Skip to content

Perplexity Perplexity Behavior

This page covers the settings that control how IntenseRP interacts with Perplexity (perplexity.ai).

Perplexity is powerful, but its web app is fairly strict. IntenseRP drives the same composer you use in the browser, watches the perplexity_ask stream, and forwards only the assistant answer text to the OpenAI-compatible API.

Free tier limitations

It really only makes sense to use Perplexity with a paid (Pro or higher) account. IntenseRP can work with free accounts, but the experience is rough due to being locked mainly to the Sonar 2 model without Thinking controls. It's also heavily rate-limited, so expect slow responses and possible temporary blocks if you send too many requests.

Special thanks

Huge thank you to Yurushia for their financial support and giving me access to a paid Perplexity account for development and testing. Without them, this integration quite literally would not exist. So, wholeheartedly, thank you!! 💙

Modes (model IDs)

In IntenseRP Next v2, the model you select in SillyTavern is mostly a behavior preset, not the same thing as Perplexity's paid model picker.

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

Model ID Behavior
perplexity-auto Uses your IntenseRP settings
perplexity-chat Forces Thinking off
perplexity-reasoner Forces Thinking on when the account/model allows it

Real Perplexity model selection (web UI)

IntenseRP can switch Perplexity's real model picker in the web UI when your account exposes it:

Settings -> Provider Behavior -> Perplexity -> Model

Current options:

  • Best (Auto)
  • Sonar 2
  • GPT-5.4
  • GPT-5.5
  • Gemini 3.1 Pro
  • Claude Sonnet 4.6
  • Claude Opus 4.7
  • Kimi K2.6
  • Nemotron 3 Super

Paid-account controls

Perplexity only exposes model and Thinking controls on paid accounts. Free accounts can still send prompts, but IntenseRP skips those controls and uses whatever the Perplexity UI allows.

Spaces Mode

Spaces mode makes IntenseRP use an IntenseRP Next Space instead of the normal Perplexity chat page.

Settings -> Provider Behavior -> Perplexity -> Use Perplexity Spaces

When enabled, IntenseRP opens https://www.perplexity.ai/spaces, looks for an existing Space named IntenseRP Next, and creates it if it isn't there yet. The Space ID is remembered only for the current browser session, so if the Space is deleted IntenseRP will warn in the logs and recreate it next time.

After the Space is ready, prompts, model selection, Thinking, Search, and file uploads keep using the same Perplexity composer flow as normal chat mode.

Space Instructions

You can also move leading system messages into the Space instructions field:

Settings -> Provider Behavior -> Perplexity -> Paste System Instructions Into Space Instructions

Only leading system messages are moved. Once a user or assistant message appears, later system messages stay in the regular prompt. Perplexity caps Space instructions at 8000 characters, so anything that doesn't fit also stays in the prompt.

Loadouts

Both Spaces settings are Perplexity Behavior settings, so Perplexity loadouts can turn Spaces mode and Space-instruction syncing on or off per loadout.

Thinking

Enable Thinking

When enabled, IntenseRP tries to toggle Perplexity Thinking before sending the prompt.

Settings -> Provider Behavior -> Perplexity -> Enable Thinking

Some models force Thinking on, and some account tiers do not expose the toggle. In those cases IntenseRP logs what happened and keeps the request moving.

Send Thinking

Perplexity does not currently expose usable thinking traces in the intercepted stream. The setting exists for mode/loadout consistency, but IntenseRP does not forward <think> content for Perplexity yet.

Settings -> Provider Behavior -> Perplexity -> Send Thinking

Perplexity Web search can be toggled through the composer tools menu.

Settings -> Provider Behavior -> Perplexity -> Enable Search

Search payloads

Perplexity streams search/source data around the answer. IntenseRP strips those provider-specific payloads and forwards only the assistant text.

File Upload Mode

Instead of pasting your full prompt into Perplexity's editor, IntenseRP can upload it as a text file.

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

If Perplexity indicates that the upload cap is reached, IntenseRP falls back to pasting the prompt as normal text for that request.

Text File Message

Optional text pasted alongside the uploaded prompt file.

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

File Upload Timeout

Controls how long IntenseRP waits for the send button to become available after selecting the file.

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

Message Send Timeout

Controls how long IntenseRP waits for the send button to appear after text entry in normal, non-file mode.

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

Login notes

Perplexity uses email-code login. Auto Login can enter the saved email and click through the first step, but you still need to type the 6-digit code in the browser window.

Persistent Sessions strongly recommended

Perplexity sessions are much nicer when the browser profile stays signed in. Keep Persistent Sessions on unless you have a specific reason to reset the provider profile each time.

See: Login & Sessions

UI language requirement

The Perplexity driver currently expects the web UI language to be English (en / en-US). If the page is in another language, IntenseRP may fail to find the Sign In label, model menu, or tools menu reliably.

If you see a warning about Perplexity UI language:

  1. Change Perplexity 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 Perplexity Behavior settings for that request only.

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

Macro Effect
[[think]], [[r1]] Force Thinking on
[[nothink]], [[r0]] Force Thinking off
[[search]] Force Search on
[[nosearch]], [[no_search]] Force Search off
[[file]] Force Send As Text File on
[[nofile]] Force Send As Text File off

Scope

Only macros from the latest user message apply. They do not persist across requests.

Quick Reference

Setting What It Does Default
Model Selects Perplexity's real model picker (paid accounts) Best (Auto)
Enable Thinking Toggles Perplexity Thinking where available Off
Send Thinking Reserved; no thinking traces are forwarded yet Off
Enable Search Enables Perplexity Web search Off
Use Perplexity Spaces Uses an IntenseRP-managed Space instead of normal chat Off
Paste System Instructions Into Space Instructions Moves leading system messages into Space instructions Off
Send As Text File Uploads prompt as .txt Off
Text File Message Text pasted alongside uploaded file (empty)
File Upload Timeout Seconds to wait after upload 20
Message Send Timeout (s) Seconds to wait for send button 8

Back to Providers

Providers Overview