Providers

AI Providers

Configure and manage AI providers for Karolina’s intelligence features.

Overview

Karolina supports multiple AI providers, allowing you to choose the best model for your needs. Each provider has different capabilities, pricing, and performance characteristics.

Supported Providers

OpenAI

Model	Use Case	Speed	Cost
GPT-4o	Complex reasoning, analysis	Medium	Higher
GPT-4o Mini	Fast, efficient tasks	Fast	Lower
GPT-4 Turbo	Long context, complex tasks	Medium	Higher
GPT-3.5 Turbo	Simple, quick tasks	Fast	Lowest

Anthropic

Model	Use Case	Speed	Cost
Claude Opus	Complex analysis, long documents	Slow	Higher
Claude Sonnet	Balanced performance	Medium	Medium
Claude Haiku	Fast, simple tasks	Fast	Lower

Groq

Model	Use Case	Speed	Cost
Llama 3 70B	Fast inference	Very Fast	Competitive
Llama 3 8B	Quick, simple tasks	Very Fast	Low

GLM-5

Model	Use Case	Speed	Cost
GLM-5-Pro	Complex reasoning	Medium	Competitive
GLM-5-Flash	Fast, efficient	Fast	Lower

Provider Configuration

Setting Default Provider

Go to Settings → AI
Click Default Provider
Select provider and model
Save

Organization vs. Personal

Setting	Scope	Who Configures
Organization	All users in org	Admin only
Personal	Your account only	Individual user

Provider Priority

Set fallback order:

Go to Settings → AI
Click Provider Priority
Drag to reorder
First available is used

Provider Settings

OpenAI Configuration

Enter API Key
Select default model
Set organization (optional)
Test connection
Save

Anthropic Configuration

Enter API Key
Select default model
Test connection
Save

Groq Configuration

Enter API Key
Select default model
Test connection
Save

GLM-5 Configuration

Enter API Key
Select default model
Test connection
Save

Model Selection

When to Use Each

Task	Recommended Provider
Quick summarization	GPT-4o Mini, Haiku
Complex analysis	GPT-4o, Claude Opus
Long document analysis	GPT-4 Turbo, Claude
Cost-sensitive tasks	Groq, GPT-3.5
Fast responses	Groq, GPT-4o Mini

Automatic Routing

Let Karolina choose:

Enable Smart Routing
Karolina selects based on:
- Task complexity
- Latency requirements
- Cost optimization
- Quality needs

Task-Specific Settings

Override defaults per task:

Go to Settings → AI → Tasks
Configure per task:
- Summary generation
- Email drafting
- Risk analysis
- Report creation

Usage and Costs

Usage Tracking

Monitor AI spend:

Metric	Description
Requests	Number of API calls
Tokens Used	Input + output tokens
Cost	Estimated spend
By Provider	Breakdown per provider
By User	Usage per team member

Cost Controls

Set limits:

Control	Description
Monthly budget	Cap total spend
Per-user limits	Individual caps
Model restrictions	Block expensive models
Auto-disable	Turn off if over budget

Usage Reports

View detailed reports:

Go to Settings → AI → Usage
Select date range
Filter by:
- Provider
- User
- Task type
Export data

Security

API Key Management

Practice	Description
Encryption	Keys encrypted at rest
Rotation	Regularly rotate keys
Scopes	Use minimal required scopes
Monitoring	Track key usage

Data Privacy

Concern	How Karolina Handles
Data sent to AI	Minimized, anonymized where possible
Training on data	Providers don’t train on Karolina data
Retention	AI responses not persistently stored
Compliance	SOC 2 compliant providers

Troubleshooting

Issue	Solution
API key invalid	Verify key, regenerate if needed
Rate limit hit	Wait, use fallback model
Slow responses	Switch to faster model
High costs	Enable cost controls, use cheaper models
Provider down	Switch to alternative provider
Connection failed	Check API key, network