Add function to parse replay

[pazqo]

After completing a puzzle on CTC, it's possible to save the replay, which is an encoded list of actions from the very beginning to the completion. There is already a function in script.js (Replay.decode) but we can expose this via rest api.
There could be another route in the app, which could be /parse_replay.

I already have a functioning version of the code (maybe needs some extra check as the replay might be broken), can I just submit a pull request or do you have other plans in mind?

I think this could be great to start from a solve, generate the replay, have an LLM clean, split and describe list of actions and generate more training data.

[jsseely]

Hi @pazqo !

We have some snippets like this internally. Namely, we have used

with open(replay_path, "r") as f:
    replay = json.load(f)

driver.execute_script(f"Framework.features.replaysave.loadPayload(JSON.stringify({replay}))")
replay_stack = json.loads(driver.execute_script("Framework.app.puzzle.replayStack"))

where replay_stack is a list of actions in the SudokuPad format ('sl:r1c1/10' means select r1c1 10*.05ms after the previous action).

(I recall there are several equivalent ways to call Replay.decode and that loadPayload was somehow the most reliable)

We opted to not include replay parsing right away but if you have several .replay files from valid solves and would find this functionality useful then we'll happily review a PR

[pazqo]

I'll see if I can make this more robust, then I'll submit a PR. Is there a slack/discord where I can communicate directly with the team? I actively working on this right now and I can iterate faster with a more direct support. Thanks!