Cloud AI Providers
RTILA X supports cloud AI providers for when you need more powerful models, faster responses, or don’t want to run models locally. Cloud providers process your prompts on remote servers and return responses over the internet.
Changed in v8.3.2: The “Custom Provider” has been removed and merged into the unified Local LLM provider. If you were using a custom OpenAI-compatible endpoint, switch to Local LLM → Custom API Endpoint mode. See Migration Guide below.
OpenRouter (Recommended Cloud Provider)
Section titled “OpenRouter (Recommended Cloud Provider)”OpenRouter gives you access to hundreds of AI models through a single API key. Instead of signing up for individual AI services (OpenAI, Anthropic, Google, etc.), you get one key that works with all of them.
Why OpenRouter?
Section titled “Why OpenRouter?”| Benefit | Description |
|---|---|
| One API key | Access hundreds of models with a single key |
| Pay per use | Only pay for the tokens you actually use |
| Model variety | Switch between models instantly without changing configuration |
| Competitive pricing | Often cheaper than direct API access |
| Free models | Some models are available at no cost |
Setup Steps
Section titled “Setup Steps”- Go to openrouter.ai/keys.
- Sign up for an account (or log in if you already have one).
- Click Create Key.
- Give your key a name (e.g., “RTILA X”).
- Copy the generated key (it starts with
sk-or-). - In RTILA X, go to Preferences → AI Assistant.
- Set AI Provider to OpenRouter.
- Paste your API key into the API Key field.
- Select a model from the dropdown (or enter a custom model name).
- Start chatting with the AI Assistant!
Available Models
Section titled “Available Models”RTILA X includes a curated list of popular models in the dropdown. Here are the recommended options:
| Model | Description | Speed | Quality | Cost |
|---|---|---|---|---|
minimax/minimax-m3 |
MiniMax’s latest model | Fast | Good | Low |
deepseek/deepseek-v4-flash |
DeepSeek’s fast variant | Very Fast | Good | Low |
z-ai/glm-5.3-flash |
Zhipu AI’s flash model | Very Fast | Good | Low |
google/gemini-3.5-flash-lite |
Google’s lightweight model (default) | Very Fast | Good | Very Low |
google/gemini-3.1-flash-lite |
Previous Gemini flash | Very Fast | Good | Very Low |
openai/gpt-5.6-luna |
OpenAI’s latest | Fast | Excellent | Medium |
xiaomi/mimo-v2.5 |
Xiaomi’s model | Fast | Good | Low |
google/gemini-2.5-flash-lite |
Older Gemini flash | Very Fast | Good | Very Low |
google/gemini-2.5-flash |
Gemini flash (full) | Fast | Very Good | Low |
Tip: The default model is
google/gemini-3.5-flash-lite, which offers an excellent balance of speed, quality, and cost for automation tasks.
Custom Model Names
Section titled “Custom Model Names”You’re not limited to the models in the dropdown. You can enter any OpenRouter model identifier in the model field:
- Browse available models at openrouter.ai/models.
- Find a model you want to use.
- Copy its full identifier (e.g.,
anthropic/claude-sonnet-4.5). - Paste it into the model field in RTILA X.
Verify Button
Section titled “Verify Button”If you enter a custom model name, you can verify it’s valid:
- Enter the model name in the model field.
- Click the Verify button.
- RTILA X checks the model against OpenRouter’s model list.
- You’ll see a success or error toast notification.
| Result | Meaning |
|---|---|
| ✅ “Model verified successfully” | The model exists and is available |
| ❌ “Model not found” | The model identifier is incorrect or the model doesn’t exist |
| ⚠️ “Verification failed” | Could not connect to OpenRouter to verify |
Note: Verification requires an active internet connection. If verification fails due to network issues, the model may still work — try using it directly.
Context Injection Strategy
Section titled “Context Injection Strategy”Changed in v8.3.2: When using OpenRouter (or any cloud provider), the full system prompt + complete command reference is injected into every request. This replaces the previous RAG-based context injection.
How It Works
Section titled “How It Works”Every time you send a message to the AI Assistant, RTILA X constructs a request that includes:
| Component | Description | Size |
|---|---|---|
| System Prompt | Behavioral instructions telling the AI how to act as an RTILA automation architect | ~2,000 tokens |
| Complete Command Reference | Every command, parameter, selector strategy, and configuration rule | ~8,000 tokens |
| Project Context | Your active project configuration (if chatting about a specific project) | Varies |
| Project Examples | Recent project configurations (if configured in chat settings) | Varies |
| Message History | Your conversation history (limited by chat settings) | Varies |
| Your Message | The message you just typed | Varies |
Why Full Injection?
Section titled “Why Full Injection?”Modern cloud models have context windows of 128K+ tokens, which is more than enough to hold the complete reference. This approach:
- Ensures the AI always has access to every command, parameter, and selector strategy
- Eliminates retrieval errors (no missing context, no irrelevant chunks)
- Provides consistent, deterministic responses
- Requires no additional latency from embedding generation or vector search
Proxy Mode (No API Key)
Section titled “Proxy Mode (No API Key)”If you don’t enter an API key, RTILA X uses a complimentary shared AI provided by RTILA Corporation:
- No API key required
- Limited availability (may change over time)
- Model is selected automatically by RTILA
- Displayed as “RTILA Courtesy AI (Preview)” in the interface
Important: The complimentary AI is a gift to help you get started. It is not a permanent plan and may be limited, modified, or discontinued at any time. Add your own OpenRouter API key for unlimited, guaranteed access.
Proxy Mode Banner
Section titled “Proxy Mode Banner”When using the complimentary shared AI (no API key), you’ll see a banner in the chat interface:
“Enjoying RTILA’s complimentary AI? It’s on us while you get started — availability may vary. Add your own key anytime for guaranteed access.”
Clicking this banner takes you to the AI Assistant preferences where you can enter your OpenRouter API key.
Migration from Old Custom Provider
Section titled “Migration from Old Custom Provider”Changed in v8.3.2: The “Custom Provider” option has been removed. If you were previously using it, follow this migration guide.
What Changed
Section titled “What Changed”The old “Custom Provider” allowed you to specify a custom OpenAI-compatible API endpoint. This functionality has been merged into the Local LLM provider’s Custom API Endpoint mode.
Migration Steps
Section titled “Migration Steps”- Go to Preferences → AI Assistant.
- Change AI Provider from “Custom Provider” to Local LLM.
- Select Custom API Endpoint mode.
- Enter your API settings in the new fields (see mapping table below).
Settings Mapping
Section titled “Settings Mapping”| Old Setting (Custom Provider) | New Location | Notes |
|---|---|---|
| Custom Provider → API Base URL | Local LLM → Custom API Endpoint → API Base URL | Same value, new location |
| Custom Provider → API Key | Local LLM → Custom API Endpoint → API Key | Same value, new location |
| Custom Provider → Model Name | Local LLM → Custom API Endpoint → Model Name | Same value, new location |
Example Migration
Section titled “Example Migration”Before (v8.3.151):
- AI Provider: Custom Provider
- API Base URL:
http://localhost:11434/v1 - API Key: (empty)
- Model Name:
llama3.1
After (v8.3.2):
- AI Provider: Local LLM
- Mode: Custom API Endpoint
- API Base URL:
http://localhost:11434/v1 - API Key: (empty)
- Model Name:
llama3.1
The functionality is identical — only the UI location has changed.
Choosing Between Local and Cloud
Section titled “Choosing Between Local and Cloud”| Factor | Local LLM | Cloud (OpenRouter) |
|---|---|---|
| Privacy | ✅ 100% private, data never leaves your machine | ⚠️ Prompts sent to cloud servers |
| Speed | Depends on your hardware | Generally faster (powerful servers) |
| Quality | Good (7B–14B models) | Excellent (access to frontier models) |
| Cost | Free (after hardware) | Pay per token used |
| Internet Required | No (after model download) | Yes |
| Setup Complexity | Moderate (model download, hardware tuning) | Simple (one API key) |
| Hardware Required | 6+ GB RAM minimum | None |
When to Use Local
Section titled “When to Use Local”- You need maximum privacy (scraping sensitive data)
- You have limited or no internet access
- You want zero ongoing costs
- You have sufficient hardware (8+ GB RAM)
When to Use Cloud
Section titled “When to Use Cloud”- You need the highest quality AI responses
- You want the fastest response times
- You don’t have powerful hardware
- You’re working on complex multi-step automations
- You want access to frontier models (GPT-5, Claude, Gemini)
Using Both
Section titled “Using Both”You can switch between Local LLM and OpenRouter at any time:
- Go to Preferences → AI Assistant.
- Change the AI Provider dropdown.
- Your chat history is preserved regardless of which provider you use.
Tip: Use Local LLM for sensitive scraping tasks and OpenRouter for complex automation building. Switch providers in Preferences whenever needed.
Troubleshooting
Section titled “Troubleshooting”| Issue | Cause | Solution |
|---|---|---|
| “Invalid API key” error | API key is incorrect or expired | Generate a new key at openrouter.ai/keys |
| “Model not found” error | Model identifier is incorrect | Verify the model exists at openrouter.ai/models |
| Slow responses | Model is overloaded or too large | Try a different model or use a “flash” variant |
| “Rate limit exceeded” | You’ve hit OpenRouter’s rate limit | Wait a moment or upgrade your OpenRouter plan |
| No response at all | Network issue or OpenRouter outage | Check your internet connection and try again |
| Complimentary AI unavailable | Shared AI is at capacity | Add your own OpenRouter API key |
Cost Estimation
Section titled “Cost Estimation”OpenRouter charges per token. Here’s a rough estimate for typical RTILA X usage:
| Task | Tokens per Message | Estimated Cost (Gemini Flash Lite) |
|---|---|---|
| Simple question | ~500 tokens | ~$0.0001 |
| Build a basic project | ~3,000 tokens | ~$0.0006 |
| Complex multi-step automation | ~8,000 tokens | ~$0.0016 |
| Full conversation (20 messages) | ~50,000 tokens | ~$0.01 |
Note: Costs vary by model. “Flash” and “Lite” models are significantly cheaper than frontier models. Check current pricing at openrouter.ai/models.