Settings
Configure SQL Server connections, AI providers and prompt rules, tuning memory, licensing, the local app lock, and analysis thresholds; changes save automatically.
Getting Started
Settings is the product-wide control surface for SQL Server connections, AI providers, generation behavior, prompt rules, tuning feedback memory, licensing, the local app lock, and analysis thresholds. Open it from the Settings item in the sidebar or from the gear icon in the top bar. Most users visit Settings during first-run setup and then return only when infrastructure, credentials, or workflow policies change. Settings is always reachable: until a license or trial is active, it is the only page available and the other modules stay hidden.
- Add, test, connect, and edit SQL Server connection profiles
- Add and test an AI provider and choose the default model
- Tune generation temperatures, the AI response timeout, and prompt rules
- Review tuning feedback memory from earlier AI-generated actions
- Activate a license, start a trial, or check license status
- Maintain your identity details and the local app lock password
- Choose which modules appear in the sidebar and adjust analysis thresholds
- General: Navigation Menu, Query Analysis, Security Audit, Blocking Analysis, Application Info
- Appearance: Theme
- Database: Connections, Query Settings, Cache
- AI / LLM: Providers, Generation, Prompt Rules
- Tuning Memory: Tuning Feedback Memory
- License: License Status, Advanced
- Security: Identity, Local App Lock
- A page tree on the left lists the pages above. Expand a page to jump to one of its parts; General opens first.
- The Search settings... box above the tree filters pages and parts by name and related keywords, and shows No matching settings when nothing fits.
- Most rows carry a one-line description under the label that says what the value controls.
- Changes are saved automatically. Checkboxes and lists save as soon as you change them; text and number fields save when you leave the field. There is no Save Settings button.
- A status line at the bottom of Settings shows Saving..., All changes saved, or Not saved with the reason and a Retry button.
- Two areas keep an explicit button: Save Prompt Rules in AI / LLM and Update Password in Security.
- Connections and AI providers are saved from their own dialogs, and actions such as Test Connection or Check License Now run immediately.
The most common onboarding sequence after installation is to configure Database first, then add one provider in AI / LLM, and only after that revisit thresholds or prompt rules. The Quickstart walks through that sequence.
Thresholds and Application Info
The lower half of the General page in version 1.1.0, with every value at its default. Navigation Menu and Query Analysis sit above this part; the local folder path is blurred.
- Backup Encryption Window: the last Security Audit field, 90 days of backup history checked for unencrypted backups.
- Blocking Analysis (Monitor and Severity Thresholds): both switches are on, so one blocking check runs in the background right after you connect and Blocking Analysis opens with its monitor refreshing. Monitor Refresh Interval is 5 s; the Low, Medium, and High thresholds are 5, 30, and 60 s of blocked wait time.
- Application Info: Version 1.1.0, Build, Author, and App Data Folder. Scrolling further shows Log File, Database File, View License Agreement, and Show Application Logs.
- Status line: Changes are saved automatically.
- Navigation Menu has a Main Menu group (Overview, Object Explorer, Query Statistics, Index Advisor) and a Tools group (Blocking Analysis, Security Audit, Scheduled Jobs, Wait Statistics). Clearing a box hides the module from the sidebar; it does not uninstall anything. At least one item must stay enabled.
- Query Analysis (Bottleneck Thresholds) sets how Query Statistics classifies queries. Defaults:
- Dominant Wait Threshold (%): 40
- Signal Wait Threshold (%): 15
- Storage IO Latency (ms): 15
- Log Write Latency (ms): 10
- CPU High (per exec, ms): 50
- CPU Low Reads (reads): 5,000
- IO High Reads (reads): 20,000
- CPU Low Reads must not exceed IO High Reads, otherwise the CPU-bound and IO-bound bands would overlap. These values affect interpretation inside the application, not SQL Server itself.
- Security Audit: Inactive Login Threshold (30 to 3650 days, default 90) and Backup Encryption Window (30 to 365 days, default 90).
- Blocking Analysis: Monitor Refresh Interval 2 to 300 seconds (default 5). Low, Medium, and High are 0 to 3600 seconds (defaults 5, 30, 60) and must satisfy Low ≤ Medium ≤ High; otherwise the value is not saved.
- Application Info shows where the application keeps its data, log file, and local database on this machine. Show Application Logs opens the latest log lines; see Where Secrets Are Kept below for what is masked there.
- Use threshold changes sparingly and validate on known workloads.
Theme
The Appearance page has one setting.
- Theme: Light, the only theme available in this release.
Adding a SQL Server Connection
The Add Connection dialog, opened from Database Connections, after a successful Test Connection. The profile is a demo entry.
- Profile: Connection Name, Server / Instance, Environment (here PROD - Production, shown with a red PROD badge), optional Port, and Default Database.
- Authentication: SQL Server Authentication with Username, an optional Domain, and Password. With Windows Authentication the current Windows session is used and no password is stored.
- ODBC Driver: Auto-select best available.
- Options: Encrypt Connection and Trust Server Certificate are both checked, so the connection stays encrypted but the server certificate is not validated.
- Result: Connection Successful! at the bottom, with Test Connection, Cancel, and Save.
- Database Connections lists the saved profiles by environment, then name, with Connect or Disconnect, Edit, and Delete.
- Test Connection uses the values in the dialog without saving them, gives up after at most 15 seconds, and shows the server, version, and edition on success. A typed password is used for the test only; Save stores it.
- Editing a connected profile in a way that changes how it connects (server, port, database, authentication, username, domain, driver, or the encryption options) disconnects the live session.
- General Query Settings: Query Timeout (1 to 600 seconds, default 30), Jobs Long Running Threshold (1 to 1440 minutes, default 30), and Jobs Failed Export Limit (default 200).
- Cache: Clear Cache empties the local application cache, and Cache Info shows what it holds. Use it for freshness troubleshooting, not routine cleanup.
AI Provider Management
The Providers page with two configured models.
- Default Ollama: type Ollama, model codellama. This entry is added automatically if the last provider is removed.
- DeepSeek: type Deepseek, model deepseek-v4-pro, marked with the star as the default.
- Actions: Add AI Model at the top; Set Default and Edit below the list. Double-clicking a row also opens Edit, where a provider can be removed.
Adding and Testing a Model
The Add AI Model dialog after Test.
- Provider: one of Ollama, OpenAI, Anthropic, Azure OpenAI, or DeepSeek. Choosing a type suggests a model name.
- Name, Model, API Key: the key is masked and can be revealed with Show. Ollama asks for a host instead of a key.
- Test result: DeepSeek API key valid, the model that was tested, and the models the key can use.
API keys are not written to the settings files; they are kept in Windows Credential Manager. Provider changes are saved when you click Add or close Edit.
Generation Temperatures
The Generation page with the default temperatures. Each value is 0.00 to 2.00; lower values give more deterministic output. The capture leaves out the first row, Default (0.10), which applies to workflows without their own setting.
- Core Analysis Temperatures: Object Analysis 0.02, Query Analysis 0.10, Code Optimization 0.10.
- Advanced Workflow Temperatures: Index Advisor (Preclassified) 0.10, Self-Reflection Refinement 0.05, AI Safety Validation 0.00, SQL Continuation 0.10, Report Continuation 0.10.
- Runtime (below the capture): AI Response Timeout, 10 to 1800 seconds, default 900, is the upper bound for one AI analysis; raise it if analyses finish as partial. Batch Pause Between Analyses, 0 to 3600 seconds, default 10, is the wait between two object analyses in a batch run and can be overridden per run.
Prompt Rules
The Prompt Rules page with the Global rule open. Prompt rules shape the AI analysis output; this is an advanced area for repeated quality problems that justify a prompt-level change.
- Rule list: Global, Query Analysis, SP Analysis, SP Code Only, Index Preclassified, Object Analysis, and Self-Reflection.
- Editor: Global holds shared instructions inserted wherever another rule uses {global_instructions}. Other rules have system and user prompt fields.
- Placeholders: the values a rule can use, shown under the editor; None for this rule here.
- Saving: not automatic. Edits are kept while you move around Settings and apply only after Save Prompt Rules; Reset Prompt Rules returns to the built-in templates.
Tuning Feedback Memory
Tuning Memory before any action was recorded. It keeps the review history of AI-generated Object Explorer actions in the local database on this machine.
- Object Name: searches by object name, schema, or database.
- Columns: Date, Object, Type, Priority, Outcome, Block, Confidence, Fingerprint, and Action.
- Buttons: Mark Hard Block, Mark Improved, and Reset to Suggested for the selected entry, with its details in the panel below.
It is a review surface, not an execution engine: outcome, block state, and confidence are operator guidance that keeps earlier decisions visible.
License Status and Advanced Details
The License page on an activated machine with Advanced expanded. Device ID, Server URL, and the configuration folder are blurred.
- License Status: Active, with Change License (Activate License before activation) and Check License Now.
- Cards: Expires, Last Validated, and Licenses / Devices 1 / 1.
- Advanced: Device ID with Copy ID, Server URL, Installed License, Imported .lic File, App Config Folder, Refresh After, and Offline Grace Until.
- Buttons: View License Agreement and Remove License From This Device.
Server URL is the only field on this page that saves itself, when you leave it; change it only if support asks you to.
Choosing How to Activate
The first step of the Activate License wizard: Method, then Details, then Result.
- Online: enter the activation code from your website account, or sign in with your website password and let the application fetch one for this device.
- Manual (.lic file): for offline installations, import a .lic file issued for this device.
Entering the Activation Code
The Details step for online activation; the email address and code are blurred.
- Email address and Activation code: use the email of the website account that owns the code, then press Activate.
- No code yet? Get one with my website account: asks for the website account password and requests a code for this device. The password is not saved.
- Free trial: the wizard offers it only while a trial is still available on this machine. It lasts 30 days and each machine can use it once.
- If activation fails, the Result step shows Activation could not be completed with the reason.
- Remove License From This Device clears the saved license on this machine; trial dates are kept.
- Licenses are counted per device: see Pricing.
Identity and Local App Lock
The Security page with the app lock on; the email address is blurred.
- Identity: Name Surname, Company Name, and Email Address. These fields are shared by the local app lock, trial activation, device ID generation, and online license activation.
- Local App Lock: Enable local app lock asks for the local password when the application starts; Configured shows Yes once a password is set.
- New Password and Confirm Password: take effect only after Update Password; everything else on the page saves when you change it.
- Automatic Sign-In: Disabled here. Clear Remember Me turns it off so the next start asks for the password again.
An email address is required while the app lock is on, and the password must be at least 6 characters. If the local password is forgotten, use Forgot Password on the sign-in screen and verify the website account.
- SQL Server passwords and AI provider API keys are kept in Windows Credential Manager, not in the settings files.
- The local app lock password is stored as a salted PBKDF2 hash, not as plain text.
- Show Application Logs displays the latest 100 log lines with user names, passwords, tokens, and API keys masked. Server names and SQL text are not masked, and the log file on disk is not changed.
- Use Settings > Database to add a connection, test it, save it, and click Connect before opening the analysis modules.
- Use Settings > AI / LLM > Providers to add and test one stable provider before changing prompt rules or temperatures.
- Use Settings > License to check trial or licensed state, and open the activation wizard when you receive a license.
- Use Settings > Security to keep your identity details current and to change or disable the local app lock password.
- Use Settings > General after your first successful analysis if you need sidebar visibility changes or threshold tuning for Query Statistics, Security Audit, or Blocking Analysis.
- Keep one stable default AI provider instead of rotating defaults frequently.
- Use descriptive connection names and set the Environment so production servers are easy to recognise.
- Change Query Analysis and blocking thresholds gradually and validate the result on a known workload.
- Use Tuning Memory as an operator review log, not as an automatic truth source.
- Keep Encrypt Connection enabled for SQL connections unless your environment requires an exception.
- Treat the local app lock password as a real machine-level access control.
- Watch the save status line at the bottom of Settings; if it shows Not saved, correct the value or click Retry.
Use Installation for first-run onboarding and SQL prerequisites, Quickstart for the shortest path to a working setup, and Security for what each module sends to an AI provider.