AI Models & Setup

Before building AI-powered flows, you need to configure at least one AI model. Flow-Like supports a wide range of providers—from cloud APIs like OpenAI and Anthropic to local models via Ollama.

Supported Providers

Flow-Like supports these AI providers out of the box:

Provider	Type	Best For
OpenAI	Cloud	GPT-4, GPT-4o, o1, o3 models
Azure OpenAI	Cloud	Enterprise deployments
Anthropic	Cloud	Claude 3.5, Claude 4 models
Google	Cloud	Gemini models
Ollama	Local	Self-hosted, privacy-first
Groq	Cloud	Ultra-fast inference
DeepSeek	Cloud	Reasoning models
Mistral	Cloud	European AI provider
Together AI	Cloud	Open-source models
OpenRouter	Cloud	Model aggregator
Perplexity	Cloud	Search-enhanced AI
Cohere	Cloud	Enterprise NLP
HuggingFace	Cloud	Open-source models
xAI (Grok)	Cloud	X’s AI models
VoyageAI	Cloud	Embedding models
Hyperbolic	Cloud	High-performance inference
Moonshot	Cloud	Multilingual models

Adding a Model to Flow-Like

Step 1: Open Profiles

Go to Settings → Profiles in Flow-Like Desktop:

1. Click your profile avatar in the sidebar
2. Select Profiles from the menu
3. Click on your active profile

Step 2: Add a Provider

In the Models section of your profile:

1. Click Add Provider
2. Select your provider from the list
3. Enter your API key (or connection details for local models)
4. Click Save

Getting API Keys

Each provider has their own signup process:

• OpenAI: platform.openai.com
• Anthropic: console.anthropic.com
• Google AI: aistudio.google.com
• Ollama: No key needed—just run Ollama locally!

Step 3: Select Models

After adding a provider, Flow-Like will fetch the available models. Choose which ones you want to use:

1. Browse the available models
2. Toggle on the models you want active
3. Optionally, set a default model for quick access

Using Models in Your Flows

Once configured, you can use AI models in your flows with these nodes:

Find Model Node

The Find Model node automatically selects the best available model based on your preferences:

┌─────────────────┐     ┌─────────────────┐
│  Make Model     │────▶│   Invoke LLM    │
│  Preferences    │     │                 │
└─────────────────┘     └─────────────────┘

How it works:

1. Add a Make Preferences node
2. Set preferences like speed, cost, or capability requirements
3. Connect to your LLM node
4. Flow-Like automatically picks the best matching model

Direct Model Selection

Alternatively, use a specific model by its identifier:

1. Add your provider’s Prepare Model node (e.g., “Prepare OpenAI”)
2. Select the specific model from the dropdown
3. Connect to your LLM invocation node

Model Preferences

The Model Preferences system helps you choose the right model dynamically:

Preference	Description
Speed	Prioritize fast response times
Cost	Prefer cheaper models
Quality	Prefer more capable models
Context Size	Require specific context window
Capabilities	Require features like vision or tools

This is especially useful when you want your flows to:

• Automatically fall back to alternatives if a model is unavailable
• Balance cost vs. quality based on the task
• Use different models in development vs. production

Local Models with Ollama

For privacy-sensitive applications or offline use, you can run models locally with Ollama:

Setting Up Ollama

1. Download Ollama from ollama.ai
2. Install and run Ollama on your machine
3. Pull a model: ollama pull llama3.2 or ollama pull mistral
4. In Flow-Like, add Ollama as a provider (usually auto-detected)

Recommended Local Models

Model	Size	Good For
llama3.2	3B/8B	General purpose, fast
mistral	7B	Coding, reasoning
phi-4	14B	High quality, balanced
deepseek-r1	7B/32B	Complex reasoning
nomic-embed-text	-	Text embeddings

Hardware Requirements

Local models require a capable machine. For 7B models, you’ll want at least 8GB RAM. For larger models (32B+), 32GB+ RAM or a GPU is recommended.

Model Configuration Tips

For Chat Applications

• Use models with large context windows (32K+ tokens)
• Enable streaming for better user experience
• Consider cost for high-volume applications

For RAG & Knowledge Retrieval

• Pair a generative model with an embedding model
• Use the same embedding model for indexing and querying
• Consider specialized models like nomic-embed-text

For Agents & Tools

• Use models explicitly designed for tool use (GPT-4, Claude 3.5+)
• Larger models generally perform better with complex tools
• Test with simpler tasks first

Troubleshooting

”No models available”

• Check your API key is valid and has credits
• Verify your internet connection (for cloud providers)
• For Ollama, ensure it’s running (ollama serve)

“Model not responding”

• Check rate limits on your API key
• Try a different model from the same provider
• For local models, check system resources

”Unexpected responses”

• Verify the model supports your use case (e.g., vision, tools)
• Check your system prompt and temperature settings
• Try a more capable model

Next Steps

With your models configured, you’re ready to build:

• Simple chat: Continue to Chat & Conversations
• Knowledge retrieval: Jump to RAG & Knowledge Bases
• Autonomous agents: See AI Agents