How to use Janitor AI with SiliconFlow APIs

8 Apr 2026

Daftar Isi

Janitor AI is a leading AI platform serving millions of users worldwide, built for immersive roleplay, character-driven storytelling, and interactive fiction. Whether you're into romantasy, fanfiction, or building entirely original worlds, Janitor AI delivers safe, engaging, and innovative experiences, pushing the boundaries of what AI-powered storytelling can be.

To power these experiences, Janitor AI connects to external LLM APIs. In this guide, we'll show you how to use SiliconFlow APIs in Janitor AI, giving you access to 200+ open-source and frontier models at competitive pricing.

Step 1: Prerequisites

Get Your SiliconFlow API Key

Before you begin, ensure you have a valid SiliconFlow account:

Register a SiliconFlow account at https://cloud.siliconflow.com/. You can sign up with Google or GitHub.
Log in and navigate to API Keys in the dashboard.
Click 🔑Create API Key and add a name to your key to help you identify it later, then confirm to create it.
Click the API key to copy it automatically.

Please keep your API key somewhere safe, as you'll need it to configure Janitor AI in the next step.

Step 2: Configure SiliconFlow APIs

Open Janitor AI: Log in at janitorai.com, then start or create a chat.
API Settings: In the chat, click the gear icon (top-right) → "API Settings" → "Proxy" → "Add Configuration".

Fill in the details:
- Config Name: SiliconFlow
- Proxy URL: https://api.siliconflow.com/v1/chat/completions
- API Key: Paste the SiliconFlow API key you copied in Step 1.
- Model Name: Enter the model you want to use, for example deepseek-ai/DeepSeek-V3.2. You can browse the full model list at https://www.siliconflow.com/models

Save & Test: Click "Test", then "Save Settings".

Refresh the page and send a test message to confirm everything is working.

You're all set. Select or create a character in Janitor AI now!

With SiliconFlow connected to Janitor AI, you can easily experiment with a wide range of open-source and frontier models for roleplay and storytelling. Different models bring different strengths, from expressive dialogue and immersive narration to long-context reasoning and character consistency.

Try a few models, explore their behaviors, and discover which ones best bring your characters and stories to life.

Which Models Should You Try First?

If you're not sure where to start, the OpenRouter Roleplay leaderboard is a useful reference. As of April 2026, the most-used models by Janitor AI users on Openrouter are largely dominated by a handful of open-source models. SiliconFlow supports all of them ranked within the Top 10, including DeepSeek V3.2, Step 3.5 Flash, DeepSeek V3 0324, GLM 4.5 Air, DeepSeek V3.1, GLM 5, DeepSeek R1 0528, and DeepSeek V3.2 Exp.

While the leaderboard highlights which models are most widely used, it doesn't fully explain why they perform well in roleplay scenarios. To better understand their strengths, the following sections summarize roleplay-related capabilities for DeepSeek models and the GLM model series.

Model	Roleplay Strengths	Typical Use Cases	Source
DeepSeek V3.2 / V3-0324	Highly adaptable and moldable to user style and conversational tone. Performs well in natural dialogue and cooperative storytelling. Sometimes too accommodating	Slice-of-life, romance scenarios, casual interactions	Janitor AI
DeepSeek R1-0528	Strong reasoning capability with explicit chain-of-thought (<think> tags). Tends to follow character traits strictly, producing more intense or dramatic character behavior.	Action-heavy, dramatic narratives, strong character personalities	Janitor AI
GLM-5	Strong long-context memory and planning abilities. Maintains consistent character settings, narrative logic, and emotional continuity across long interactions.	Long-form roleplay, complex storylines, multi-stage narrative scenarios	Z.ai Doc
GLM-4.7	Produces more vivid and descriptive narrative text, with stronger atmosphere building and sensory details. Maintains world-building consistency and character archetypes.	Immersive storytelling, narrative-driven RP, IP content creation	Z.ai Doc
GLM-4.5 Air	Designed for multi-turn character dialogue with stable tone and behavior across conversations.	Dialogue-heavy RP systems, character chat applications	Z.ai Doc