Your first successful run

Role: Model user. Prerequisites: CLI installed and logged in, with access to a ready model. Files: A new private analysis directory; optional original local file inputs. Commands use the default profile; add --profile research for another configured service.

1. Discover the contract

enumeray model list
enumeray model get example

Use your actual model ID. Read its input schema, named examples, current publication, availability, execution location and scientific-data access disclosure. If no current publication is accessible, follow discovery and access.

2. Initialize and review

Select a published example by its exact name:

enumeray analysis init example --directory first-run --example starter-example

Alternatively, supply your model-specific JSON with --inputs inputs.json. Initialization creates a private directory and saves the exact publication, contract, destination and disclosure. It performs no uploads or execution. Edit first-run/inputs.json as needed, then review the printed destination and disclosure before authorizing submission.

For declared file slots, add a local binding such as --file dataset=dataset.csv. The CLI uploads that source for remote staging and constructs the artifact reference automatically. See analysis directories for file types and metadata.

3. Submit and observe

enumeray analysis submit first-run
enumeray analysis status first-run
enumeray analysis wait first-run --check-outcome --interval 2 --deadline 3600

The first submission freezes the scientific inputs and persists operation state before uploads or run acceptance. Repeat analysis submit first-run after a lost reply or interruption to recover the same operation. A new intended execution requires a new directory.

queued means accepted work is waiting for assignment. running includes preparation, backend waiting, execution and result delivery. Terminal states are succeeded, failed and cancelled. With --check-outcome, wait exits success 0, failure 3, cancellation 4 or local deadline 5. A deadline stops observation while execution continues; resume waiting with the same directory. Default wait behavior exits 0 for every terminal outcome.

For live logs or cancellation, use enumeray analysis logs first-run --follow or enumeray analysis cancel first-run; both retain the same accepted run through the directory.

4. Collect the result

enumeray analysis collect first-run

Collection saves final inspection, effective inputs, scientific outputs, provenance and verified file outputs. first-run/bundle.json maps declared output names to their files and evidence. Repeat collection after interruption; matching files are verified and reused, and conflicting files are preserved.

For the scalar-double fixture with input 2, results/outputs.json contains {"doubled":4}. Interpret other models using their scientific documentation. Collection completeness is separate from scientific success: check complete: true and bundle.json for every verified output, and read the retained scientific state independently. Keep original local data separately: the analysis retains source paths/checksums and artifact references without copying input bytes.

Expected outcome: One accepted run and a verified evidence bundle. Next: Analysis-directory recovery, file inputs and outputs, or logs and provenance. Use the noninteractive JSON recipe for scripts; run history and inspection remain available by ID.