Free assessment did not finish

Open the original assessment

Open the conversation where you used the maintained Compute assessment prompt. Check whether your agent returned a fit recommendation, asked for missing details, or said the workload could not be assessed safely. Work from that observed response instead of repeating the whole review.

The assessment is a decision aid. It does not issue a Lithi credential, approve access, reserve capacity, accept a quote, or start paid work.

It is read-only: it does not ask for private bills or files, install tools, configure credentials, create a paid quote, or run work.

Copying the prompt leads to one canonical review prompt. It does not create an account, grant access, upload anything, or authorize a job, run, batch, or other execution.

Give it one bounded question

If the first response is missing or unclear, return to Usage fit and Set up Compute, then copy the maintained assessment prompt from Get a free estimate. Describe one outcome, the material you are permitted to inspect, the approximate volume, and how you would check a result.

For example, suppose you want to group 10,000 support documents using a fixed rubric. Give your agent the authorized description, the document category, the approximate volume, the desired groups, and the review method. The assessment should identify the useful input, expected result, checking step, and whether current Compute support is established. A clear no-fit answer is useful when the operation, data handling, or availability cannot be confirmed.

If the fit looks wrong, narrow the workflow to one checkable task. State its input type, desired result, failure cases, and review method. Do not change the data category merely to obtain a match.

Do not add customer files, credentials, private records, or bills just to make the description longer. Your own coding-agent provider has its own access, retention, and usage terms.

If the agent asks a focused question, answer only what the assessment needs. If it reports that a capability, price, or data rule is unverified, keep that fact visible rather than guessing.

Keep price and access separate

If current public pricing identifies its response as information_only, your agent may use the published planning rates of $0.05 per million input tokens and $0.10 per million output tokens. For illustration, 10 million input tokens plus 2 million output tokens is $0.70 before applicable adjustments. Lithi measures text, code, documents and media in tokens using the published metering profile. Another provider’s token count cannot replace Lithi’s count.

That illustration is information only. It is not a signed quote, capacity reservation, access approval, charge, or execution authority. Compare only like-for-like scope and billing units, and treat unavailable pricing as unavailable rather than zero.

If the fit looks right, application preparation is a separate choice. When you explicitly ask to apply or request a link, the current flow may prepare it from known authorized facts. Review the result yourself: preparation does not transmit an application, grant access, create a key, or authorize paid work. Transmission requires your explicit approval.

If a prepared link opens a search page instead of the Lithi form, copy the whole link and paste it into your browser's address bar. If the form says the link could not be read, ask your agent to rebuild it from the template in the assessment prompt, or fill in the form yourself.

Know when the assessment is complete

A useful assessment gives you a clear fit or no-fit decision, one small test you could evaluate, and the facts that still need confirmation. Stop when the data handling, operation, or price cannot be established from current sources.

If you choose to continue, Apply for Compute access. Applying requests review; it does not issue a key or approve paid work.

For an assessment problem you cannot resolve, use Contact Lithi with the workflow category, the safe public operation name when known, and what your agent returned. Do not include private source material, bills, credentials, or the agent's full conversation.