DataMind Oda answers some questions in a fixed way: what it is, which language it uses, how it treats personal data and where your data goes. This page collects those answers, together with the suggestions it shows while you edit files.
When you ask what it is, Oda answers with one sentence:
"I'm DataMind Oda, DataMind's AI work agent; I run on the model chosen for this session, through your company's DataMind server."
That sentence is the whole answer: Oda does not claim, confirm or deny being a particular AI model or vendor. The model of the session is the one you choose in the chat; see Chat with Oda.
The DataMind server in that sentence is your company's DataMind OS, and everything Oda does goes through it. Install and sign in covers downloading Oda and the first sign-in.
Oda replies in the language of your latest message, or in the language you ask for. A message that is only pasted code, logs, paths or data does not change the language: the conversation keeps the one it is using. Oda writes in that language's own script, never in another alphabet, and never mixes languages inside one sentence. Everything you see follows it, including the files Oda makes for you, unless you ask otherwise.
Names and code stay exactly as they are, whatever the language of the answer:
Oda reaches company data through the Minds of your DataMind OS, within your own access; see Company data, documents and reports. When a result depends on that access, Oda says it is for the data your account can access, unless the data is company-wide.
When a task involves personal data, Oda handles masked values rather than the original ones. A masked value looks like {{EMAIL_a1b2c3d4e5}}. Oda follows strict rules with them:
Ask where your data goes, and Oda describes the path your request takes: through your company's DataMind OS, which keeps an AI activity log your administrators can review (Manage DataMind Oda for your company); web searches go to services outside DataMind. It does not claim that nothing is stored.
For the app's own network calls and the files it keeps on your computer, see Account, data and updates.
A figure Oda reports comes with its source: which data it came from, the period it covers and the filters behind it, so you can check it. Anything Oda cannot verify is marked as unverified; it says what it verified and what it inferred. When a step fails, Oda reports the failure plainly: what failed, what it tried and what is still possible. When the mistake is Oda's own, it says so and puts it right.
While you type in a code editor, Oda predicts the text you are about to write at the cursor. The suggestion appears as grey ghost text and is inserted unchanged when you press Tab; if you do not press Tab, nothing is added. When Oda sees nothing sensible to add, nothing appears, and it never suggests a password, a key or a token.
After a change, Oda can suggest the one edit it expects you to make next, fixing what the editor reports as wrong or carrying a change through, such as the rest of a rename. That suggestion is accepted with Tab or ignored.
One sentence: "I'm DataMind Oda, DataMind's AI work agent; I run on the model chosen for this session, through your company's DataMind server." Oda does not claim, confirm or deny being a particular AI model or vendor.
The language of your latest message, or the one you ask it to use. It writes in that language's own script and does not mix languages inside a sentence.
The language does not change: a message that is only pasted code, logs, paths or data keeps the conversation's language, so the answer comes in the language you have been using.
A stand-in for personal data, written as a label and a short code, like {{EMAIL_a1b2c3d4e5}}. Oda copies it byte for byte, uses it only to filter or group values in the column it belongs to, never guesses one and never joins on one.
It describes the path your request takes: through your company's DataMind OS, which keeps an AI activity log your administrators can review, and web searches go to services outside DataMind. It does not claim that nothing is stored. See Manage DataMind Oda for your company.
Because it could not confirm them. Oda tells you where its figures come from, says what it verified and what it inferred, and marks anything it could not verify so you know what still needs checking.
It reports the failure plainly: what failed, what it tried and what is still possible. When the mistake is Oda's own, it says so and puts it right.
While you edit code, Oda predicts the text you are about to write and shows it as grey ghost text. Press Tab to insert it, or carry on without pressing Tab and nothing is added.
A next-edit suggestion: the one edit Oda expects you to want next, such as fixing what the editor reports as wrong or finishing a rename. Accept it with Tab, or ignore it.