Core Concept: Providers
Providers are the underlying AI engines (Large Language Models) that process your text and generate responses. Rephlo acts as a universal interface, allowing you to swap these brains in and out as needed.
Provider Types
Rephlo supports 9 BYOK providers plus a Dedicated API, giving you flexibility to choose the best AI for each task:
Cloud Providers (Internet Required)
- Azure OpenAI: Azure-hosted OpenAI models (GPT-4, GPT-4o) with enterprise-grade security, compliance, and regional data residency. Requires an Azure subscription with an OpenAI resource deployed. Has its own dedicated configuration panel (endpoint URL, deployment name, API version).
- OpenAI: The creators of GPT-5.1 and o3. Industry standard for reasoning, coding, and creativity.
- Anthropic: Creators of Claude Sonnet 4.5 and Opus 4.5. Known for high-quality writing and large context windows (200K+ tokens).
- Groq: An ultra-fast inference provider using custom LPU chips -- near-instant responses.
- Google: Access to Gemini 3 models with strong multimodal capabilities.
- xAI (Grok): Grok 4.1 models with real-time knowledge and unique personality.
- OpenRouter: A unified API gateway to 100+ models (LLaMA, Mixtral, Command-R, and more) with automatic fallback.
- OpenAI-Compatible: Connect to any OpenAI SDK-compatible service including vLLM, LocalAI, LiteLLM, LM Studio, and other self-hosted deployments. Requires API Key, Base URL, and Model Name.
Local Providers (Complete Privacy)
- Ollama: Run open-source models (Llama 4, Mistral, Gemma, DeepSeek, Qwen 3) directly on your computer. Your data never leaves your machine.
Managed Provider
- Dedicated API: Rephlo's managed cloud option with OAuth 2.0 authentication -- simplest setup for users who don't want to manage API keys.
Note: Azure OpenAI is a distinct, first-class provider in Rephlo with its own configuration control. It is not a variant of OpenAI-Compatible. If you use Azure-hosted OpenAI models, select Azure OpenAI as the provider type, not OpenAI-Compatible.
Connection Modes
You can connect to these providers in two ways:
┌─────────────────────────────────────────────────────────────────────────┐
│ CONNECTION MODE COMPARISON │
├─────────────────────────────────────────────────────────────────────────┤
│ │
│ DEDICATED MODE (Managed) BYOK MODE (Bring Your Own Key) │
│ │
│ ┌─────────────┐ ┌─────────────┐ │
│ │ Rephlo │ │ Rephlo │ │
│ │ App │ │ App │ │
│ └──────┬──────┘ └──────┬──────┘ │
│ │ │ │
│ ▼ │ │
│ ┌─────────────┐ │ │
│ │ Rephlo │ │ Direct │
│ │ Cloud │ │ Connection │
│ └──────┬──────┘ │ │
│ │ │ │
│ ▼ ▼ │
│ ┌─────────────┐ ┌─────────────┐ │
│ │ AI Provider │ │ AI Provider │ │
│ │ (OpenAI, │ │ (OpenAI, │ │
│ │ Anthropic) │ │ Anthropic) │ │
│ └─────────────┘ └─────────────┘ │
│ │
│ Pros: Pros: │
│ - Zero setup - Maximum privacy │
│ - Managed billing - Direct vendor pricing │
│ - Latest models │
│ Cons: Cons: │
│ - Less control - Requires vendor account │
│ │
└─────────────────────────────────────────────────────────────────────────┘
1. Dedicated Mode (Managed)
- Description: Rephlo manages the connection and billing. You might use a generic "Rephlo Cloud" account (if applicable to your license).
- Pros: Zero setup.
- Cons: Less control over specific model parameters.
2. BYOK Mode (Bring Your Own Key)
- Description: You use your own personal API Key directly from the vendor (e.g., your own OpenAI Platform API Key).
- Pros:
- Privacy: Your data goes directly from your machine to the vendor, bypassing Rephlo servers.
- Control: You pay the vendor directly (often cheaper for heavy users).
- Flexibility: Access new models immediately upon release.
- Cons: Requires setting up an account with the vendor.
Getting API Keys (BYOK)
When using BYOK, you are responsible for the API usage costs directly with the vendor.
| Provider | Where to Get Keys | Notes |
|---|---|---|
| Azure OpenAI | portal.azure.com | Requires Azure subscription + deployed OpenAI resource |
| OpenAI | platform.openai.com | GPT-5.1, o3, DALL-E |
| Anthropic | console.anthropic.com | Claude Sonnet 4.5, Opus 4.5 |
| Groq | console.groq.com | Free tier available |
aistudio.google.com | Gemini 3 Pro/Flash | |
| xAI | console.x.ai | Grok 4.1 |
| OpenRouter | openrouter.ai | One key, 100+ models |
| OpenAI-Compatible | Varies by service | vLLM, LocalAI, LiteLLM, LM Studio |
Switching Providers
You can switch providers globally (for all commands) or per-command:
- Example: Use Groq for quick spell-checks (fast/cheap), but switch to Claude for complex legal analysis (smart/nuanced).
┌─────────────────────────────────────────────────────────────────────────┐
│ PROVIDER CONFIGURATION DIALOG │
├─────────────────────────────────────────────────────────────────────────┤
│ │
│ ┌─────────────────────────────────────────────────────────────────┐ │
│ │ Add Provider [X] │ │
│ ├─────────────────────────────────────────────────────────────────┤ │
│ │ │ │
│ │ Provider Type: [OpenAI v] │ │
│ │ │ │
│ │ Display Name: [My GPT-5____________________________] │ │
│ │ │ │
│ │ ───────────────────────────────────────────────────────── │ │
│ │ API Key: [sk-xxxx...xxxx____________________] [Show] │ │
│ │ │ │
│ │ Model: [gpt-5.1 v] │ │
│ │ [ ] gpt-5.1 │ │
│ │ [ ] gpt-5.1-mini │ │
│ │ [ ] o3 │ │
│ │ │ │
│ │ ───────────────────────────────────────────────────────── │ │
│ │ Status: [Test Connection] │ │
│ │ │ │
│ │ Connection successful! │ │
│ │ │ │
│ │ [ ] Set as default provider │ │
│ │ │ │
│ │ [Cancel] [Save] │ │
│ │ │ │
│ └─────────────────────────────────────────────────────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────────────┘
Provider Comparison Guide
| Provider | Best For | Speed | Context Size |
|---|---|---|---|
| Azure OpenAI (GPT-4o) | Enterprise, compliance | Fast | 128K |
| GPT-5.1 | General, Coding, Analysis | Fast | 256K |
| Claude Sonnet 4.5 | Writing, Long docs | Fast | 200K |
| Groq | Quick tasks | Ultra-fast | 128K |
| Gemini 3 | Multimodal, Research | Fast | 2M+ |
| Grok 4.1 | Real-time knowledge | Fast | 128K |
| OpenRouter | Multi-model access | Varies | Varies |
| OpenAI-Compatible | Self-hosted, custom | Varies | Varies |
| Ollama | Privacy, Offline | Depends on HW | 8-128K |
Next Concept: Explore the Dashboard & Navigation.