couple of web settings that can silently change how your prompt gets sent (for example, turning big messages into file attachments).

To keep things predictable, IntenseRP tries to auto-disable these when the driver starts:

• Large Text as File
• Split Large Chunks
• Memory
• History Memory

If any of those get changed, you might see the Qwen tab reload once. That is normal.

Reuse Matching Chat

Reuse Matching Chat tries to keep chats tidy: when you send the exact same prompt twice in a row, IntenseRP clicks Qwen's "Regenerate" instead of creating a brand new chat. Especially useful if you swipe a lot in SillyTavern and want to hit ratelimits less.

Settings -> Provider Behavior -> QwenLM -> Reuse Matching Chat

Delete Chat After Reply

If you want Qwen's chat list cleaned up automatically, IntenseRP can delete the completed Qwen chat after a successful reply finishes.

Settings -> Provider Behavior -> QwenLM -> Delete Chat After Reply

See also: Chat Auto-Deletion

Login notes

QwenLM supports Auto Login (email + password). Persistent Sessions are still recommended, since they reduce how often you need to log in.

See: Login & Sessions

UI language requirement

The QwenLM driver currently expects the Qwen web UI language to be English (en / en-US). If QwenLM is set to another language, IntenseRP may fail to find buttons/toggles reliably.

If you see a warning about QwenLM UI language:

1. Change QwenLM language to English
2. Reload the page (F5 / Ctrl+R)
3. Retry / restart the browser from IntenseRP if needed

Per-message macros

You can add simple [[...]] macros to the latest user message in SillyTavern to override certain QwenLM Behavior settings for that request only.

All macros are stripped from the message before sending it to QwenLM.

Quick Reference

Back to Providers

Providers Overview