AI Configuration

SetGet includes AI-powered features that help teams work faster — intelligent issue suggestions, automated summaries, content generation, and conversational assistance. The AI Configuration page in the Admin Panel lets you connect an LLM provider and control how AI features are made available across the instance.

Navigate to Admin Panel > AI Configuration or go directly to /backoffice/settings/ai.

How AI works in SetGet

SetGet's AI features communicate with a configurable large language model (LLM) through a standard API. The platform sends context-aware prompts (based on workspace data, project context, and user queries) to the LLM and presents the responses within the UI.

AI features include:

Feature	Description
AI Assistant	Conversational thread-based assistant accessible from the sidebar
Issue Suggestions	AI-generated issue titles, descriptions, and property recommendations
Content Generation	Draft page content, comments, and descriptions using AI
Smart Summaries	Summarize long threads, cycles, or module progress
Action Suggestions	AI recommends actions like assigning, labeling, or prioritizing

LLM provider configuration

SetGet supports any LLM provider that exposes an OpenAI-compatible API. This includes OpenAI, Anthropic (via compatible proxies), local models (Ollama, vLLM), and hosted alternatives.

Configuration fields

Field	Description	Required	Environment Variable
Provider Name	Display label for the provider (e.g., "OpenAI", "Verinova")	Yes	SETGET_AI_PROVIDER_NAME
API Endpoint URL	Base URL of the LLM API	Yes	SETGET_AI_API_ENDPOINT
API Key	Authentication key for the LLM API	Yes	SETGET_AI_API_KEY
Model	Model identifier to use (e.g., gpt-4o, claude-sonnet-4-20250514)	Yes	SETGET_AI_MODEL
Max Tokens	Maximum tokens per response	No	SETGET_AI_MAX_TOKENS
Temperature	Creativity parameter (0.0 = deterministic, 1.0 = creative)	No	SETGET_AI_TEMPERATURE

Step-by-step setup

1. Open Admin Panel > AI Configuration.
2. Toggle Enable AI Features to on.
3. Enter the API Endpoint URL (e.g., https://api.openai.com/v1 for OpenAI).
4. Enter your API Key.
5. Select or type the Model identifier.
6. Optionally adjust Max Tokens and Temperature.
7. Click Test Connection to verify the provider responds.
8. Click Save to persist the configuration.

WARNING

The API key is stored encrypted on the server and is never exposed to the frontend. However, you should still rotate keys periodically and use keys with the minimum required permissions.

Provider examples

OpenAI

Field	Value
Provider Name	OpenAI
API Endpoint URL	https://api.openai.com/v1
API Key	Your OpenAI API key (starts with sk-)
Model	gpt-4o or gpt-4o-mini
Max Tokens	4096
Temperature	0.3

Self-hosted (Ollama)

Field	Value
Provider Name	Ollama
API Endpoint URL	http://localhost:11434/v1
API Key	ollama (placeholder, not validated)
Model	llama3.1 or your preferred model
Max Tokens	4096
Temperature	0.3

Custom / enterprise

Field	Value
Provider Name	Your provider's name
API Endpoint URL	Your provider's API base URL
API Key	Your API key or token
Model	Consult your provider's documentation

TIP

For air-gapped or high-security environments, use a self-hosted model via Ollama or vLLM. This keeps all data within your network and avoids sending workspace content to external APIs.

Workspace-level AI controls

After configuring the LLM provider at the instance level, you can control AI feature availability per workspace.

Setting	Description	Default
AI enabled globally	Master switch for all AI features	Off (until configured)
Per-workspace override	Allow workspace admins to enable/disable AI for their workspace	Enabled
Workspace allowlist	Restrict AI to specific workspaces only	Empty (all workspaces)
Workspace blocklist	Block AI for specific workspaces	Empty

Controlling AI by workspace

1. Navigate to Admin Panel > AI Configuration > Workspace Controls.
2. You will see a list of all workspaces with their current AI status.
3. Toggle AI on or off for individual workspaces.
4. Alternatively, use the allowlist or blocklist to manage access at scale.

WARNING

When AI is disabled for a workspace, all AI-related UI elements (assistant, suggestions, generation buttons) are hidden from users in that workspace. Existing AI threads and generated content remain accessible but new AI interactions are blocked.

AI usage monitoring

The AI Configuration page includes a usage dashboard that helps you track consumption and costs.

Metric	Description
Total requests	Number of AI API calls in the selected period
Total tokens (input)	Sum of prompt tokens sent to the LLM
Total tokens (output)	Sum of completion tokens received from the LLM
Average response time	Mean latency of AI API calls
Error rate	Percentage of failed AI API calls
Top workspaces	Workspaces with the highest AI usage
Top users	Users with the highest AI usage

Usage data can be filtered by date range and exported as CSV for billing or audit purposes.

Setting usage limits

To prevent unexpected costs, you can set usage limits:

Limit	Description	Default
Monthly token limit	Maximum total tokens per month	Unlimited
Per-user daily limit	Maximum AI requests per user per day	Unlimited
Per-workspace monthly limit	Maximum tokens per workspace per month	Unlimited

When a limit is reached, AI features are temporarily disabled for the affected scope until the limit resets.

Test connection

The Test Connection button sends a minimal prompt to the configured LLM API and verifies:

1. The endpoint URL is reachable.
2. The API key is valid.
3. The specified model is available.
4. A response is received within the timeout period.

If the test fails, the error message indicates the issue (network error, authentication failure, model not found, etc.).

Security and privacy

Concern	Mitigation
Workspace data sent to external API	Use self-hosted models for sensitive data
API key exposure	Key is encrypted at rest and never sent to the browser
AI-generated content accuracy	AI responses include a disclaimer; users must review before acting
Cost control	Use usage limits to cap spending

Context sent to the LLM

When a user interacts with AI features, SetGet sends contextual data to the LLM to produce relevant responses. Understanding what data leaves your instance is important for security and compliance.

Context type	When sent	Data included
User query	Every AI interaction	The user's message or prompt
Workspace context	Workspace-scoped queries	Workspace name, description
Project context	Project-scoped queries	Project name, description, member list
Issue context	Issue-related queries	Issue title, description, status, priority, labels, assignees
Cycle/Module context	Planning queries	Cycle or module name, date range, progress
Page context	Page-related queries	Page title and content
Conversation history	Follow-up messages	Previous messages in the AI thread

WARNING

If your organization handles sensitive data (healthcare, financial, classified), evaluate whether sending workspace content to an external LLM complies with your data handling policies. Self-hosted models eliminate this concern entirely.

Troubleshooting AI

Problem	Cause	Solution
Test connection fails	Wrong endpoint URL	Verify the API base URL format (e.g., must include /v1)
"Model not found" error	Model identifier incorrect	Check the model name against the provider's documentation
Slow responses	Model or network latency	Try a smaller/faster model, or use a closer endpoint
Responses cut off	Max tokens too low	Increase the Max Tokens setting
Incoherent responses	Temperature too high	Lower the Temperature setting (try 0.2-0.3)
429 rate limit errors	Too many concurrent requests	Set per-user daily limits or upgrade your API plan

Related pages

• Admin Panel Overview — Navigate the Admin Panel
• Feature Flags — Global on/off for AI features
• Security Settings — Data privacy and access controls
• General Settings — Instance-wide preferences