FAQ and troubleshooting
Answers to common questions and fixes for issues you may run into across the Console, deployment, and Studio. If you don't find your answer here, contact your account team.
General
Where does my data live?
Everything you create in Studio — agents, conversations, documents, datasets, connector credentials, and knowledge graphs — stays inside your own AWS account (and your own knowledge-graph database). The only information that flows out is license and usage telemetry sent to the Console. See Data residency and security.
Does Prometheus Research Labs ever see my data?
No. The Console is a management portal and never connects to your AWS account. It receives only signed license activation, heartbeat, and usage information. See Licensing and telemetry.
What is the difference between the Console and Studio?
The Console is the hosted portal for account, billing, licensing, and downloads. Studio is the product you deploy into your AWS account and work in. See What is AlphaAgent?.
Which AWS regions and residency zones are supported?
Model inference runs on Amazon Bedrock pinned to a single residency zone — United States or European Union — that you choose at deploy time. Inference never crosses zones. See Data residency and security.
Account and licensing
I lost my license token. What do I do?
The license token (the aalk_<id>.<secret> value) is shown only once when you create the key. If you lost it before deploying, create a new license key and use its token. The public license id is always visible in the Console. See License keys.
My Studio shows a license warning banner. What does it mean?
Studio periodically checks in with the Console via a heartbeat. If those check-ins cannot complete — for example because of a billing issue or a network problem reaching the Console — Studio shows a banner. Check your billing status (any unpaid invoice) in the Console and confirm Studio can reach the Console. See Licensing and telemetry.
My agents are disabled and a banner says to contact my administrator. What happened?
Studio disables agent execution when it cannot confirm a valid heartbeat with the Console. The usual causes are (1) the deployment has lost outbound connectivity to the Console, or (2) there is an account or billing hold on your deployment. Studio resumes on its own once connectivity is restored or the hold is cleared — no redeploy is needed. If it persists, ask your administrator to confirm outbound HTTPS from the deployment to the Console and to check the account billing status (any unpaid invoice) in the Console. See Licensing and telemetry and Deployment troubleshooting.
Will a Console outage stop my work?
Agent execution depends on Studio being able to confirm its check-in with the Console, so if the deployment loses contact, agent runs pause and a banner appears. This clears automatically within about a minute of contact being restored. Your usage data is not lost during a brief outage — it is queued in your own account and delivered once the Console is reachable again.
Deployment
Where do I get the install bundle?
Download it from the Console's download page after creating a license key. See Downloading Studio.
The pre-deployment checks failed. What should I look at?
The installer runs preflight checks before making changes. Common causes are missing AWS permissions, an unverified domain or certificate, model access not yet granted in Bedrock, or your identity provider details not being ready. Review the Prerequisites and the Deployment troubleshooting guide.
How do I update Studio to a new version?
Download the new bundle from the Console and run studioctl update (it takes no options). It auto-detects whether the release is a code-only change or also changes infrastructure and does the right thing. Infrastructure changes are shown as a reviewed change set and any destructive change requires confirmation. If an update can't finish (for example, your AWS session token expires), it rolls back automatically or stops with the exact recovery command (studioctl rollback). See Updating Studio.
How do I remove Studio?
Run python3 studioctl.py teardown. This permanently destroys the entire deployment, including all of your data, in the region you deployed to — it is irreversible, so back up anything you need first. It is heavily gated (a typed account ID + region + DESTROY confirmation and a 30-second abort countdown), and your TLS certificate (ACM) is preserved so you can redeploy later. To move to a new release without losing data, use studioctl update instead, not teardown. See Teardown.
Sign-in and access
How do users sign in to Studio?
Studio uses single sign-on through your own identity provider (SAML). There is no separate username and password to manage in Studio. See Single sign-on (SAML).
Why can't I see another user's agents or conversations?
Resources are owned by the user who created them. Each user sees their own agents, conversations, datasets, and other resources. This isolation is by design. See Data residency and security.
Building agents
I edited my agent but chat still behaves the old way. Why?
Editing an agent creates a new version but does not change what is live. Activate the new version so it becomes the active configuration. See Agents.
Why can't I deactivate an agent?
If the agent is used as a supervisor or specialist in an active swarm, AlphaAgent prevents deactivation so the swarm doesn't break. Remove it from the swarm or deactivate the swarm first. See Swarms.
Why won't my swarm activate?
A swarm can only be activated when all of its member agents have an active configuration. Activate the members first. See Swarms.
My agent complains about infrastructure problems or says it can't run code. What's wrong?
The most common cause is an execution environment that hasn't finished provisioning yet. Creating an environment stands up real infrastructure in your AWS account, and until it reaches the active state an attached agent can't reach it — so the agent reports infrastructure or code-execution errors instead of returning a result. Wait for the environment to become active before activating or running agents that depend on it. If the errors persist, confirm the environment actually became active and that a custom image (if used) can be pulled. See When an environment is still provisioning.
I updated a knowledge graph and some agents stopped using it. What happened?
A knowledge graph keeps a single live version. When you re-ingest, AlphaAgent guides you through an impact and reactivation flow for the agents and swarms that depend on it. See Knowledge Graphs (AMPG).
Chat and data
When should I use Express versus Deep Traversal?
Use Express for quick, direct questions and short tasks. Use Deep Traversal for complex questions that benefit from planning and knowledge-graph retrieval, and when you want to review a plan before the agent proceeds. Note that Deep Traversal requires a Knowledge Graph: every Deep Traversal flow starts from AMPG grounding, so an agent without a Knowledge Graph wired in can't run it — use Express instead, or add a Knowledge Graph to the agent. See Chat and Knowledge Graphs (AMPG).
My results table isn't showing every row in the chat. Is that a bug?
No. Large result tables are registered as datasets and shown as compact previews, charts, or grids so the conversation stays readable. The full data remains in your workspace. See Datasets and data flow.
How long do datasets last?
Session datasets persist for the life of the conversation/run — there is no fixed expiry, so even long, multi-hour runs keep the datasets they register available throughout. See Limits and quotas.
Do runs time out?
Not in the way you might expect. Express and Deep Traversal chat have no fixed turn timeout — the agent streams until it finishes the task (and you can always stop it yourself). Deep Traversal and multi-node Workflows are built for long, multi-step work and can run for extended periods, up to several hours. The only hard limit is on a single code execution, which is capped by your execution environment's timeout (at most 15 minutes per run); the overall chat or workflow keeps going across many such steps. See Timeouts.
Can I attach a document in any chat mode?
Yes. You can attach documents in Express, Deep Traversal, and workflow document nodes alike. An attached document is saved to the conversation store and analysed by the coder; images are read with model vision. See Chat.
If I close the tab during a run, do I lose it?
No. Chat runs are detached — they keep executing server-side even if you close the tab or browser. When you return to the conversation, Studio rehydrates the in-flight run and replays it into the UI (steps, reasoning, streaming output, datasets, and the coder's progress) so you pick up where you left off. See Chat.
My document upload was rejected. Why?
Check the file type and size against the Limits and quotas. Knowledge-graph documents are limited to 100 pages and 50 MiB; chat attachments are limited to 4.5 MB per document (3.75 MB per image) and to specific file types.
Workflows
Why can't I add more nodes to my workflow?
A workflow can contain up to 20 nodes. See Limits and quotas. Consider splitting a large process into multiple workflows.
How do I schedule a workflow?
Describe the schedule in plain language (for example, "every weekday at 8am") and AlphaAgent translates it into a recurring schedule. See Workflows.
Can a later workflow step use data produced by an earlier step?
Yes. A step can load a dataset produced by an upstream step, even though each step runs in its own workspace. See Datasets and data flow.