These are single-purpose scripts - not only useless by themselves, but even when integrated (or planned to be integrated) into a larger pipeline, of interest only to people who 1) speak Czech, and 2) are interested in Czech Open Data and/or public procurement.
www.hlidacstatu.cz indexes and curates data made available (compulsorily and not very conveniently) by Czech public administration - public contracts, companies servicing them, politicians on their boards etc. Among other things, it publishes a developer API (described at https://hlidacstatu.docs.apiary.io/ ) allowing registered 3rd parties to upload structured data they consider relevant to the site's mission.
The data is modelled as a dataset - a sequence of JSON objects with a mandatory "Id" attribute. Every dataset has a JSON schema, defined by dataset's author. When one has some JSON data, the schema can be produced semi-automatically by a JSON schema generator and incorporated in the dataset description, quoting the schema JSON as a single string value (as described in the documentation referenced above). For presentation, the dataset description also requires a couple of templates specifying how to format the JSON data as HTML, also quoted when included.
The dataset-roundtrip.py script moves these structured values from the dataset description into separate files and back, so that they can be edited in their native format. What should be moved where is specified by a "matrix" JSON file - basically the dataset description where the quoted values were replaced by strings starting with '@' and continuing with file names. It's the matrix's author responsibility to use unique and valid names - the script just uses them as specified. dataset-roundtrip.py can run in 2 modes: "pack" combines matrix and the files named in it into the dataset description (which can then be manually uploaded to hlidacstatu.cz) while "unpack" copies values from the dataset description into files specified in the matrix. All parameters of this script are specified on its command line - for details, run it without any.
Individual dataset items can also be uploaded manually, but for collections of useful size that's impractical, because each item must be POSTed separately. The pusher.py script automates this cycle. For configuration, it requires a config.json file in the current directory. config.json has 2 mandatory settings: "token" for the API access token, available from hlidacstatu.cz upon registration, and "datasetId" for the uploaded dataset name, returned when it's created. The "mode" setting specifies how to handle uploaded items that already exist on the server: "rewrite" to replace them and "merge" to combine them. The "sourceDir" setting specifies the directory with files to upload; its value is "json" by default. Note that to mark progress, contents of the "sourceDir" are destroyed after succesfull upload - make sure it's not your only copy!
Files in the "json" directory represent dataset items to be uploaded - every file must have the item's Id as a name, plus a ".json" extension (e.g. "a1fa5ee3-d57a-4065-bde3-50d40a1ce269.json" for an item with Id "a1fa5ee3-d57a-4065-bde3-50d40a1ce269"). The contents must be a JSON object conforming to the dataset's schema, including the same Id that's in the name of the file. pusher.py simply goes through the "json" directory, POSTs the files, checks the API's return value and either removes the uploaded file on success, or ends on error.