Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code: DatasetGenerationError
Exception: CastError
Message: Couldn't cast
episode_id: string
family: string
type: string
mode: string
program: string
user_request: string
system: string
server_url: string
transcript_note: string
steps: list<item: struct<call: struct<name: string, arguments: string>, result: string>>
child 0, item: struct<call: struct<name: string, arguments: string>, result: string>
child 0, call: struct<name: string, arguments: string>
child 0, name: string
child 1, arguments: string
child 1, result: string
final_message: string
split: struct<part1_end_index: int64, recap: struct<recap: struct<episode_id: string, part: int64, paths_wr (... 156 chars omitted)
child 0, part1_end_index: int64
child 1, recap: struct<recap: struct<episode_id: string, part: int64, paths_written: list<item: string>, last_result (... 117 chars omitted)
child 0, recap: struct<episode_id: string, part: int64, paths_written: list<item: string>, last_result: struct<statu (... 102 chars omitted)
child 0, episode_id: string
child 1, part: int64
child 2, paths_written: list<item: string>
child 0, item: string
child 3, last_result: struct<status: string, path: string, content: string, truncated: bool, total_tokens: int64>
child 0, status: string
child 1, path: string
child 2, content: string
child 3, truncated: bool
child 4, total_tokens: int64
child 4, user_request: string
to
{'episode_id': Value('string'), 'family': Value('string'), 'type': Value('string'), 'mode': Value('string'), 'program': Value('string'), 'user_request': Value('string'), 'system': Value('string'), 'server_url': Value('string'), 'transcript_note': Value('string'), 'steps': List({'call': {'name': Value('string'), 'arguments': Json(decode=True)}, 'result': Json(decode=True)}), 'final_message': Value('string')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1827, in _prepare_split_single
for key, table in generator:
^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
for item in generator(*args, **kwargs):
~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
episode_id: string
family: string
type: string
mode: string
program: string
user_request: string
system: string
server_url: string
transcript_note: string
steps: list<item: struct<call: struct<name: string, arguments: string>, result: string>>
child 0, item: struct<call: struct<name: string, arguments: string>, result: string>
child 0, call: struct<name: string, arguments: string>
child 0, name: string
child 1, arguments: string
child 1, result: string
final_message: string
split: struct<part1_end_index: int64, recap: struct<recap: struct<episode_id: string, part: int64, paths_wr (... 156 chars omitted)
child 0, part1_end_index: int64
child 1, recap: struct<recap: struct<episode_id: string, part: int64, paths_written: list<item: string>, last_result (... 117 chars omitted)
child 0, recap: struct<episode_id: string, part: int64, paths_written: list<item: string>, last_result: struct<statu (... 102 chars omitted)
child 0, episode_id: string
child 1, part: int64
child 2, paths_written: list<item: string>
child 0, item: string
child 3, last_result: struct<status: string, path: string, content: string, truncated: bool, total_tokens: int64>
child 0, status: string
child 1, path: string
child 2, content: string
child 3, truncated: bool
child 4, total_tokens: int64
child 4, user_request: string
to
{'episode_id': Value('string'), 'family': Value('string'), 'type': Value('string'), 'mode': Value('string'), 'program': Value('string'), 'user_request': Value('string'), 'system': Value('string'), 'server_url': Value('string'), 'transcript_note': Value('string'), 'steps': List({'call': {'name': Value('string'), 'arguments': Json(decode=True)}, 'result': Json(decode=True)}), 'final_message': Value('string')}
because column names don't match
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1879, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
episode_id string | family string | type string | mode string | program string | user_request string | system string | server_url string | transcript_note string | steps list | final_message string |
|---|---|---|---|---|---|---|---|---|---|---|
crud_e1_crud_flask_inventory_01 | crud | E1 | [mode: app:crud] | crud_flask_inventory_01 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5376 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5376 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_02 | crud | E1 | [mode: app:crud] | crud_flask_inventory_02 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5377 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5377 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_03 | crud | E1 | [mode: app:crud] | crud_flask_inventory_03 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5378 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5378 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_04 | crud | E1 | [mode: app:crud] | crud_flask_inventory_04 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5379 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5379 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_05 | crud | E1 | [mode: app:crud] | crud_flask_inventory_05 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5380 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5380 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_06 | crud | E1 | [mode: app:crud] | crud_flask_inventory_06 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5381 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5381 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_07 | crud | E1 | [mode: app:crud] | crud_flask_inventory_07 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5382 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5382 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_08 | crud | E1 | [mode: app:crud] | crud_flask_inventory_08 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5383 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5383 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_09 | crud | E1 | [mode: app:crud] | crud_flask_inventory_09 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5384 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5384 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_10 | crud | E1 | [mode: app:crud] | crud_flask_inventory_10 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5385 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5385 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_11 | crud | E1 | [mode: app:crud] | crud_flask_inventory_11 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5386 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5386 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_12 | crud | E1 | [mode: app:crud] | crud_flask_inventory_12 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5387 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5387 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_13 | crud | E1 | [mode: app:crud] | crud_flask_inventory_13 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5388 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5388 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_14 | crud | E1 | [mode: app:crud] | crud_flask_inventory_14 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5389 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5389 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_15 | crud | E1 | [mode: app:crud] | crud_flask_inventory_15 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5390 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5390 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_16 | crud | E1 | [mode: app:crud] | crud_flask_inventory_16 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5391 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5391 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_17 | crud | E1 | [mode: app:crud] | crud_flask_inventory_17 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5392 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5392 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_18 | crud | E1 | [mode: app:crud] | crud_flask_inventory_18 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5393 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5393 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_19 | crud | E1 | [mode: app:crud] | crud_flask_inventory_19 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5394 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5394 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_20 | crud | E1 | [mode: app:crud] | crud_flask_inventory_20 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5395 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5395 (left running). Health plus exercise asserts all pass. |
crud_e1_crud_flask_inventory_21 | crud | E1 | [mode: app:crud] | crud_flask_inventory_21 | Build a tiny inventory manager: JSON API plus a plain HTML page using fetch. Backend file 60 lines or fewer. | You are Engine 2, a coding agent that acts through four tools — read_file, write_file, edit_file, execute_terminal_command. Shell commands are bash (POSIX subset); Python is invoked as 'python'. Build only from file contents you have actually read; never invent file contents, tool results, or designs. | http://127.0.0.1:5396 | server left running in transcript; generator killed the instance after the verdict (never in transcript) | [
{
"call": {
"name": "write_file",
"arguments": {
"path": "requirements.txt",
"content": "flask>=3.0\npython-dotenv>=1.0\n"
}
},
"result": {
"status": "ok"
}
},
{
"call": {
"name": "write_file",
"arguments": {
"path": "app.py",
... | Done. Server running at http://127.0.0.1:5396 (left running). Health plus exercise asserts all pass. |
End of preview.
SilentBeast-Corpus
Supporting data artifact for a Kaggle research grant application (project B9 / Engine2, a from-scratch code-pretraining run).
This repository contains the synthetic portion of the training corpus, the BeastTok tokenizer trained on the full corpus, and the provenance manifests documenting how the (not-included) real corpus was assembled. It is designed to be fully redistributable: everything here was generated by this project or authored by this project.
Contents
| Path | What it is | Volume |
|---|---|---|
data/snippets_synthetic.jsonl |
Synthetic code snippets (DS/algo families) with per-row license + provenance fields. Every row is kind: synthetic_snippet. |
4,756 rows, 5.1 MB |
data/episodes_tk/ |
Synthetic Tkinter tool-use episodes (calculator + calendar apps) as per-episode JSON: system prompt, user request, multi-step tool-call transcript, program, verifier report. | 1,722 episodes, 11.4 MB |
data/episodes_apps/ |
Synthetic CRUD/landing web-app tool-use episodes (Express/Flask/HTML) in the same format. | 1,309 episodes, 8.8 MB |
tokenizer/ |
BeastTok: a 32,000-vocab BPE tokenizer trained from scratch on the full corpus (real + synthetic sample), plus its config, locked special-token list, and training report. | ~5 MB |
manifests/ |
Aggregate merge statistics (manifest.json), exact token counts under both tokenizers, dedup exemptions, eval-carveout composition, and the full corpus-merge report (REPORT7). |
metadata |
What is NOT included, and why
The real-code corpus (2,642,045 documents, 11.9 GB, 2,384,653,839 BeastTok tokens) is intentionally not in this repository. It is a merged derivative of third-party datasets whose licenses govern redistribution, and the merged form carries no per-row license attribution. It was used for research training under source-appropriate terms; it is not re-publishable here. The upstream sources are all publicly available on Hugging Face and are cited below. Also excluded: the 763-item evaluation carveout (frozen for evaluation only) and raw tokenizer-training sample material.
Corpus provenance (reference — real corpus not hosted here)
The real training corpus was assembled from five Hugging Face sources, license-filtered (per-row permissive allowlist for codeparrot and The Stack v2 dedup; 204,337 copyleft rows dropped), content-hash deduplicated, and tokenized with BeastTok:
| Source | Docs | Tokens (StarCoder-ref) | Share |
|---|---|---|---|
| starcoderdata (The Stack v1) | 1,475,263 | 1,374,758,434 | 47.8% |
| The Stack v2 dedup | 580,490 | 776,995,995 | 27.0% |
| codeparrot (github-code) | ~488,600 | 643,196,734 | 22.4% |
| opc-annealing-corpus | 97,105 | 81,775,362 | 2.8% |
| CodeSearchNet | 626 | 166,053 | ~0% |
| Total | 2,642,045 | 2,876,892,578 |
Exact BeastTok recount: 2,384,653,839 tokens (−17.1% vs the
StarCoder-ref accounting; an in-domain 32k BPE is more efficient than
the generic 49k reference vocab). Full methodology: manifests/.
Synthetic data design
The synthetic set teaches tool-use discipline for a code agent
harness: episodes are multi-step transcripts (read → plan → edit →
verify) over small generated programs, with verifier reports and
genuine failure/repair paths; snippets are single-turn
prompt→code pairs. All content is generated — no real user code is
embedded in these files. Volumes are locked by the project spec
(SYNTH_SPECS v5); the snippet family is MIT per its per-row license
field.
BeastTok
Trained from scratch (HF tokenizers, BPE + Whitespace,
BpeTrainer(min_frequency=1)) on a 504.1 MB stratified sample of the
merged corpus; vocab 32,000 realized exactly, 11 special tokens (ids
0..10, id 10 = document/episode separator), byte-verified. See
tokenizer/TRAIN_REPORT.md.
Upstream datasets (cited, not redistributed)
- The Stack v1 (starcoderdata) — https://huggingface.co/datasets/bigcode/starcoderdata
- The Stack v2 dedup — https://huggingface.co/datasets/bigcode/the-stack-v2-dedup
- codeparrot (github-code) — https://huggingface.co/datasets/codeparrot/github-code
- CodeSearchNet — https://huggingface.co/datasets/code_search_net
- OpenCoder opc-annealing-corpus — https://huggingface.co/datasets/OpenCoder-LLM/opc-annealing-corpus
License
Repository content (synthetic data, tokenizer, documentation,
manifests): Apache-2.0. Individual snippet rows additionally carry a
per-row license: MIT field describing the intended license of the
generated snippet content.
- Downloads last month
- 135