Something fun to do - thanks.

Here is one solution (using many of the tricks already expressed by others) and some Common Table expressions to weed out the rest.

Assumptions:

• There is always a full path given
• The filename always contain both the year-month and the file number
• The year-month is always 6 digits long
• The file-number is always 6 digits long

Everything else should be dynamically handled.

  -- SQLite version 3.35.4  [ Release: 2021-04-02 ]  on  SQLitespeed version 2.1.3.11.
  -- ================================================================================================

  -- Create an example table containing example paths...

CREATE TABLE paths(pathname TEXT);

INSERT INTO paths VALUES
 ('D:\DOSYALAR\2021-07\0012090829\0012090829-202107-K-000000.zip')
,('D:\DOSYALAR\2021-07\0012090829\GIB-0012090829-202107-KB-000000.zip')
,('D:\DOSYALAR\2021-08\16270558554\GIB-16270558554-202108-YB-000000.zip')
,('D:\DOSYALAR\2021-08\35345023780\35345023780-202108-K-000000.zip')
,('D:\DOSYALAR\2021-08\9920362895\GIB-9920362895-202108-YB-000000.zip')
,('D:\DOSYALAR\2021-08\3310378950\3310378950-202108-K-000001.zip')
;


-- Show how the path-length identifier works. It keeps looking for the
-- next slash in the name, skips to it and cuts off the handled part,
-- using the row-id to identify the rows. (It could use the primary key
-- or unique index too, but my example did not have any):

WITH pathpos(path_id, slash_idx, remaining_path) AS (
    SELECT rowid, 0, pathname FROM paths
    UNION ALL
    SELECT path_id, slash_idx + instr(remaining_path,'\'), substr(remaining_path, instr(remaining_path, '\') + 1)
      FROM pathpos
     WHERE instr(remaining_path, '\') > 0  LIMIT 1000
), pathlen(path_id, last_slash_idx, remaining_path) AS (
    SELECT path_id, MAX(slash_idx), remaining_path
      FROM pathpos
     GROUP BY path_id
)
SELECT path_id, last_slash_idx AS directory_length, remaining_path AS filename
  FROM pathlen
;

  --             |director-|                                      
  --    path_id  | y_length|filename                              
  -- ------------|---------|--------------------------------------
  --       1     |    31   |0012090829-202107-K-000000.zip        
  --       2     |    31   |GIB-0012090829-202107-KB-000000.zip   
  --       3     |    32   |GIB-16270558554-202108-YB-000000.zip  
  --       4     |    32   |35345023780-202108-K-000000.zip       
  --       5     |    31   |GIB-9920362895-202108-YB-000000.zip   
  --       6     |    31   |3310378950-202108-K-000001.zip        




-- The next query shows using the above and more string searches and
-- sub-string handlers to get the rest of the information.

WITH pathpos(path_id, slash_idx, remaining_path) AS (
    SELECT rowid, 0, pathname FROM paths
    UNION ALL
    SELECT path_id, slash_idx + instr(remaining_path,'\'), substr(remaining_path, instr(remaining_path, '\') + 1)
      FROM pathpos
     WHERE instr(remaining_path, '\') > 0  LIMIT 1000
), pathlen(path_id, last_slash_idx, remaining_path) AS (
    SELECT path_id, MAX(slash_idx), remaining_path
      FROM pathpos
     GROUP BY path_id
), ym_start(path_id, yearmonth_idx, yearmonth_part) AS (
    SELECT path_id,
           instr(substr(remaining_path,5), '-') + 1,
	   substr(remaining_path, 5)
      FROM pathlen
), ym_end(path_id, yearmonth, fileno_idx, yearmonth_part) AS (
    SELECT path_id,
           substr(yearmonth_part, yearmonth_idx, 6),
           instr(substr(yearmonth_part, yearmonth_idx + 7), '-') + 1,
           substr(yearmonth_part, yearmonth_idx + 7)
      FROM ym_start
), all_info(path_id, directory, filename, type, yearmonth, fileno) AS (
    SELECT pathlen.path_id,
           substr(paths.pathname, 1, pathlen.last_slash_idx),
	   pathlen.remaining_path,
	   substr(ym_end.yearmonth_part, 1, ym_end.fileno_idx - 2),
	   ym_end.yearmonth,
	   substr(ym_end.yearmonth_part, ym_end.fileno_idx, 6)
      FROM pathlen
      JOIN ym_end ON ym_end.path_id = pathlen.path_id
      JOIN paths  ON paths.rowid    = pathlen.path_id
)
SELECT *
  FROM all_info
;


  --    path_id  |directory                         |filename                              | type|yearmonth| fileno 
  -- ------------|----------------------------------|--------------------------------------|-----|---------|--------
  --       1     |D:\DOSYALAR\2021-07\0012090829\   |0012090829-202107-K-000000.zip        |  K  |  202107 | 000000 
  --       2     |D:\DOSYALAR\2021-07\0012090829\   |GIB-0012090829-202107-KB-000000.zip   |  KB |  202107 | 000000 
  --       3     |D:\DOSYALAR\2021-08\16270558554\  |GIB-16270558554-202108-YB-000000.zip  |  YB |  202108 | 000000 
  --       4     |D:\DOSYALAR\2021-08\35345023780\  |35345023780-202108-K-000000.zip       |  K  |  202108 | 000000 
  --       5     |D:\DOSYALAR\2021-08\9920362895\   |GIB-9920362895-202108-YB-000000.zip   |  YB |  202108 | 000000 
  --       6     |D:\DOSYALAR\2021-08\3310378950\   |3310378950-202108-K-000001.zip        |  K  |  202108 | 000001 

  -- ------------------------------------------------------------------------------------------------

Please note: On a rather large dataset, this query will get the job done, but it will not be fast. For speed, nothing will beat the UDF suggestion made in another post.

Good luck!