Documentation
Open platform

Get started

Platform tour

Learn how sign-in, workspaces, evaluation areas, execution tools, and Personal Access Tokens fit together in the current AIDX platform.

The active workspace is the operating boundary of AIDX. It determines the evaluations, credits, members, and enabled modules you can use. Within that boundary, BenchDX and RobustDX provide their own creation, execution, progress, and result views.

Sign in and create your first workspace

AIDX uses passwordless email verification and may also offer Google or Microsoft sign-in when those providers are configured.

AIDX Account Login page with an email field, Terms of Service checkbox, Login button, and Google and Microsoft sign-in options.
Accept the Terms of Service, then continue with email verification or an available OAuth provider.

  1. Accept the Terms of Service

    Accept the terms before using email, Google, or Microsoft sign-in.

  2. Verify your identity

    For email sign-in, enter your address, select Login, then enter the six-digit code sent to your inbox. Use Resend code after the countdown or Use a different email when needed.

  3. Complete first-time setup

    New users enter an Account name and Workspace name, then select Create workspace. The new workspace becomes active and AIDX opens the Home page.

If sign-up began from an invitation, complete the first-workspace step first. AIDX then returns you to the pending invitation so you can accept the invited workspace.

Understand workspace context

Existing users select a workspace after sign-in. Use the workspace selector in the application header to change the active workspace. Always confirm the active workspace before creating an evaluation or spending credits.

Select a workspace page showing two workspace cards, a selected workspace indicator, role labels, and a Continue button.
Workspace selection determines the resources, access, modules, and credits used in the session.
Workspace controlsWhat they affect
MembershipWhich users can enter the workspace and what member-level access they receive.
Enabled modulesWhich evaluation menus and capabilities are available.
CreditsThe balance used when an evaluation is created.
Evaluation dataThe evaluations, progress, and results visible in the current context.

Find the supported evaluation areas

MenuUse it forMethod-specific input
LLM → BenchDXBenchmark safety evaluation against selected categories or standardsModel name, model version, and benchmark categories or standards
LLM → Adversarial TestRobustDX adversarial testing using generated attack promptsUploaded CSV or TXT seed prompts and selected attack methods

Know the evaluation tools

Open an evaluation to access the tools used after creation:

Request Tutorial
The post-creation guide that points you to the available request workflow for that evaluation.
API Guide
The integration path for retrieving prompts and submitting one model response for each traceId.
Evaluation Trigger
The path for mapping AIDX prompts into a target API, extracting its responses, and sending pending requests without a custom integration script.
Process Details
The case-level view of Awaiting Trigger, Sent, Responded, Evaluated, and Error states.
Completed dashboard
The result view containing the evaluation summary, conclusions, charts, detailed scores, and score definitions.

The API Guide and Evaluation Trigger are alternative execution paths. Use one deliberate path for a request batch and monitor it through Process Details.

BenchDX API Guide showing the GET promptsAll endpoint, bearer PAT header, evaluation ID, curl example, and the start of the response submission step.
The first API Guide step retrieves all prompts for the selected evaluation.
BenchDX API Guide showing the POST response endpoint and the traceId, llmResponse, and evaluationId form fields with a curl example.
The second API Guide step submits one target-model response for every returned traceId.

Create and manage a Personal Access Token

A Personal Access Token (PAT) is the supported credential for scripts and external automation using the LLM evaluation APIs. Evaluation Trigger does not require you to create a PAT.

Upper-right AIDX user menu with Personal Access Token and Logout options.
Open the user menu in the upper-right corner to reach Personal Access Token management.

  1. Open Personal Access Token

    Open the user menu in the upper-right corner and select Personal Access Token.

  2. Create the token

    Select Create PAT, enter a name, and choose an expiration date or a 30-, 60-, or 90-day shortcut, then confirm creation.

  3. Copy it immediately

    Store the raw token securely when the creation result appears. The token value is shown only once; the list later shows only its name, status, and expiration.

  4. Use it with the evaluation guides

    Send Authorization: Bearer <YOUR_PAT> in each supported external request. Open the evaluation’s Request Tutorial for its evaluation ID. The documented API base URL is https://trial.aidx.pro. Use API Guide to retrieve prompts and submit one response for every traceId.

  5. Revoke it when no longer needed

    Delete the token from the Personal Access Token page to revoke it before expiration. Expired tokens cannot be used.

Common workflows

Tutorial

Run your first evaluation

Create a BenchDX evaluation and complete it through Evaluation Trigger.

Evaluation guide

Understand BenchDX

Learn the benchmark method, scope, scoring model, and limitations.

Evaluation guide

Explore RobustDX

Create adversarial tests from seed prompts and selected attack methods.

Reference

Troubleshoot a run

Diagnose access, trigger, response, processing, or result issues.