# Submission flow

> From an archive on your machine to a score on the leaderboard.

Source: <https://frontieror-challenge.com/docs/main/submission-flow>

You submit solver code for the open round's problems; we run it on their private instances and score the results. What the archive must contain is on [Submission format](https://frontieror-challenge.com/docs/main/submission-format).

## Three ways to submit

|           | How                                                                                                                                                                                                 |
| --------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Dashboard | **Submissions** page: choose the archive and the environment, check which problems it covers, then **Submit**.                                                                                      |
| Starter   | `uv run frontieror submit --wait` packs the solvers in `solvers/` and submits them. See [Example code](https://frontieror-challenge.com/docs/main/example-code).                                    |
| API       | [`POST /main/submissions`](https://frontieror-challenge.com/docs/main/api#submit-an-archive), then poll [`GET /submissions/{id}`](https://frontieror-challenge.com/docs/main/api#get-a-submission). |

All three run the same checks and count toward the same daily limit.

## Before you submit

- Your team is on the Main track. See [Choosing a track](https://frontieror-challenge.com/docs#choosing-a-track).
- A round is open. See [Rounds](https://frontieror-challenge.com/docs#rounds).
- Your team has a submission left today: one per UTC day, 4 during the testing stage, reset at 00:00 UTC.

## What is checked at upload

The upload is checked straight away. A refused upload stores nothing and does not count toward the day's limit. The checks, in order:

1. The file is at most 4 MB.
2. The image is one of ours and the extra packages are valid.
3. A round is open.
4. The file is a zip, tar or tar.gz archive with at most 10,000 entries, no links and no unsafe paths.
5. At least one problem of the open round has a solver. If some have none, you confirm that they score 0.
6. Your team has a submission left today.

The upload form reads the archive in your browser first and shows, for each problem, whether its solver was found.

## Statuses

| Status    | Meaning                                                                                                            |
| --------- | ------------------------------------------------------------------------------------------------------------------ |
| Queued    | Stored and waiting for a sandbox.                                                                                  |
| Running   | Being evaluated: first the extra packages are installed, if any, then the instances are run and checked.           |
| Completed | Every instance has a result. The score counts on the leaderboard if it is the team's best in the round.            |
| Failed    | The submission could not be evaluated at all. The reason is shown with it. It still counts toward the day's limit. |

Each instance runs in a fresh sandbox, so the submission evaluation time depends on number of problems and time limits per problem.

**Why a submission can fail**

| Message                                                   | What to do                                                                               |
| --------------------------------------------------------- | ---------------------------------------------------------------------------------------- |
| The archive has no solver for any problem of Round *n*    | Put one folder per open-round problem, named by its slug, with `solve.py` inside.        |
| The archive cannot be used: …                             | The archive could not be extracted. Pack it again with a standard zip or tar tool.       |
| The extra packages could not be installed                 | Read the build log on the dashboard, then fix the list.                                  |
| Installing the extra packages took longer than 10 minutes | Remove packages that build from source, or pin versions with wheels for Linux.           |
| Installing the extra packages ran out of memory           | As above.                                                                                |
| Evaluation did not complete                               | Something went wrong on our side and the submission stopped for two hours. Submit again. |

A problem on our side, such as a sandbox that fails to start, is retried automatically. If an instance still fails after 3 tries, it scores 0 with “Evaluation failed on our side”.

## What you see afterwards

On the dashboard's **Submissions** page, each submission shows its status and score. While it is queued or running, an estimate of when its results arrive takes the score's place. Open the submission to see:

- the environment: the image, the extra packages and their installed versions, or the build log if the install failed;
- every instance's result, grouped by problem: **Feasible** with its score, or the reason it scored 0;
- the archive you uploaded, to download.

The same data, except the estimate, is in [`GET /submissions/{id}`](https://frontieror-challenge.com/docs/main/api#get-a-submission). Solver logs, objective values and run times are not shown, since they could reveal the private instances. How the scores are computed is on [Evaluation and scoring](https://frontieror-challenge.com/docs/main/scoring).