LLM Providers

All supported AI model providers and how to configure them.

Supported providers

Provider	Type	Notes
Ollama	Local	Free, private. Any model you pull — Llama, DeepSeek, Qwen, Mistral, CodeGemma, Phi, etc.
OpenAI	Cloud	Any OpenAI model — GPT-4o, GPT-4.1, o3-mini, etc.
Claude	Cloud	Any Anthropic model — Claude Sonnet, Opus, Haiku
Gemini	Cloud	Any Google model — Gemini Pro, Flash, etc.
Grok	Cloud	Any xAI model — Grok 2, Grok 3, etc.
Qwen	Cloud	Any Alibaba model — Qwen 2.5, Qwen-Max, etc.
Fireworks	Cloud	Any model on Fireworks — Llama, Mixtral, DeepSeek, etc.
Together	Cloud	Any model on Together — Llama, Code Llama, etc.

✓

You type the model name yourself in Settings — Codeteel doesn't restrict which models you can use. Any model available from the provider will work, as long as it supports the OpenAI-compatible chat completions API.

Web vs Platform

Codeteel has two separate LLM configurations:

	Web LLM	Platform LLM
Used by	Browser chat interface	Slack, Telegram, Discord
Ollama	Yes (direct browser → localhost)	No (server can't reach localhost)
Cloud providers	Yes (via SSE proxy)	Yes (via Lambda)
API key storage	Encrypted (AES-256-GCM)	Encrypted (AES-256-GCM)

Ollama for web: When using Ollama, LLM calls go directly from your browser to localhost:11434. No proxy, no API key, no server involved. Your code and prompts never leave your machine.

Setting up Ollama

Install Ollama from ollama.com, then:

# Start the server
ollama serve

# Pull a model (example)
ollama pull deepseek-coder-v2:16b

# Codeteel auto-discovers available models in Settings

CORS required for web app. If you're accessing Codeteel from a deployed URL (not localhost), you must enable CORS on Ollama:

Windows: set OLLAMA_ORIGINS=* && ollama serve
Mac/Linux: OLLAMA_ORIGINS=* ollama serve

Without this, the browser cannot reach Ollama due to cross-origin restrictions.

Setting up cloud providers

For any cloud provider:

Go to Settings
Click "Add Provider"
Select the provider
Paste your API key
Enter the model name (e.g., gpt-4o)
Set as active

✓

API keys are encrypted with AES-256-GCM before storage. The UI only shows the first 7 characters.

Next: Security & Encryption