No row of your firm's data is ever sent to the AI provider. CaseQuery never connects to your database itself. Your firm's connector does: it holds the database login, opens the database read-only, checks every query before it runs, and connects out to CaseQuery. Query results go from the connector, through the CaseQuery service, to the browser of the person who asked. The service does not keep them.
What does go to your firm's AI provider, to plan each query: the question (with up to 12 earlier turns of the same conversation), the database's structure (table and column names and types, with your sensitive columns taken out), and your firm's own rules, definitions and notes. A database's error text is not sent; the plain-English answer is put together by CaseQuery from the plan and the result. One thing to know: a column note may list the codes a column holds (for example A, C, P), whether you wrote it or accepted it from preparation, and a note travels as a note.
What the service keeps for your firm: what your firm writes (business rules, definitions, column notes, Schema rules, starter questions, settings), your people and their roles, and your firm's audit journal. Never rows or results from your database; a value from it is kept only where your firm chooses to keep a code in a column note, settled definition or rule. Columns you hide under Schema rules never leave your firm's network. Chat history is kept only in each person's own browser. Your AI provider key is kept encrypted by the service and is never shown again, only its last four characters.
The service is hosted by NetDirector on Microsoft Azure in the United States (East US 2), encrypted in transit and at rest, and each firm's people, records and connector are kept apart from every other firm's.
Nothing is installed on staff computers. Everyone uses CaseQuery in a modern web browser, at your firm's CaseQuery web address. The only piece your firm may run is the connector, on a server near your database (see 1.4); if your database is in a cloud service, NetDirector can run the connector for you instead.
There are no shared keys in the cloud edition. Each person has their own sign-in (their email address) and their own password, and one of three roles:
| Role | What it opens | Who holds it |
|---|---|---|
| User | Chat: asking questions, voting on the floor, sending corrections, Casey. | Everyone who asks questions. |
| Lead | Everything a user can do, plus the lead tier on Fine-tuning: settling definitions, marking baseline answers, handing questions to the DBA. It does not open Settings. | Operations Leads. Optional: a firm with no separate lead tier can leave it out. |
| Administrator | Everything a lead can do, plus every tab under Settings, writing business rules and terms, confirming or withdrawing settled definitions, the DBA screens, and managing people. | Your IT team or nominated administrator. Your firm always keeps at least one. |
Administrators manage people on Settings › Security (see 3.2).
A session ends after 30 minutes without use, and after eight hours in any case. Repeated failed sign-ins are limited, and after that even the right password is refused for a while.
Settings › Setup checklist lists what to do, in order. The heading reminds you that each step builds on the one before, and to pair your firm's connector first: until it is, nothing else there can be checked.
Each step shows one of two labels on the right:
Open beside a step title takes you to the screen where it is done. The progress line counts steps done. The business-rules step is optional and does not hold the checklist open.
Why: the connector is the only way CaseQuery reaches your database. Nothing else on the list can be checked until one is connected.
How: Open goes to Database. Pair a connector as described in 1.4. If the connector offers more than one database, choose the one to work in under Which database. The step ticks when a connector is connected and a database is chosen.
Why: nobody can sign in until they have been added, and a role decides what they may do. The step shows how many people have a sign-in and how many of them are administrators.
How: Open goes to Security. Under People at your firm, add each person with their role (see 3.2). Consider a second administrator, so that someone else can reset your password if you lose it. Then confirm.
Why: questions cannot be answered until your firm's key is saved. One key serves everyone at your firm; nobody needs to enter one of their own.
How: Open goes to AI provider. Choose the provider, paste the key, and press Save (see 3.4). The step ticks when a key is in force, and names the provider and model.
Why: a hidden column is never read when the database is prepared, so nothing about it is ever written down. And until you save the Schema rules for a database, none of its data is read at all: only its structure, even if nothing needs hiding.
How: Open goes to Schema rules. Review what is flagged, tick only columns nobody should see, and press Save & lock policy (see 3.8). Hiding nothing is a legitimate choice, because a hidden column also stops the AI using it, but save either way. Then confirm. The step is confirmed database by database: another database the connector offers has not been reviewed just because this one has.
Why: preparation describes the coded columns so the AI does not guess at them, and lists the definitions your firm needs to settle. Answers are unreliable until it has run.
How: Open goes to DB Tuning › Prepare the database (see 2.2). Your firm's connector does the reading, where your data is. The step ticks when preparation has run, and says who ran it and when. If the database's structure changes later, the step opens again and asks you to prepare it again.
Why: leaving the defaults is a real decision, and it looks identical to never having opened the screen. The step shows the current row limit and whether question checking and safe response mode are on.
How: Open goes to Safeguards (see 3.3). The defaults are sensible. They apply to everyone at your firm. Confirm when you are happy with them.
Why: Casey explains the product and uses your AI provider account. Whether Casey may also read what your firm has written down is a separate permission people miss.
How: Open goes to Assistant (see 3.5). Decide both switches, then confirm.
Why: rules are things only your practice knows, such as what counts as referred. They come from the teams using the product, not from whoever sets it up. Invented rules are worse than none: they steer every future answer and nobody remembers agreeing to them.
How: nothing to do now. Tick it and leave it. Rules arrive later when someone meets a number that looks wrong and reports it from the answer. Open goes to Business Rules.
The connector holds the database login, opens the database read-only, checks every query itself, and connects out to CaseQuery over HTTPS. No firewall port is opened inbound. There are two ways to run it.
| Where the database is | Who runs the connector |
|---|---|
| On your own network, such as a server in your office | Your firm, on a Windows server near the database. It reads Microsoft SQL Server and Azure SQL, PostgreSQL, MySQL and MariaDB, and Oracle. |
| In a cloud service | NetDirector, on Azure, kept apart from the web service. It reads PostgreSQL today; other engines on request. The database must let the connector's addresses through its firewall. |
C:\Program Files\CaseQuery Connector, checks every file, and opens a setup form in the server's browser.The form pairs the connector with your firm, tests the database read-only, seals the password to that server (it is never sent to CaseQuery and never written in plain text), and sets the connector running on its own: it starts with Windows and restarts if it stops. In Task Scheduler it is listed as CaseQuery Connector. Nothing else needs installing; it brings its own runtime.
The connector then shows as connected under Connect your databases. Before anyone asks a question, save the database's Schema rules (checklist step 4).
For IT staff who prefer it, or who need a proxy: the CONNECTOR_INSTALL.md file in the download gives the console route. In a console opened as Administrator in the connector's folder, connector setup pairs the connector and adds each database (each is tested read-only before it is kept), connector doctor --as-service checks everything as the connector will run, and connector install-service sets it running on its own. A proxy that decrypts traffic is refused, by design.
On the connector's server, in a console opened as Administrator:
| To | Run |
|---|---|
| See what it is doing | connector status, and its log at %ProgramData%\CaseQuery Connector\runtime\logs\connector.log |
| Check everything again | connector doctor --as-service |
| Offer another database | connector add-database, then restart the connector (Task Scheduler › CaseQuery Connector › End, then Run). Then save that database's Schema rules. |
| Stop offering one | connector remove-database <name>, then restart the connector |
| Stop it answering anything, at once | connector off, and connector on to undo. This is your firm's switch: CaseQuery cannot turn it back on. |
| Check its statement record is intact | connector log |
| See which version is installed | connector version |
| Remove it | connector uninstall-service. Its settings folder is kept; delete %ProgramData%\CaseQuery Connector yourself to remove its identity and sealed logins. |
The connector does not update itself. To upgrade, follow "Upgrading" in CONNECTOR_INSTALL.md, and keep %ProgramData%\CaseQuery Connector: it holds the connector's identity, its sealed logins and its records.
From CaseQuery's side, Revoke beside the connector on Settings › Database stops it at once; your database cannot be reached until a connector is paired again.
With it, every request that reads values needs a one-time ticket, fetched by the person's own browser from a "ticket desk" on your firm's network. A stolen sign-in used from outside your office or VPN then still cannot read data. Reading the database's structure (table and column names) does not need a ticket.
It is set up once by your firm's IT: an internal name for the desk, a certificate for that name that your PCs trust, a firewall rule letting office PCs and the VPN reach it, and the desk switched on in the connector's settings, then a restart of the connector. "Switching on the office-network check" in CONNECTOR_INSTALL.md, and HOW-IT-WORKS.html in the same download, give each step. Without it, protect sign-in with Microsoft sign-in and with the list of addresses people may sign in from (see 3.2); the connector's limits and off switch remain the backstop. Trials do not include the office-network check.
| Guide | Give it to |
|---|---|
| Getting Started | Everyone, first |
| Admin / DBA Guide (this guide) | Administrators, and the DBA who works with them |
| Lead Guide: Fine-tuning | Operations Leads |
| User Guide | Everyone who asks questions |
| Working with Casey | Everyone, if Casey is offered |
Six kinds of tuning shape answers. They are different tools with different reach. Each belongs to the database it was made for.
| What | Where | Who writes it | What it changes in an answer |
|---|---|---|---|
| Column notes | Schema rules; written first by preparation | Administrator | How the AI reads an abbreviated or coded column. Sent with questions as part of the schema. |
| Business rules | Business Rules › Rules | Administrator | A standing instruction ("always do Z"). Applies to every question it concerns, including ones nobody has asked yet. |
| Terms | Business Rules › Terms | Administrator | What a phrase means ("when someone says X, they mean Y"), optionally with the exact condition and tables. |
| Settled definitions | Fine-tuning › Definitions to settle | Lead settles; administrator confirms or withdraws | Which reading of an ambiguous word or measurement the firm means. Applies from the moment it is settled, exactly as a rule does. |
| Baseline questions | Fine-tuning › Baseline questions | Lead marks answers | Nothing directly. A record of which answers were right on a given date, and a source of work: a Wrong answer becomes a definition to settle, and a Right answer is offered to the DBA as a pair to settle. |
| Verified queries (Question-SQL pairs) | DB Tuning › Question-SQL pairs | Administrator, with the DBA | That exact question is answered by running your query, word for word: no AI call, no wait, and the same query every time (the figure follows your data as it changes). |
| Ground truth | DB Tuning › Ground truth | Administrator | Nothing directly. A check of answers against figures your firm already trusts, run when you choose. |
Up to a dozen rules, terms and settled definitions together are sent with every question. Past that, each travels with the questions it concerns: a rule naming a table goes with questions about that table, and a rule naming nothing is always sent. Tick Always apply on a rule to send it with every question whatever it names — for a firm-wide rule that happens to mention one table.
Go to Settings › DB Tuning › Prepare the database. Preparation reads the database once. It describes the coded columns so the assistant is not guessing, and finds the definitions your firm needs to settle. It is done by your firm's connector, against your database, and computed with SQL: no AI tokens. Table data is not sent to the AI, and what the connector reads stays with it: only the column notes you apply and the relationships it proved come back to the service. The profile holds structure, counts and characteristics, not your records.
.sql, Word .docx or Excel .xlsx). Press Read it. Only the documentation is sent through your AI account, never database records. The report shows where your documentation and your database disagree, which statements the data confirmed, which were taken from your documentation without a check, and which could not be checked. The report is for you to act on; nothing from it is saved.Press Prepare Database. Preparation times each query and eases off while the database is busy, so it is safe on a live system and quicker outside peak hours. Stop ends a run without confirmation, because nothing is written until preparation is applied. When it finishes, read the report before anything is written.
The badge beside the heading reads Not run, Prepared, Skipped or Schema changed — run again.
After a new status value or a large data load, Check for new definitions, on the same tab, re-reads the database for words it could read more than one way and refreshes the leads' list. Nothing already settled is undone or asked again, and it uses no AI tokens.
Leads settle or set aside definitions, mark baseline answers, run baseline sections, add baseline questions, use Check the wording and See what this changes, and hand questions to the DBA with Send to the DBA. They cannot open Settings, save rules or terms, or confirm or withdraw settled definitions. Their day-to-day work is described in the Lead Guide: Fine-tuning.
Every recorded decision is recorded under the sign-in of the person who made it; nobody types a name.
Go to Fine-tuning › 3 · Awaiting sign-off. It lists the definitions a lead has settled that are already in force and still waiting for an administrator's signature. Confirmed definitions are on Business Rules › Rules › Definitions in effect from Fine-tuning, where a confirmed one can be withdrawn; withdrawn and cleared ones are on Business Rules › History.
| State | Meaning |
|---|---|
| provisional | Settled, not yet reviewed. It applies. |
| stale | Unreviewed for over 30 days. It still applies. |
| archived | Unreviewed for over 90 days. It still applies. |
| active | Reviewed and confirmed by an administrator. |
| withdrawn | No longer applied. |
For an administrator, the unreviewed ones are pinned at the top under a heading saying how many definitions the firm has not stood behind. For each:
Confirming is a second signature, recorded under the administrator signed in, because the point is that the firm, not only the lead who settled it, stands behind it. The same definitions also appear on Business Rules › Rules › Definitions in effect from Fine-tuning, marked In force · awaiting sign-off, because they already apply. A header badge appears on every screen while reviews are owed.
Business Rules › Has it worked? is observed from questions people actually asked. It is advisory only: nothing here changes an answer.
A question settling shows the wording stopped being ambiguous. It does not prove which rule did it.
Floor voting asks the people doing the work what an ambiguous word means, instead of the product choosing quietly. It is off by default. Like every setting on the Safeguards tab, it is your firm's: the same for everyone at it, whichever computer or browser they use.
Go to Settings › Safeguards › On Floor Voting and switch on Ask the floor when a question is ambiguous.
A vote appears under an answer only when both are true:
It reads, for example, Counted "open" as Open. Is that right?, with a button per value and Not sure. A line says who will see the vote.
Nothing is asked about "anything except", "one of" or "like" filters, numbers or yes/no flags, or queries filtering more than one status column. A column with 7 or more values, or a person who has switched asking off, gets only the line Counted "active" as Active. with nothing to vote on. Questions with a Verified query never ask.
| Setting | Default | What it does |
|---|---|---|
| Votes needed to settle it | 4 (2 to 50) | When that many different people agree, it becomes a provisional settled definition that applies straight away. You then confirm or withdraw it under Awaiting sign-off. A small firm may only have three or four people who would know. |
| Stop asking after | 10 (2 to 200) | Once this many people have answered without agreement, nobody else is asked and it goes to the leads. Only answers count; being shown the question does not. |
After voting, the person is told how many more matching answers would settle it.
A question goes to Fine-tuning › Definitions to settle, under Floor needs a ruling, only when:
While votes are still being collected and everyone agrees so far, there is nothing there for a lead to act on. The ruling row names who voted which way, so a lead does not simply approve the bigger number. A lead presses one reading to settle it, and it can be withdrawn under Awaiting sign-off. Not sure counts towards stopping the asking, never towards settling.
Everyone signs in as themselves in the cloud edition, so every vote is recorded under the voter's sign-in name, and each person counts once.
Off by default. While off, each person sees an Ask me what words mean tick box beside the question box and a Stop asking me link under a vote, so someone who never wants to be interrupted can switch asking off for themselves. They still see which reading each answer used. Turn it on to ask everyone, with no way to opt out.
Anyone can read Business Rules. Only an administrator can save. The screen has tabs: Rules, Terms, History, Has it worked? and How this works. It opens on Rules, which has two sections, both folded until you open them, each with its count. Rules in effect by Administrator is your own policy on how to count, edited here: the list, with Always apply and Remove on each, the text editor, the save buttons, and where each rule came from. It opens by itself when you add a rule or a save needs your attention. The count is what is saved; a rule you have added but not saved is not yet in effect. Directly below it, Definitions in effect from Fine-tuning lists every definition in force, marked awaiting sign-off or confirmed: they reach the AI exactly as rules do, but they are decisions about your data settled on Fine-tuning, so they are changed only by withdrawing. A confirmed one has Withdraw it; one awaiting sign-off has Open in Fine-tuning. Save written rules and Clear written rules never touch settled definitions or terms. History holds withdrawn definitions and earlier versions of the rules; Has it worked? includes how floor voting is going.
The screen's own advice is Write definitions, not answers. A good rule defines terminology and conventions: what a term means, how to count, which records to exclude. Avoid recording conclusions. Telling the AI what it should find stops it finding anything you did not already know.
| Defines (good) | Answers (avoid) |
|---|---|
| Any status beginning 'On Hold' is held, not active. | We have 412 active files. |
| Always count distinct record identifiers after joining detail tables, never raw rows. | The Smith matter is the oldest open file. |
| Our fiscal year starts on 1 July. | Revenue this year is higher than last year. |
Limits: up to 200 rules, 400 characters each, about 30,000 characters in total. Anything too long is refused and named, not shortened for you. Edit all as text is for pasting several at once or reordering. Clear written rules removes them all; settled definitions and terms are not affected.
A rule can use wildcards: % for any run of characters, _ for exactly one. You do not have to use them; "match the name anywhere, not just the beginning" works too. Whether a match ignores capitals differs by database, so say so in the rule if capitals matter.
When Check rules against a second model is on (under AI provider), a second model plans the same question when a rule is previewed. If the two write different queries, the wording still admits more than one reading.
Open Terms, open Define a term (Administrator) (folded, and open by itself when no terms exist yet), and fill it in:
FileStatus IN ('Active', 'On Hold')) and Tables it applies to. Naming tables keeps the term out of unrelated questions; leaving it blank sends it with every question.Press Add the term, then Save terms.
History keeps the last ten versions of the rules, saved automatically before each change, with when and by whom. A save that empties the rules is never kept, so it cannot push a good version out. Restore replaces the rules on screen with that version, after you confirm. The version being replaced is kept as a backup, so a wrong restore is recoverable. Nothing in History is sent to the AI.
Go to Settings › DB Tuning › Question-SQL pairs. A pair joins a question your firm asks constantly to the query that answers it. From then on that question is answered by running your query, word for word: no AI call, no wait, and the same query every time, so the figure changes only when your data does.
In chat, an answer from a settled pair carries Verified query with the name and date of whoever settled it, and costs nothing from your AI account. Floor voting never asks on these questions. If nobody has re-checked the pair for a long time, the mark says how many days.
Settled pairs are listed under Settled questions, with the query, who settled it, and any attached wordings.
Anyone can type @ in the question box to say which tables a question is about. A short list appears as they type, and choosing one puts its name into the question. Nothing shows until the @ is typed, so it stays out of the way of people who do not want it. This is the feature most likely to be useful to you and to a DBA who knows the schema: it stops the product guessing which of three similarly named tables was meant.
Beside each table, Often needed with it offers the tables it is joined to. Those come from two places and the tooltip says which: a key your database declares, or a join that queries run here have actually used — which preparation recorded. It is never a guess about what tables are "usually" used together, because this product does not measure that.
It is a preference, not a restriction. If the question genuinely needs another table it is still used, and the answer names what it added. This is deliberate: refusing to answer teaches the user nothing, and honouring a scope that cannot answer produces a confident figure about the wrong population. There is no strict mode.
Three things worth knowing when somebody asks you about it:
Sales.invoices and Archive.invoices — typing @invoices asks which they meant rather than choosing one. Naming it in full works.On a database small enough to send whole, naming tables changes what the AI is asked to concentrate on, not what is sent — the schema travels either way. On a large one, where only the relevant part of the schema is sent with each question, it changes both.
Go to Settings › DB Tuning › Ground truth. These are figures your firm has already established as correct, from reports it already runs and trusts. The system answers each question and shows which figures it reproduced exactly. Nothing is ever edited to match an answer: if a figure and an answer disagree, that is the finding. The figures are what your firm asserts, so the service keeps them as firm-written content.
A figure belongs to the database it was reported from. If you switch database with unsaved drafts on screen, they are dropped and the screen says so. Saved figures can be changed in the grid and saved with Save figures.
Advanced: edit as text holds every case, including settings the grid does not show: which column carries the figure, label aliases, how close counts as equal, and whether rows your report leaves out are expected. It saves only if every case is valid, and tells you everything wrong at once.
Each question is planned, billed and recorded exactly as if someone had typed it. Asking 3 times costs three times as much. The planner does not write identical SQL every time, so a single pass cannot tell a real fix from a lucky one. Use 3 or 5 after a change you want to prove. The Last check column shows each result, and a line gives how many matched exactly. Re-run after any change to your business rules.
Go to Settings › DB View Assistant. It helps you design a database view for questions that keep joining the same tables. It writes a script; it never runs it.
It lists table combinations worth a view, from the questions answered at your firm on the chosen database and the tables their queries used. A combination is considered once at least five answered questions, and at least 5% of them, have used the same set of two or more tables, and is listed where a view would help: a join that had to be inferred rather than read from a declared key, answers that varied, or three or more tables. Combinations already settled are left out. Each shows the joins, whether each is declared or proved by reading rows, and where there are several child rows per parent. An empty list on a firm that has only just started is normal.
The counts behind this list hold the database, table names, counts and dates, not values. The card says where the counts come from.
Press Use these tables for a draft rather than a blank form. The recommendation knows the tables and how they can join, but not which columns, which rows or how they should be totalled, because it keeps no SQL. So the draft selects the tables, ticks the columns they join on (unless the hidden-column list could not be read, when it says so and ticks none), and offers a suggested description you adopt with Use this description or replace. It then asks two things, and Estimate & Generate View Script waits until both are answered:
When no join between the tables is known yet, there is nothing to choose from, so you write both instead: a line starting Join: in Filters and business rules, and Row grain. The button waits for those.
Your choices fill Row grain and a Join: line in Filters and business rules. Everything else is yours to add or leave. Choosing other tables drops the draft.
Give the script to whoever creates objects in your database, usually the DBA. They review it and create the view with their own rights; the connector's login can only read and cannot. Once it exists, press I have created it — re-read the schema on the recommendation, or refresh the object list on Schema rules. The new view is then available to questions. On Schema rules, Use in the DB View Assistant takes any table or view you are inspecting into this form.
Go to Settings › Security › Import & export tuning. The service keeps point-in-time backups of its own records for 35 days; this copy is your firm's own, kept where you choose.
Under Take a copy of your tuning work, press Download the copy. One file is saved to your computer, holding everything your firm has taught CaseQuery: business rules and terms, settled definitions, column notes, the questions you have checked, your reference figures, your preparation and your settings. The screen names the file and its size.
It contains no AI provider key or other secret, no connector registration, and no rows or results. It does hold what your set-up says about your database: table and column names, your notes on them (and any codes those notes or your settled definitions quote), the ground-truth figures your people typed, which columns are withheld, and what preparation measured of its shape. Your people and their roles are listed in it for reference, and are never restored from it. Keep it as carefully as you would your rules themselves.
Restore puts back business rules, settled definitions, column notes, the sensitive-column policy, preparation and settings. Only your own firm's copies are taken.
A user clicks Not right? beside a measurement under How this was measured. In Correct this measurement they pick anything under What went wrong?, write What should it say?, and press Send for review. It is recorded under their sign-in name. It is not sent to the AI and does not change the answer.
The correction goes to the top of Fine-tuning › Definitions to settle with the badge reported wrong, where a lead can settle it as a definition or send it to the DBA.
The amber Worth checking panel offers Help me sort this out…, a guided flow using Casey that ends in a definition, a correction or a handover depending on who is asking.
Two records, kept in two places, so neither side has to be taken on trust.
connector log checks that it is intact. It shows an edit; it cannot show its last entries, or the whole record, being deleted, so protect the server as you would the database.A link is one or more fields that tie your data together: the same reference kept in two of your databases, so that one question can be answered across both. One database might hold your records and another the invoices raised against them; a link says which field in each holds the shared reference, and how to read it. CaseQuery looks at both databases where they are, proposes the fields that tie them with its evidence, and you check and confirm.
Linking is not the same as Connect (3.6). Connect adds up counts across databases that hold different records. A link joins related records, so a question can count or total one database's records by what is true of their counterparts in the other. The two are set up separately and neither changes the other.
Nothing in linking is written for one kind of business. The words for your kinds of record come from a starting vocabulary you choose and can change, and a link is proved from your data, not from what its fields are called; names also help choose and screen the fields to look at, and inform the set-up proposal.
"linking": { "scan": true, "accept": true, "ask": true, "assist": true } to databases.json in %ProgramData%\CaseQuery Connector, then restarts it (Task Scheduler › CaseQuery Connector › End, then Run). Until then, the card says your firm's connector needs updating.On Settings › Database, Link your databases comes after Add a cloud database when that is shown, and before More than one database. It is folded like the others: press its heading to open it. Folding it again clears from the page anything read from your databases.
The first time, the card is five numbered steps. Only the next step to do is marked Next; a later one says in one line which step comes first, and a finished one is ticked and stays usable.
Under The words your firm uses, choose a starting vocabulary under Start from and press Use this vocabulary: the kinds of record your databases hold, and what each one's reference is called. Two are offered: General, one kind, Record, whose reference is called Reference, a sound start for any business; and Law firm, an example of a ready-made vocabulary for one kind of business, starting with File, Invoice, Bill and Time entry.
The vocabulary is your firm's own from then on. Edit the kinds shows each Kind of record and what Its reference is called; change them to the words your people use. Add a kind of record adds one, up to 32; Remove takes one away, as long as at least one is left and no accepted link is of that kind. Press Save the kinds, or Cancel. The names are used in the wording of answers and are given to the AI as your firm's words. Nothing in the product decides anything from them.
Choose One database and The other, and press Use these two. They must be different. A link joins exactly two databases; to tie a third, make a second link. You can choose two others at any time, which sets the later steps back. If the connector offers only one database, the card says linking needs two.
Nothing of a database's data is read until its Schema rules are saved, so the connector will not look for links in it until then. A database whose rules are saved is ticked. For each that is not, Open Schema rules for its name makes it the database you work in and opens Schema rules: hide any column nobody should see and press Save & lock policy, even if nothing is to be hidden. Then come back to the Database tab: the step is checked again by itself, or press Check again. Columns you hide there are never read for linking, never shown as examples, and never usable in a question.
Under What would you like to ask across them?, choose a goal: Totals of amounts by linked record, Time by linked record, Counts by linked record, or Let CaseQuery suggest. Then press Find the links. Your firm's connector reads both databases where they are, compares their columns' values in its own memory, and sends back only names and rounded statistics. This can take a minute or two.
The connector considers text and whole-number columns your Schema rules allow, that are not free text and hold enough different values. It does not read a column whose name says it holds people's names or a structured personal identifier, and it drops any column with a value that looks like one. It reads each value several ways: exactly as written; ignoring capitals and spaces at either end; after the last separator; letters and digits only; and without leading zeros. So EAST:AB-0123 in one database can be found to match AB-0123 or AB0123 in the other. Where no single field ties two tables, it tries pairs of fields together. A pairing of two columns of small whole numbers that would match as well by chance is not proposed.
What it finds is listed under The links your connector found. Each names the two tables and the field or fields that tie them, with its evidence in plain words: how many values tie the two and how they are read, what share of each side is found in the other, how the records relate, and the common shapes of each key field's values (A for a letter, 9 for a digit, such as AAAA-999 (62%)). Counts are rounded, and no value from either database is in a proposal. Under Also looked at, and not proposed: the card says why others were set aside. A link your connector already holds is marked Already linked ✓ and not offered again. Find the links again looks again.
If nothing stands out, the card says No link stood out clearly enough to propose. and step 4 stays the one to do. CaseQuery looks at the likeliest tables first. If you know which tables hold the shared reference, press Choose the tables to look at, choose up to eight in each database (hold Ctrl or Shift to choose more than one), and press Find the links in these tables. Only table names are sent for this. Otherwise the two may not share a reference: pick two others.
Once links are found, CaseQuery sets up the likeliest one not already linked, using your firm's AI provider, and first asks, in a pop-up:
It is asked each time, and the answer is never remembered. Where your firm has an AI provider set up, the analysis uses it, and its credit, either way. Your firm's audit journal records that it was asked and the answer, with counts of columns and rows, never a value. Where your firm has no AI provider set up, the link is proposed from names and statistics without one. Beside each proposal, Set it up sets up that one instead, and Show examples shows real values from both databases to you alone (see below).
One screen says what will be linked and how, for example Link files to invoices by File number (read after the last separator)., and what people will be able to ask, beginning You'll be able to ask:. Below it, before anything is confirmed, Left out: lists each field that will not be used and why:
| Reason given | Meaning |
|---|---|
| not needed for this | The analysis set it aside for this goal; where it gave a reason, that is shown instead. |
| its name or values look like personal information | Refused, whatever the analysis said. |
| too many different values to filter or group by | Not a category. |
| its values could not all be checked | The connector could not screen it. |
| its kind of value cannot be used that way | For example, a date can only narrow a question to a whole period. |
| a whole number is added up only if you choose it under Advanced | Whole numbers are never added up unasked. |
| more fields than one link may use | A link may use up to 16 fields on each side. |
A line says how the link was proposed: from names, statistics and a few rows of each table, which were not kept; from names and statistics, no rows sent; or from names and statistics alone, where your firm has no AI provider set up.
Press Confirm. Your connector checks the link on the data before keeping it. If it would leave out anything more, nothing is saved, and the confirm screen comes back with those fields listed first; look again, and confirm.
Advanced: choose each field yourself opens the full review, started from what was proposed:
Press Accept this link, or Cancel. The card then says it was accepted and is held by your connector, and names any field the connector did not keep, and why.
Show examples beside a proposal shows real values from both databases, to you alone: a few values of each key field, pairs that match, and for fields proposed as categories, their codes with how many records have each (with a code's description where your database keeps one in a lookup table). They may include people's names. Never shown: columns your Schema rules withhold, addresses, and identifiers such as national ID, telephone, email or account numbers. Examples go from your connector to your browser only: they are not sent to the AI provider and not kept. They are cleared when the password you confirmed is no longer recent (15 minutes), and by Hide examples, folding the card or finding the links again.
Once your connector holds a link, the steps give way to Set-up complete. Your linked databases are ready to ask. On the Chat screen the linked pair is an entry in the Database list, beside the single databases. Try it in Chat opens the Chat screen with that pair chosen, which starts a new chat. Under each link, a line says what it can answer, and a few example questions are offered, written from the fields you permitted: pressing one puts it in the question box on the Chat screen, not sent.
Everyone at your firm can ask across the pair, once both its databases are available in Chat (Make available in Chat, under Which database). In the cloud there is no All databases box beside the question box: the Database list is where a linked pair is chosen.
When your connector already holds a link, the card opens on a summary rather than the steps: each link with its name and its pair (A + B), the two tables and key fields, what it can answer, a badge, and Change and Remove.
Several links can tie the same two databases. They share one entry in the Database list, and CaseQuery chooses the link each question needs.
The badge says whether your connector agrees:
| Badge | Meaning |
|---|---|
| Held by your connector | The connector holds exactly this link and version. People can ask across it. |
| Not held by your connector | The service keeps it but the connector does not, for example after the connector was replaced. Remove it and set it up again. |
| Your connector holds a different version | The two disagree. Use Change, or remove it and set it up again. |
| Your connector could not be asked | The connector is not connected, or did not answer just now. Look again later. |
An accepted link holds names only: tables, columns, how they are read and what each may be used for, who accepted it and when. Your connector answers a question only for a link and version it holds itself, bound to both databases and their Schema rules as they were when it was accepted. Every statement the connector runs for linking is written to its statement record, and stops when your firm switches the connector off with connector off.
With the pair chosen in the Database list, people ask in the question box as they would of one database. The User Guide describes it for them.
To answer, your firm's AI provider turns the question into a plan using names only: the question, today's date, the links' names, your firm's word for the kind of record, and each side's database, table, key field names and the fields you permitted. Your connector checks the plan against the link it holds, reads each side, joins the two in its own memory and does the arithmetic. The answer goes to the person who asked and is not kept by the service. Your firm's audit journal records each question across a link with counts only.
A total can read many records and return one figure, so repeated narrow questions could otherwise reveal one record's figure. The connector holds a figure back in these cases, and the answer says so:
| What the answer says | Why |
|---|---|
| not shown: fewer than 5 | The figure covers fewer than 5 linked records. |
| not shown: one record carries most of it | One linked record gives more than 80% of the figure. |
| not shown: it would reveal a group that is not shown | Seen only on an answer from an older connector, which held back a second group beside a small one. A current connector holds back only the small group itself and shows every other group. |
| not shown: with an answer given in the last 30 days, it would reveal a figure for fewer than 5 | Together with a figure already shown across your firm's links in the last 30 days, it would let a figure for fewer than 5 records be worked out. The same question asked again is not affected. |
The history behind the last row is kept on your connector only, as keyed fingerprints and counts, for 30 days: never a reference, a value, a figure or a question. It withholds a figure that differs by fewer than 5 records from one shown in that time; it makes a small group harder to work out from several answers, not impossible. Grant the fields a question may use with this in mind: a category that splits your records into very small groups is of little use. Other limits, fixed on the connector: a question of up to 1,000 characters, up to 2 groupings and at most 50 groups; at most 500,000 different references and 1,000,000 rows on each side, or the question is refused whole, never answered in part; and hourly and daily allowances of questions, of finding links, of examples and of set-up analyses.
Where your firm's connector uses the office-network check (1.4), finding the links, showing examples, setting up and accepting a link and asking across one each read values, so each needs a ticket from the desk on your network, as any other read of values does.
| What you see | Likely cause, and what to do |
|---|---|
| No Link your databases card on Settings › Database | Linking is not switched on for your firm, or you are not an administrator. Ask NetDirector. |
| The card says your firm's connector needs updating | The connector does not yet offer linking, or this part of it. Upgrade it and switch linking on in its settings, as above; on a connector NetDirector hosts, ask NetDirector. |
| The card says linking needs two databases | The connector offers only one. Add another (1.4). |
| A step says another step comes first | Do the step it names. Step 4 waits for a vocabulary, two databases and both their Schema rules saved. |
| No link stood out clearly enough to propose. | Read the reasons under Also looked at, and not proposed:. Use Choose the tables to look at to name the tables that hold the shared reference. It may be in a column your Schema rules hide, in free text, or written too differently in the two databases to match; a view that presents the reference cleanly can help (2.8). |
| Every link found between the two is already linked | To look for another, choose the tables to look at. |
| A field you wanted is under Left out: | Read its reason. A whole number can be ticked for adding up under Advanced: choose each field yourself. A field refused as personal, or as having too many values, cannot be used; that is by design. |
| The confirm screen comes back with fields listed first | Your connector checked the link on the data and would leave those out. Look again, and confirm. |
| CaseQuery has set links up several times recently | The allowance for set-up analyses is spent. Try again later, or use Advanced to set the link up yourself. |
| Those proposals are no longer held by your firm's connector | More than 30 minutes have passed, or the connector restarted. Find the links again. |
| Your connector would not accept the link as adjusted | An adjusted link may use only the two tables proposed, and columns your Schema rules allow. |
| The examples disappear, and the card says the password is no longer recent | The 15 minutes since you confirmed who you are are up. Press Show examples again and confirm. |
| A kind of record cannot be removed | An accepted link is of that kind. Remove the link first, or change it to another kind. |
| The pair is not in the Database list on Chat | Both its databases must be available in Chat (Which database), and the link must show Held by your connector. |
| People are told a database, or its Schema rules, changed since the link was accepted | Use Change on the link and accept the change. |
| People are told the question would read more than the connector allows | Narrow it to a period or a category. |
The tabs across the top of Settings, in screen order. Each needs the Administrator role.
The ordered rollout list. Every step is described in 1.3. Return to it after any change to the connector, its databases, people or the AI provider, and press Re-check.
Two tabs: Who may administer, which holds who may sign in and from where, and Import & export tuning (see 2.9).
Everyone who signs in for your firm, with their Role and when and by whom they were added. A user asks questions; a lead also settles definitions and reviews answers; an administrator also manages Settings and these sign-ins.
Nobody changes or removes their own sign-in here; another administrator can. Your firm always keeps at least one administrator.
A firm can let its people sign in with their Microsoft work accounts (Microsoft Entra ID). NetDirector links your firm's Microsoft directory to CaseQuery on request, and can also make Microsoft sign-in the only way in for your firm's people. Once linked:
Optional. List your office's internet addresses, and your VPN's, one address or range per line (for example 203.0.113.10 or 203.0.113.0/24), under Allowed addresses, and press Save. Add my address adds the one you are using now. Nobody can then sign in to your firm from anywhere else, even with the right password, and anyone already signed in is signed out the moment they use it from elsewhere. Leave it empty to allow sign-in from anywhere. Your connector is not affected.
Safe default: empty, unless your firm's people always work from the office or the VPN. When to change: when your office's or VPN's addresses change. If they change and nobody can sign in, contact NetDirector, who will clear the list.
For: how much data each question may return, what reaches the AI provider, floor voting, and what each answer shows.
| Control | Default | What it does, and when to change it |
|---|---|---|
| Safe response mode | On | Favours totals and constrains detailed listings. For databases too large to send in full, switching it off may name more unclassified columns to the AI, so classify columns in Schema rules first. |
| Confirm token estimate | Off | Requires Okay or Cancel before each AI provider call. The send button becomes Estimate & Run and a Review estimated token use dialog appears. Turn on only where every call must be approved. |
| Maximum detail rows | The service's ceiling | The service answers with at most 500 rows, whatever safe mode says; safe mode changes how the AI is asked, not how many rows come back. Lower it to keep listings short. |
| Control | Default | What it does |
|---|---|---|
| Check questions before sending | On | Warns the user when a question appears to contain a Social Security number, a payment card, an email address, a telephone number, or a real value from a column treated as sensitive. Nothing is sent until they decide. It reduces mistakes rather than preventing them. Leave on. |
| Also watch name columns | Off | Also catches a person's or party's name typed into a question. Surnames overlap with ordinary words, so it warns more often. Turn on if names must not travel. |
| Control | Default | What it does |
|---|---|---|
| Ask the floor when a question is ambiguous | Off | Asks people which reading of a status word they mean. Changes how the product speaks to everybody. |
| Votes needed to settle it | 4 | Different people agreeing makes a provisional settled definition. |
| Stop asking after | 10 | Asked this many without agreement, it goes to the leads. |
| Everyone is asked | Off | Off lets each person switch asking off for themselves. |
How a vote becomes a settled definition folds away a short explanation. The full behaviour, including when the leads rule, is in 2.4.
| Control | Default | What it does, and when to change it |
|---|---|---|
| Explanation level | Standard | Under Around each answer: Plain — answer only; Standard — adds how it was measured; Reasoning — adds the assistant's reasoning; Expert — adds the SQL it wrote. Every level shows the answer in words and any warning that an answer may be wrong, except which business rules were left out of a question: that line shows from Reasoning up. A practical ladder is Plain for executives, Standard for most staff and Reasoning for leads, who can spot a rule that should have applied and report it to an administrator, who owns the written rules. The result table follows Show result table. |
| Everyone uses this level | Off | Off lets people choose their own Explanation beside the question box, so someone checking the system can go deeper without changing what anyone else sees. On holds everyone at the level above and hides their choice. |
| Show result table | On | Displays the returned rows beneath each answer. |
| Suggest follow-up questions | Off | Offers two or three related questions under each answer, ready to click. Useful for people who do not know what to ask next. |
| Result columns | Requested columns plus supporting columns (recommended) | Right after Show result table, because it decides what that table contains. See below. |
Result columns. Supporting columns can make an answer easier to understand. Only the requested columns adds no supporting figures or joined names, and does not add sorting or rounding.
Hard protections remain active whatever is set above: writes, multiple statements, unsafe engine functions and system-catalog access are always blocked, by the service and again by your firm's connector; an answer is at most 500 rows; and columns classified as sensitive in Schema rules never leave your firm's network.
For: the account that answers questions, the model, and how hard it thinks.
Your firm uses its own account with one of four providers: OpenAI, Anthropic, AWS Bedrock or Microsoft Foundry. The AI is shown your database's structure and your questions, never a record from it. The key is sealed by the service and never shown again; only its last four characters are, so you can compare it with your provider's console.
Your firm can keep a key for more than one provider. Each is listed with its model and the last characters of its key; Use this one switches questions to a provider already set up, without entering its key again. The one in use is marked In use. One key serves everyone at your firm.
How long the model thinks before answering. Each level shows a measured typical time per question; your own questions will differ. Quicker is not a worse answer for most questions. The setting applies to Claude models; the screen says if the active provider ignores it.
Fast mode, off by default: the same model generating faster, billed at roughly double. Your provider organisation must be granted it; if not, questions run at normal speed and the screen says so.
Check rules against a second model: before a rule is recorded, a second model plans the same question. If both write the same query, the rule says one thing. If they differ, the wording still admits more than one reading. It runs only when a rule is previewed, never on an ordinary question, at one extra call each time, on your own account. The second model is chosen for you from the answering model; where none is paired, the switch says so.
Safe default: the provider's usual model, default speed, fast mode off. When to change: turn on the second-model check while rules are being written in earnest; save a new key when one is rotated or may have leaked.
The example questions on the chat screen before anyone has asked anything. They are presentation only and never reach the AI, so word them in your firm's own vocabulary. Each database keeps its own list, and everyone at your firm sees it. Type One question per line, then Save questions.
A database starts with none. Suggest questions drafts some from the database's table and column names alone, through your AI provider, for you to keep, edit or discard; nothing is saved until you press Save questions. No rows, values or notes are sent, and columns your firm has hidden and tables left out of preparation are not named.
Casey is the guide to this product: it explains and drafts, changes nothing and never runs a query. It uses your AI provider account. Casey's questions are not kept by the service. Full detail is in Working with Casey.
| Control | Default | What it does |
|---|---|---|
| Offer Casey on the chat screen | On | Everyone at your firm gets the Ask Casey switch beside the question box and the Ask Casey panel (Ctrl+K). |
| Let Casey read what your firm has written down | On | Lets Casey read your terms, business rules, settled definitions and column notes, so it can answer "why is my rule not firing" rather than only "how does the Rules screen work". Still no rows from your database, and never notes on hidden columns, though a column note can name the values a column holds. |
| Starter questions for the assistant | Supplied set | Casey's own suggested questions, one per line. Restore the supplied set puts the originals back. |
Press Save. Casey accepts questions up to 2,000 characters.
The tab's cards, in screen order. Every card is the full width of the page and folds the same way: press its heading, beside the small arrow, to open or fold it. Some are shown only where they apply.
The connector, how to install it and how to look after it are described in 1.4. On this card:
Your firm's connector can offer more than one database. The one each person chooses is where their questions are answered, and what the schema, preparation and fine-tuning screens show them. It is each person's own choice: nobody else's changes with it. Choosing is offered only when the connector offers more than one: Use this one beside a database makes it yours, and the one you work in shows In use.
Administrators also decide which databases are offered on the Chat screen: Make available in Chat or Remove from Chat beside each, and Opens by default for the one people start in. People other than administrators see only the databases made available in Chat. A linked pair (2.12) is offered on the Chat screen only once both its databases are available in Chat.
Shown only when your firm's connector takes databases from here, as a connector NetDirector hosts can. Give it a Name it is listed by, the Server, Port, Database, User and Password, and press Test and add. PostgreSQL, over an encrypted connection whose certificate is checked; nothing is added that does not open. The login is sealed in your browser so that only your firm's connector can open it; the service passes it on and keeps no copy. Use a login that can only read. On a connector your firm runs, add databases on its server with connector add-database instead.
Whichever way a database is added, save its Schema rules before anyone asks a question: until then only its structure is read.
Shown to administrators where linking is switched on for your firm. It ties two of your databases together by a shared reference, so a question can be answered across both, and lists the links your firm has. Described in full in 2.12. Where it is shown, the two cards below it each begin with a line pointing to it: they are for counting across databases that hold different records, not for tying two databases together.
Shown when the connector offers more than one database. In the cloud edition, Connect answers a question across databases with a count, and nothing else: up to four databases, each database's own count shown.
Counts are added together only across pairs you have declared cannot hold the same matter at once, declared from the current main database. For each database added, you are asked whether the same matter can ever be active in both it and the main database at once, and who at the firm answered. This is asked, never detected. No — combine them lets the counts be added; Yes — keep them separate keeps each count on its own. Make available in Chat then offers the databases together as one entry in the database list beside the question box, for everybody.
Connect does not list records, add up amounts, or match records across systems. To match related records in two databases and count or total across them, link the databases instead (see 2.12).
Appears when combining databases. For each database, say which column holds the firm's case number, the status and the rest. Nothing is assumed: a field bound to the wrong column gives answers that look ordinary and are about something else. A database not yet described cannot be asked across. The database's own codes in these mappings stay on your firm's connector.
In the cloud there is no All databases box beside the question box. The Database list on the Chat screen is the one place to choose what to ask: a single database, a combined entry, or a linked pair.
Safe default: one connector, one database, its Schema rules saved. When to change: a new database, a replacement server, or a database password change (made on the connector's server).
Three tabs, all described in Part 2:
For: browsing the schema, hiding columns, writing column notes, and choosing what opens when someone clicks a case number. Until a database's Schema rules are saved, none of its data is read: only its structure.
Search tables, views or columns, filter to Views or Tables, and refresh with the ↻ button after the database changes. Select an object to see Columns, Sample data, Definition and Relationships. Sample data is read through your firm's connector to your screen, and is not kept.
Tick a column only if nobody should see its values: a Social Security number, a bank account or card number, a password. A ticked column is withheld by your firm's connector: never sent to the AI, never displayed to anyone, and any query naming it is refused.
Do not tick a column just because it looks personal. Party names, addresses and reference numbers are how staff and the AI identify records; hiding them stops ordinary questions working.
Each database the connector offers has its own Schema rules. Safe default: hide nothing unless it is truly never needed, but save the policy either way. When to change: before preparing, and whenever a sensitive column is added.
No database rows or query results are sent to the AI whatever you tick. A value written into a rule or column note, including code lists accepted from preparation, travels as part of that instruction.
Explain abbreviated or coded columns so the AI reads them correctly. Each visible column has a note box; hidden columns do not, because a note on them could never be read. The values a column holds are shown with counts. A full list short enough to fit can go into the note; when there are too many values, only the most common are shown and not offered, because a partial list would tell the AI the rest do not exist.
Show only columns needing a note narrows the list. Press Save notes.
What opens when someone clicks a case number in a result. Tick fields in the order you want them read; ticking adds a field to the end, so re-ticking moves it. Use what the primer suggests fills it from preparation. Press Save. Hidden columns never appear, whatever is ticked.
Designing a reviewable CREATE VIEW script from recorded questions or from sources you choose. Described in 2.8.
Records what each of your questions sent to the AI provider, so you can show an auditor what left your firm rather than describe it. The log is kept in your browser only, for you; the service keeps no copy.
Gathers what support needs to diagnose a fault into one file you can read before sending. Nothing is transmitted from here, and the service keeps no copy. It contains the shape of your set-up: your firm's name, your databases' names and your connector, how many tables, rules and notes there are, which AI provider is in use and where its requests go, and the last entries of the activity record with people's names replaced by user-1, user-2. It never contains a key, a password, a database address or login, a question or a row of your data.
Safe default: tracing off. When to change: an audit request, or a fault to report.
| Screen | User | Lead | Administrator |
|---|---|---|---|
| Chat | Ask questions, vote on the floor, send corrections, use Casey, personal choices | The same | The same |
| Fine-tuning › Definitions to settle | Read | Settle or set aside definitions; rule on Floor needs a ruling; Send to the DBA | Everything a lead can do |
| Fine-tuning › Baseline questions | Read | Run sections, mark answers, add questions | Everything a lead can do |
| Fine-tuning › Awaiting sign-off | Read | Read | Confirm this / Withdraw it |
| Writing a definition | — | Check the wording, See what this changes | The same |
| Business Rules | Read | Read | Save rules and terms, Restore from History, accept a drafted rule |
| A linked pair in the Database list on Chat | Ask across it, once both its databases are available in Chat | The same | The same, and set up links on Settings › Database › Link your databases |
| Settings (every tab) | — | — | Everything, including people, the connector, the AI provider, preparing the database, Question-SQL pairs, ground truth, the DB View Assistant, and taking and restoring a copy |
An Administrator reaches every screen. Every recorded decision carries the sign-in name of the person who made it. Changes to who can reach your firm's data, or where it goes, ask administrators to confirm who they are if they last did so more than 15 minutes ago.
Every call below is billed to your firm's own provider account.
| Activity | Cost |
|---|---|
| An ordinary question in chat | Provider calls per question. A settled Verified query costs nothing. |
| A question across a linked pair | One provider call per question, to plan it, from the question and the links' names. |
| Setting up a link: the analysis after Find the links, or Set it up | Where an AI provider is set up, one provider call each time, on your firm's provider's most capable model where it offers one. Finding links, showing examples, accepting and removing a link make no provider calls. |
| Preparing the database, and Check for new definitions | No provider calls: computed with SQL by your firm's connector. |
| Reading system documentation during preparation | Uses the provider. Only the documentation is sent; table data is not. |
| A baseline section run (Answer this section) | Each question is asked twice: up to 2 calls per question, at most 3 if a repair is needed. Questions already verified cost nothing. |
| Ground truth Check the answers | Each question is asked once, 3 or 5 times: the cost multiplies accordingly. The screen states the number before you confirm. |
| Casey, including Help me sort this out… | Uses the provider for each question. |
| Rule and definition checking: Check the wording, See what this changes | Uses the provider. Check rules against a second model adds one extra call each time. |
| Request SQL for this question on Question-SQL pairs | Uses the provider. Run it and show me runs your query against the database only. |
| View generation (Estimate & Generate View Script) | Uses the provider. Recommendations and the read-only preview do not. |
| Suggest questions for starter questions | Uses the provider. Only table and column names are sent. |
| Fast mode | Roughly double the price per call, where granted. |
| Floor voting, saved starter questions, Schema rules, copies of the set-up | No provider calls. |
| What you see | Likely cause, and what to do |
|---|---|
| You are asked to sign in again | The session ended: 30 minutes without use, or eight hours in all. Sign in again. |
| Your password is refused although it is right | Too many failed attempts; wait a while and try again. If you have forgotten it, another administrator can use Reset password. |
| Nobody can sign in since the office's internet address changed | The list under Where people may sign in from no longer includes it. Contact NetDirector, who will clear the list. |
| Someone cannot see Settings | Settings is offered to administrators only. Change their role on Settings › Security if they need it. |
| A work-account link does not work | It works once, for seven days, and only with an account your firm's Microsoft directory admits. Send a new Work-account link. |
| A checklist step will not tick | Steps marked Checked cannot be ticked by hand. Do the work, then press Re-check. Nothing can be checked until the connector is paired. |
| The connector shows as paired but not connected | On a connector your firm runs: check the server is on, run connector status and connector doctor --as-service, and check whether someone ran connector off. Check the server can reach the internet over HTTPS and its clock is right. On a connector NetDirector hosts, contact NetDirector. |
| The pairing code is refused | It works once, for 30 minutes. Create a new one. If your firm already has a connector, tick If my firm already has a connector, replace it in the setup form. |
| Questions are refused with no AI provider set | Save your firm's key on AI provider, and press Test. |
| A database shows its tables but no data is read | Its Schema rules have not been saved. Open Schema rules, choose it, and press Save & lock policy, even with nothing ticked. |
| The database looks smaller than it is | The connector's login lacks rights to some schemas or tables. Grant it read access to them on the database. |
| The preparation badge reads Schema changed — run again | Tables or columns changed. Prepare again. |
| A settled question is no longer answered as a Verified query | The database changed since it was written, so it is not used. Check the query and settle it again. |
| Floor voting never asks anyone | It is off, the question did not meet both conditions in 2.4, or the person switched asking off. |
| Nothing is waiting under Floor needs a ruling | Normal while votes are still arriving and everyone agrees so far. |
| A rule was refused on save | Too long (over 400 characters) or over the total. The refused text is kept on screen; shorten it or split it. |
| Rules disappeared | Someone saved an empty or different set, or restored a copy. Use History › Restore, or restore your last copy. |
| A new view is not available to questions | The schema has not been re-read. Press I have created it — re-read the schema or refresh on Schema rules. |
| Answers to the same question still vary | Look at Has it worked?, settle the definition behind it, or check with ground truth asked 3 times. Settle it as a Question-SQL pair if it must never vary. |
| Linking is not offered, or a linked pair cannot be asked | See the table at the end of 2.12. |
| Something else | Create a support file and send it to NetDirector (see 3.10). |