Skip to content

Provider Support

IntenseRP Next v2 is designed to support multiple providers by driving their web apps and intercepting the underlying network requests.

Today, DeepSeek, GLM Chat (Z.ai), Moonshot, QwenLM, Perplexity, HuggingChat, and Google AI Studio are implemented and usable.

Search Older Matching Chats is currently supported on DeepSeek, GLM Chat, Moonshot, QwenLM, and HuggingChat. Perplexity does not support chat reuse yet, and Google AI Studio still only has the regular single-slot Reuse Matching Chat flow for now.

GLM status (important)

The GLM driver is still beta-like. It is mostly usable, but:

  • Search and Advanced Search are supported (search results are not sent to the client)
  • Login requires solving a CAPTCHA (Persistent Sessions are strongly recommended)
  • GLM model selection is supported (GLM-5.1 / GLM-5-Turbo / GLM-5V-Turbo / GLM-5 / GLM-4.7) via Settings -> Provider Behavior -> GLM Chat -> Model
  • Reuse Matching Chat is currently unreliable with GLM

Web apps change

Provider drivers depend on the provider's web UI and internal API shapes. If a provider updates their frontend, a driver may break until it is updated.

How providers work (in v2)

All providers follow the same general approach:

  1. Launch a real browser session (Playwright/Patchright)
  2. Log in (manual or auto-login, depending on settings)
  3. Trigger a generation in the provider UI (type/upload + click send)
  4. Intercept the provider's internal streaming request
  5. Convert the provider stream into OpenAI-style SSE deltas (/v1/chat/completions or /v1/completions)

This is why IntenseRP can present an OpenAI-compatible API even though the underlying provider is a normal web chat app.

Provider lifecycle stages

Every provider goes through the same lifecycle:

Queued Planning Prototyping Driver Implementation Integration Verification Stable

Stage legend (icon + color coded)

Stage Indicator What it means
Queued In the backlog, not started
Planning Design/requirements work
Prototyping Early experiments, not production-ready
Driver Implementation Building the provider driver (UI + interception)
Integration Wiring driver into the app and API surface
Verification Stress testing, edge cases, reliability
Stable Considered reliable for normal use

Current provider roadmap

Providers are prioritized in this order:

Provider Priority Current stage
DeepSeek 1 Stable
GLM Chat 2 Stable (mostly)
Moonshot 3 Stable (mostly)
Google AI Studio 4 Verification
QwenLM 5 Stable
Perplexity 6 Verification
HuggingChat 7 Driver Implementation

What 'Verification' means for Moonshot

Moonshot is implemented and usable, but this is the first integration pass. Expect selector and stream-shape adjustments as the provider UI evolves.

Google AI Studio status

Of all the providers, Google AI Studio is the weirdest and most complex, so it is taking extra time to verify and stabilize. It is usable, but expect some rough edges until it's fully verified. Breakage is also possible, especially if Google updates the AI Studio web UI.

Perplexity status

Perplexity is implemented as an early integration. It can send prompts and stream answer text, but chat reuse/regeneration and thinking-trace forwarding are not supported yet.

HuggingChat status

HuggingChat is implemented as an early integration. It supports model selection, inference provider selection, thinking effort, Exa search, uploads, chat reuse, and account rotation, but HuggingChat's web UI and monthly credits are both easy to run into. Disable spent accounts until their credits reset.

  • DeepSeek DeepSeek Behavior

    DeepThink, Search, anti-censorship, and more.

    ➡ DeepSeek Behavior

  • Z.ai GLM Behavior

    Deep Think, modes, Search status, and GLM-specific notes.

    ➡ GLM Behavior

  • MoonshotAI Moonshot Behavior

    Thinking, Search, file uploads, and Kimi-specific caveats.

    ➡ Moonshot Behavior

  • Qwen QwenLM Behavior

    Thinking, Web search, model selection, and token counting.

    ➡ QwenLM Behavior

  • Perplexity Perplexity Behavior

    Model selection, Thinking, Search, uploads, and email-code login notes.

    ➡ Perplexity Behavior

  • HuggingFace HuggingChat Behavior

    Model selection, inference provider, thinking effort, Exa search, and quota/account notes.

    ➡ HuggingChat Behavior

  • Google AI Studio Google AI Studio Behavior

    AI Studio model selection, Thinking Level, Search, URL Context, and sampling controls.

    ➡ Google AI Studio Behavior

  • Login & Sessions

    Auto login and persistent sessions.

    ➡ Login & Sessions

  • Network & API

    Ports, LAN access, and API key auth.

    ➡ Network & API

  • API Behavior

    Request flow, streaming, cancellation, and queueing.

    ➡ API Behavior

Back to Advanced

Advanced