AI Settings

All AI features in Trendz — Chat Assistant, AI Widget Summary, and Metric Code Assistant — run through a configured LLM provider. Cloud installations include a default model with a limited request quota. To remove that limit or to use a different provider, configure your own model in Settings → AI Assistant.

Navigate to Settings → AI Assistant. The tab has two sections:

• AI Models — add and manage the LLM connections Trendz can use
• AI Features — enable AI and assign models to specific features

Always add a model first before configuring features.

1. Click New AI Model.

2. Enter a Name — any label to identify the model in the list.

3. Choose a Provider: OPEN_AI, GOOGLE, AMAZON_BEDROCK, or CUSTOM.

4. Enter the Model Name (provider-specific identifier, e.g. gpt-4o or gemini-2.0-flash).

5. Fill in the provider-specific fields (API Key, Base URL, region, etc.).

6. Set Temperature and Top P.

7. Click Save. A success message confirms the model is connected.

Temperature controls response creativity. Lower values (closer to 0) produce consistent, predictable output; higher values introduce more variety.

Top P (nucleus sampling) controls the probability range for token selection. Lower values keep outputs focused; higher values allow broader variation.

You can add as many models as needed — for example, a fast cheap model for summaries and a more capable one for metric code generation.

Click Actions → Edit, change the desired fields, then click Save.

Click Actions → Delete and confirm.

Toggle Enable AI Features on to activate all AI capabilities in Trendz.

Two assignment modes are available:

Mode	Description
Use Single Model	One model handles all AI features
Custom per Feature Configuration	Each feature uses a separately assigned model

Switch between modes using the Use Single Model checkbox.

The three assignable features are:

Feature	Used by	Recommended model type
Prompts	AI Widget Summary	Cost-efficient (low latency, simple tasks)
View Assistant	Chat Assistant	Fast (interactive, conversational)
Metric Code Assistant	Metric Explorer	Advanced (complex code generation)

If a feature is disabled in this section, no user in the tenant can access its AI capabilities.

Once configuration is complete, click Save.

Generate API keys at platform.openai.com/api-keys.

Model	Notes	Input per 1M tokens	Output per 1M tokens
gpt-4o	High performance	$2.50	$10.00
gpt-4o-mini	Cost-efficient	$0.15	$0.60
o3-mini	Cost-efficient, high performance, slower	$1.10	$4.40
o4-mini	Cost-efficient, high performance, slower	$1.10	$4.40

Generate API keys at ai.google.dev/gemini-api/docs/api-key.

Model	Notes	Input per 1M tokens	Output per 1M tokens
gemini-2.5-flash	High performance	$0.30	$2.50
gemini-2.0-flash	Cost-efficient, average performance	$0.10	$0.40
gemini-2.0-flash-lite	Lightweight	$0.075	$0.30

Generate Access Key ID and Secret Access Key following the AWS Bedrock API guide.

Family	Model	Notes	Input per 1M tokens	Output per 1M tokens
Amazon	Nova Lite	Cost-efficient	$0.06	$0.24
	Nova Pro	—	$0.80	$3.20
Anthropic	Claude 3.5 Haiku	—	$0.80	$4.00
	Claude 3.5 Sonnet	Recommended	$3.00	$15.00
	Claude 3.7 Sonnet	—	$3.00	$15.00
Meta Llama	Llama 3.1 Instruct (70B)	—	$0.72	$0.72
	Llama 3.2 Instruct (90B)	—	$0.72	$0.72
	Llama 3.3 Instruct (70B)	Recommended	$0.72	$0.72
Mistral AI	Mistral Large (24.02)	—	$3.00	$12.00

Any model that implements the OpenAI API standard can be used. Select CUSTOM as the provider, then enter the Base URL, Model Name, and API Key.