Provider Setup

Connect Chalie to an LLM provider — Ollama, Anthropic, OpenAI, Gemini, OpenRouter, or any OpenAI-compatible service.

Choosing a provider

Chalie needs an LLM (large language model) to think with. You pick the provider, and Chalie handles the rest. Your options fall into two categories:

  • Local — run a model on your own hardware with Ollama. Free, private, no internet required. Best for privacy-focused setups or when you have a capable GPU.
  • Cloud — connect to Anthropic, OpenAI, Google Gemini, OpenRouter, or any OpenAI-compatible API. Better quality on most tasks, requires an API key and internet connection.

You can switch providers at any time without losing your data.

How to add a provider

Adding a provider is a guided flow inside the Brain dashboard — pick your provider, add a key if it needs one, and choose a model. Chalie fills in the technical details for you.

  1. Open the Brain dashboard (the 🧠 brain icon in the top-right corner)
  2. Select Providers in the Brain sidebar
  3. Click Add Provider

1. Choose your provider

Chalie shows a grid of popular providers — Ollama for local models, the major cloud APIs (Anthropic, OpenAI, Google Gemini), and a wide range of OpenAI-compatible services like DeepSeek, OpenRouter, Groq, and Mistral. Tap the tile for the service you want to use.

The provider picker — a grid of popular AI providers. Pick one and Chalie pre-fills its connection details.

If your provider isn’t in the grid, pick Custom (OpenAI-compatible) at the end and enter the details yourself.

2. Add your credentials

Picking a tile opens the setup form with the connection details already filled in. You only ever supply what’s genuinely yours:

  • The Host / Base URL is pre-filled for you — you only touch it for Ollama (to point at a different machine) or Custom.
  • An API Key field appears only when the provider needs one. Local Ollama needs no key, so that field never shows.

Setting up DeepSeek — the host is pre-filled automatically and an API Key field appears because this provider needs one.

Paste your API key (if asked), and Chalie fetches the provider’s live model list in the background.

3. Pick a model and save

  1. Once your credentials are complete, the Model dropdown fills itself with the models the provider actually offers — pick one from the list rather than typing an ID, so it always matches what’s available.
  2. Optionally click Test to confirm the connection — Chalie reaches the provider and reports back.
  3. Click Save.

Once saved, Chalie will use your new provider immediately. The recommended models in the tables below are the ones to look for in that dropdown.

Ollama (local)

Run models entirely on your machine. No API key, no cloud, no cost.

Prerequisites

Install Ollama from ollama.com and pull a model:

ollama pull gemma4:31b

Configuration

Field Value
Provider tile Ollama (local)
Host http://localhost:11434 (pre-filled — the default Ollama address)
Model gemma4:31b (recommended)

No API key is needed for local Ollama.

Remote Ollama

If Ollama runs on a different machine on your network, change the Host field to point to that machine’s address (e.g. http://192.168.1.50:11434). Make sure the Ollama server is configured to accept remote connections by setting OLLAMA_HOST=0.0.0.0 on the host machine.

Anthropic (Claude)

Connect to Claude models from Anthropic — strong at writing, analysis, and following complex instructions.

Get your API key

  1. Create an account at console.anthropic.com
  2. Go to API Keys in the dashboard
  3. Click Create Key and copy it

Configuration

Field Value
Provider tile Anthropic
API Key Your Anthropic API key (starts with sk-ant-)
Model claude-opus-4-6 (recommended)

Available models

Model Speed Quality Cost
claude-haiku-4-5 Fastest Good Lowest
claude-sonnet-4-6 Fast Very good Medium
claude-opus-4-6 Moderate Excellent Higher
claude-opus-4-7 Slower Best Highest

OpenAI (GPT)

Connect to GPT models from OpenAI.

Get your API key

  1. Create an account at platform.openai.com
  2. Go to API Keys in the sidebar
  3. Click Create new secret key and copy it

Configuration

Field Value
Provider tile OpenAI
API Key Your OpenAI API key (starts with sk-)
Model gpt-5.5 (recommended)

Available models

Model Speed Quality Cost
gpt-5.4-nano Fastest Good Lowest
gpt-5.4-mini Fast Very good Low
gpt-5.4 Fast Excellent Medium
gpt-5.5 Moderate Best Higher

Google Gemini

Connect to Google’s Gemini models. Generous free tier available.

Get your API key

  1. Go to aistudio.google.com
  2. Click Get API key in the sidebar
  3. Create a key in a new or existing Google Cloud project and copy it

Configuration

Field Value
Provider tile Google Gemini
API Key Your Gemini API key
Model gemini-3.5-flash (recommended)

Available models

Model Speed Quality Cost
gemini-3.1-flash-lite Fastest Good Free tier available
gemini-2.5-flash Fast Very good Free tier available
gemini-3.5-flash Fast Excellent Free tier available
gemini-2.5-pro Slower Best reasoning Paid

Gemini’s free tier is generous and a great way to try Chalie at zero cost.

OpenRouter

Access hundreds of models from a single API key. OpenRouter sits between Chalie and providers like Meta (Llama), Mistral, Anthropic, OpenAI, and dozens more — you pick the model, OpenRouter handles the routing.

Get your API key

  1. Create an account at openrouter.ai
  2. Go to Keys in the dashboard
  3. Click Create Key and copy it

Configuration

Field Value
Provider tile OpenRouter
Host https://openrouter.ai/api/v1 (pre-filled)
API Key Your OpenRouter API key
Model deepseek/deepseek-v4-pro (recommended)
Model Provider Notes
deepseek/deepseek-v4-pro DeepSeek Top-tier reasoning, 1M context, competitive pricing
deepseek/deepseek-v4-flash DeepSeek Fast and cheap, also available free
meta-llama/llama-4-maverick Meta 400B MoE, 1M context, open-weight
google/gemini-3.5-flash Google Near-Pro quality at Flash cost
anthropic/claude-opus-4-6 Anthropic Claude via OpenRouter
openai/gpt-5.5 OpenAI GPT-5.5 via OpenRouter

Browse the full catalogue at openrouter.ai/models and copy the model ID directly into Chalie.

Why OpenRouter?

  • One key, many models — try different models without managing separate accounts
  • Fallback routing — if a model is down, OpenRouter can switch to an alternative
  • Pay-as-you-go — fund one account instead of several

Mistral AI

Connect to Mistral’s models using their OpenAI-compatible API.

Get your API key

  1. Create an account at console.mistral.ai
  2. Go to API Keys
  3. Create a new key and copy it

Configuration

Field Value
Provider tile Mistral
Host https://api.mistral.ai/v1 (pre-filled)
API Key Your Mistral API key
Model mistral-medium-3-5 (recommended)

Available models

Model Speed Quality
mistral-small-2603 Fastest Good
mistral-large-2512 Fast Very good, open-weight
mistral-medium-3-5 Moderate Best — 128B dense, 256K context

DeepSeek

DeepSeek offers high-quality reasoning models at competitive prices through an OpenAI-compatible API.

Get your API key

  1. Create an account at platform.deepseek.com
  2. Go to API Keys in the sidebar
  3. Create a new key and copy it

Configuration

Field Value
Provider tile DeepSeek
Host https://api.deepseek.com (pre-filled)
API Key Your DeepSeek API key
Model deepseek-v4-pro (recommended)

Available models

Model Best for
deepseek-v4-flash Fast and efficient, everyday use
deepseek-v4-pro Top-tier reasoning, 1M context
deepseek-r1-0528 Dedicated reasoning model

Groq

Groq runs open-weight models on custom LPU hardware, delivering extremely fast inference. Free tier available. Groq is not a model creator — it hosts models from Meta (Llama), Mistral, and others at high speed.

Get your API key

  1. Create an account at console.groq.com
  2. Go to API Keys
  3. Create a new key and copy it

Configuration

Field Value
Provider tile Groq
Host https://api.groq.com/openai/v1 (pre-filled)
API Key Your Groq API key
Model llama-4-maverick (recommended)

Groq’s model catalogue changes as new open-weight models are released. Check console.groq.com/docs/models for the current list.

Other OpenAI-compatible providers

Any service that exposes an OpenAI-compatible chat completions endpoint works with Chalie. This includes providers like Together AI, Fireworks AI, Perplexity, and self-hosted solutions like vLLM or LocalAI. Pick the Custom (OpenAI-compatible) tile at the end of the provider grid and fill in the details yourself.

Configuration

Field Value
Provider tile Custom (OpenAI-compatible)
Host The provider’s base URL (e.g. https://api.together.xyz/v1)
API Key Your API key for the provider
Model Pick from the list Chalie fetches once your host and key are set

Examples

Provider Host Example model
Together AI https://api.together.xyz/v1 meta-llama/Llama-4-Maverick-17B-128E-Instruct-Turbo
Fireworks AI https://api.fireworks.ai/inference/v1 accounts/fireworks/models/llama4-maverick-instruct-basic
Perplexity https://api.perplexity.ai sonar-pro
vLLM (self-hosted) http://your-server:8000/v1 Your loaded model name
LocalAI http://localhost:8080/v1 Your loaded model name

The key requirement is that the service implements the /chat/completions endpoint in OpenAI format. If it does, Chalie will work with it.

Switching providers

You can change your active provider at any time:

  1. Open the Brain dashboard
  2. Select Providers in the Brain sidebar
  3. Click on a different provider to make it active

All your conversations, memories, and settings carry over — only the model powering Chalie’s responses changes.

Troubleshooting

  • “API key invalid” — double-check you copied the full key including any prefix (e.g. sk-ant- for Anthropic)
  • “Connection refused” — for Ollama, make sure the server is running (ollama serve); for cloud providers, check your internet connection
  • Model dropdown is empty — Chalie loads the model list automatically once your Host (Ollama / Custom) and API Key are filled in; if it stays empty, your connection details are wrong or the provider is unreachable. Use Test to confirm Chalie can reach the provider.
  • Slow responses — local models depend on your hardware; try a smaller model or switch to a cloud provider for faster results
  • “Rate limit exceeded” — you’ve hit the provider’s usage cap; wait a moment and try again, or upgrade your plan