1. Docs
  2. Administration
  3. Project settings
  4. AI Options

AI Options

AI Options decides whether a project uses AI at all, whose AI account it uses, and which AI features are switched on: locators, assessments, auto-heal, document analysis, failure analysis and report feedback.

3 min read Updated CustomersAdmins

In short

On this page you can

  • Switch all AI features of the project on or off with Enable AI.
  • Use the SmartestQA AI account, or your own Gemini, OpenAI, Anthropic or local model.
  • Check a provider key before saving with Test Connection.
  • Turn single capabilities on or off, such as Auto-Heal or AI Failure Analysis.
  • Keep sensitive data out of AI requests with Share Sensitive Data off.

How to use it

  1. Open Configuration › Project Panel › AI Options.
  2. Turn on Enable AI; the other blocks appear.
  3. Under AI Account keep Smartest AI, or pick Own Account and fill provider, model and key.
  4. Click Test Connection and wait for “Connection successful”.
  5. Set the capability switches and click Save Changes.

Enable AI#

Enable AI (block AI Features) is the main switch. New projects have it on. When it is off, every AI feature of the project is off, the other two blocks are hidden, and AI buttons elsewhere show a tooltip such as “Enable 'Enable AI' in Project Panel → AI Options”. Tests without AI steps still run normally.

AI Account#

OptionWhat it meansYou fill
Smartest AISmartestQA’s own AI account. Model, token budget and temperature are set by SmartestQA. Default for new projects.Nothing
Own AccountYour company’s provider account. AI requests of this project go to your provider, with your key.Provider, model, key (see below)

Use your own account#

FieldWhat to enterRules
AI ProviderGoogle Gemini, OpenAI (ChatGPT), Anthropic (Claude), or OpenAI-compatible / Local (Ollama, vLLM, LM Studio...).New projects: Google Gemini.
AI ModelThe model name exactly as the provider writes it. The grey hint (gemini-..., gpt-..., claude-..., llama3.1:8b) is only an example.Empty = the provider’s default model. Changing the provider clears or replaces a model of the old provider; your own model name is kept.
Base URLThe provider endpoint. It is pre-filled with the provider’s default address.Leave the default to follow the provider. Required for OpenAI-compatible / Local, for example http://localhost:11434/v1.
API KeyYour provider key.Stored encrypted. After saving it shows as a fixed mask; typing replaces the whole key, leaving the mask keeps the saved key. Optional for local providers.
Max TokensThe longest answer the model may give.1–8192. New projects: 2000.
Temperature“Creativity (0-1)”: low values give steadier answers.0–1. New projects: 0.7.
  1. Fill provider, model and key

    Pick Own Account, then the provider. Paste the key in full.

  2. Test before saving

    Click Test Connection. It uses the values on screen, saved or not (“Testing...”).

    “Connection successful”, or the model’s short reply. A failure shows one sentence and Technical details; see AI connection errors.

  3. Save

    Click Save Changes. AI features of the project pick up the new account at once; pipelines take it from their own copy (see How settings reach a run).

A local model on your own network

With OpenAI-compatible / Local, the SmartestQA server must be able to reach the Base URL. localhost means the server itself, not your laptop. Use an address that the SmartestQA server (or your on-premise installation) can reach.

AI Capabilities#

Each switch turns one AI feature on or off. All are on in new projects. They count only while Enable AI is on.

SwitchScreen textWhat it turns on
AI for LocatorsUse AI to find element locatorsAI help in the browser extension while recording: naming controls and choosing locators.
AI for AssessmentUse AI to evaluate assertionsAI Insights on the Dashboard.
Auto-HealAuto-heal locatorsThe runner repairs a locator that stopped matching; repairs wait for review in Maintenance.
AI Document AnalysisAuto-generate test cases from uploaded analysis documentsDocument analysis in Analyze, AI steps and the AI assistant (prompt to steps, expand) in Automate.
Share Sensitive DataSend sensitive data to the AIOff: sensitive values are masked before a request leaves for the AI.
Collect for TemplatesSave repeated actions to build reusable templatesCollects repeated action sequences so they can be suggested as templates.
AI Failure AnalysisAuto-analyze failed steps with AIThe AI analysis of failed steps in reports.
AI Report FeedbackAI review of this run's results on the report summaryThe AI review card on the report summary.

Code generation in JavaScript and Python steps needs only Enable AI.

What is sent to the AI#

  • Only what the feature needs: for example a failed step and its error for failure analysis, page descriptions and notes for test generation.
  • With Own Account the data goes to your provider under your agreement with them.
  • With Share Sensitive Data off, values recognised as sensitive are masked first.
  • Use test environments and test data in any case; screenshots and texts may reach the AI.

If something goes wrong#

Didn’t find your answer?Write to support with your project name and what you see on screen.

info@smartestqa.ai
© 2026 Selenity Technology · SmartestQAsmartestqa.aiThis page as MarkdownChecked against smarttest-fe predev 2a8c79cc, smarttest-be 633c2a03, smarttest-testsuite 017fcec, recorder extension