Provider Configuration Setup

After deploying the application, you can configure LLM providers through the web interface.

Overview

Providers define which LLM backends (Ollama, Anthropic, OpenAI, Gemini, etc.) are available to the system. All provider configuration is stored in the SQLite database.

Quick Start

1. Start the Application

python backend/run.py

The web interface will be available at http://localhost:8081.

2. Configure During Onboarding

The onboarding wizard at http://localhost:8081/on-boarding/ walks you through provider setup step by step. This is the easiest path for first-time setup.

3. Configure After Onboarding

To add or change providers later:

Open http://localhost:8081/brain/
Go to Settings → Providers
Click Add Provider

4. Provider Options

Each provider needs a name (any label you want), the platform, a model ID, and for cloud providers, an API key.

Local (free, private)

Ollama — Runs entirely on your machine. No API key needed.

Platform: Ollama
Host: http://localhost:11434
Model: Any model you’ve pulled (e.g., qwen:8b). Run ollama list to see available models.

Cloud

Platform	API Key Source	Example Model
Anthropic	console.anthropic.com	Any Claude model ID
OpenAI	platform.openai.com	Any GPT model ID
Google Gemini	ai.google.dev	Any Gemini model ID

Use the exact model ID from your provider’s documentation — model names change frequently.

5. Save and Test

After entering provider details, click Save or Test Connection to verify the configuration works.

Supported Platforms

Platform	Local?	Requires API Key?	Notes
Ollama	Yes	No	Local inference, requires Ollama running on machine
Anthropic	No	Yes	Claude API from Anthropic
OpenAI	No	Yes	GPT models from OpenAI
Google Gemini	No	Yes	Gemini models from Google

Troubleshooting

“Provider connection failed”

For Ollama: Ensure Ollama is running (ollama serve) and the host URL is correct (usually http://localhost:11434)
For cloud providers: Double-check that your API key is correct and has the necessary permissions
Network issues: Verify your internet connection and that the provider endpoint is accessible

“API key is invalid”

Check that you copied the API key correctly (no extra spaces)
Verify the key hasn’t expired or been revoked
For Anthropic: Get key from https://console.anthropic.com
For OpenAI: Get key from https://platform.openai.com/api-keys
For Gemini: Get key from https://ai.google.dev

Model not found

For Ollama: Run ollama pull <model-name> to download the model first
For cloud providers: Verify the exact model name matches what the provider offers

REST API Reference

For programmatic provider configuration (scripts, automation, CI).

Create Provider

curl -X POST http://localhost:8081/providers \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "name": "claude-haiku",
    "platform": "anthropic",
    "model": "claude-haiku-4-5-20251001",
    "api_key": "sk-ant-...",
    "timeout": 120
  }'

List All Providers

curl http://localhost:8081/providers \
  -H "Authorization: Bearer YOUR_API_KEY"

Update Provider

curl -X PUT http://localhost:8081/providers/{id} \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{"api_key": "sk-..."}'

Delete Provider

curl -X DELETE http://localhost:8081/providers/{id} \
  -H "Authorization: Bearer YOUR_API_KEY"

Assign Provider to Job

curl -X PUT http://localhost:8081/providers/jobs/frontal-cortex \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{"provider_id": 1}'

Security Best Practices

Protecting API Keys

Never commit API keys to version control
Never share API keys in logs or error messages
Use environment variables or secure secret management for production
Rotate API keys regularly
Use database encryption for sensitive columns in production

Network Security

Keep Ollama instances local or behind a firewall
Use HTTPS/TLS for remote API connections
Restrict file access to the SQLite database
Enable CORS appropriately for your deployment

Chalie runs embeddings locally via ONNX (gte-modernbert-base, 768 dimensions). The model downloads automatically on first run (~300MB) and requires no configuration. No LLM provider is used for embeddings.

PROVIDERS SETUP