First-Run Checklist

Access and status

Sign in to the intended deployment. Check the application and model-service status using the controls available to your role.

Storage and model

Review the configured storage and privacy boundaries. Choose an available text-capable model appropriate for the system.

A small conversation

Request a short answer, wait for completion, then refresh and verify the conversation is still present. This checks one workflow, not the entire deployment.

If something fails

Record the error and time. Check the relevant status before restarting anything.

A guided first run

Use a non-sensitive practice task. Confirm your authorized account and intended workspace. If a control is unavailable, ask your operator rather than changing permissions yourself. Keep the first task small so you can distinguish an access problem from a model or attachment problem.

Ask Friday to guide you

In an authorized FridayLocalAI conversation, try:

Walk me through my first FridayLocalAI session one step at a time. Ask me to confirm the intended workspace, visible Agent and Model controls, a short completed response, and saved conversation history. Wait for confirmation after each check.

Tell Friday what you can see and which step you have reached. If the instructions do not match your screen, say so; ask for the relevant Help topic or contact your operator. Friday cannot be assumed to see your screen or know your current settings.

Steps to follow

1. Sign in using the access address supplied by your operator; check the displayed organization and account.
2. Open Organizer and select the intended conversation, or create a practice conversation if permitted.
3. Check the Agent and Model controls. Send: “Give me three ways to organize practice meeting notes.”
4. Wait for the completed response and check its active model attribution.
5. Return to the same conversation after one normal refresh and confirm that your message and the reply remain.

Establish what this first run is for

The goal of a first session is to verify a small, understandable workflow, not to exercise every feature. You should be able to access the intended environment, find a conversation, request a short response, review it, and return to saved history. Once those basics are clear, you can move on to attachments, memory, or more specialized capabilities. Use the account and access address supplied by your operator. The public FridayLocalAI website explains the product but is not the same as the application workspace. A store account, a book purchase, or downloading a consumer guide does not grant application access. If you have not yet been given access, use the early-access contact route rather than trying to infer an installation process from these instructions.

Check identity and orientation first

Confirm that the displayed organization and account are what you expect. Then locate the workspace regions: Organizer, the active conversation, and Dashboard where available. The layout may differ with screen size or saved preferences, so a hidden panel does not necessarily mean a feature is missing. Use the labeled layout controls to orient yourself. Ask Friday for help only after conversation is available. Before that point, use the website’s guides or your operator’s instructions. If you appear to be in the wrong environment or can see unrelated information, stop and report the issue. Do not explore another person’s material to determine how much is visible. Record the minimal observation needed for the operator to investigate.

Choose a harmless practice task

Create or select an appropriate conversation using the controls permitted to your role. Type a small task whose result you can evaluate without outside research. For example, supply three fictional meeting notes and ask for a concise summary. Avoid private documents, external sending, system changes, and complicated generation jobs during this initial exercise. State the expected format and a simple constraint, such as “Use only the information I provide; mark missing dates as unknown.” This gives you something concrete to check. If the answer invents a date or owner, ask for a correction and compare the revised result. The purpose is to learn the exchange and review process, not to mistake a pleasant response for complete deployment acceptance.

Observe the response path

Check the Agent and Model controls before sending the request, then inspect the active attribution associated with the response. The requested Model selector affects future replies, while routing may influence the active model actually used. Note that distinction if the result does not match your expectations or if you are comparing behavior across requests. Wait for completion before deciding what happened. An incomplete response, a visible error, and a completed but inaccurate answer are different outcomes. Record which one you observed. If there is an error, preserve its non-sensitive wording and the time. Do not immediately change multiple settings or restart services, because doing so can make it harder to identify the cause of the initial failure.

Verify continuity carefully

After the practice exchange completes, return to the same conversation after one ordinary refresh or navigation away and back. Confirm that the user message and assistant response are still present in the intended place. This is a practical check of that saved conversation workflow. It is not proof that every storage, recovery, or retention behavior has been validated. If the conversation appears missing, avoid creating a series of duplicate replacements. Check the selected project, conversation name, and available search first. Note what you expected to see and what is actually visible. Ask the operator for help if the discrepancy remains. Preserving a clear description of the event is more useful than repeatedly changing the workspace in an attempt to make it look right.

Decide what to learn next

If the basic conversation works, choose one next capability that matters to your intended use. Follow the attachment guide for source work, the model guide for understanding selection and attribution, or the Memory Notes guide for reusable preferences. Keep the example small and define the expected result before proceeding. If the basic workflow fails, use the troubleshooting guide and contact the operator with your observations. Include the feature, time, Agent and model information you can see, and the smallest reproduction. Leave out passwords, tokens, and private source contents. A successful first run ends with an understandable result and a clear next step, not a claim that every part of FridayLocalAI has been certified or enabled for your account.