Python Code Verification Skill#

class council.skills.python.PythonCodeVerificationSkill(code_template: str = '')[source]#

Bases: SkillBase

Skill that verifies a given python code. It verifies: - the code is parseable using ast.parse - the code follows an optional code template.

The python code is retrieved from the message content from context.try_last_message, looking for a markdown python code block.

Below is an example of code template:

def say_hi() -> str:
# COUNCIL NO EDIT BEFORE THIS LINE

    pass

# COUNCIL NO EDIT AFTER THIS LINE

print(say_hi())

The two magic comments # COUNCIL … are used to identify the reference code, respectively above and below each magic comment.

The verification process will fail if any non-empty lines of the reference code are missing in the given code.

The verification process relies on ast.parse and ast.unparse to standardize the code. As such, the reference code must be formatted in such a way it is not impacted by the standardization process, a.k.a the reference code must be stable.

Below is a non-exhaustive list for good practices for the reference code:

  • no comments, other than the magic comments

  • empty lines are ok as they will be ignored

__init__(code_template: str = '')[source]#

initialize a new instance

Parameters:

code_template – a code template use to validate the python code retrieved from the context.

build_success_message(message: str, data: Any | None = None) ChatMessage#

Builds a success message for the skill with the provided message and optional data.

Parameters:
  • message (str) – The success message.

  • data (Any, optional) – Additional data to include in the message. Defaults to None.

Returns:

The success message.

Return type:

ChatMessage

Raises:

None

execute(context: SkillContext) ChatMessage[source]#

Executes the skill on the provided chain context and budget.

Parameters:

context (SkillContext) – The context for executing the skill.

Returns:

The result of skill execution.

Return type:

ChatMessage

Raises:

None

execute_skill(context: SkillContext) ChatMessage#

Skill execution

property name#

Property getter for the skill name.

Returns:

The name of the skill.

Return type:

str

Raises:

None

render_as_dict(include_children: bool = True) Dict[str, Any]#

returns the graph of operation as a dictionary

render_as_json() str#

returns the graph of operation as a JSON string

run_in_current_thread(context: ChainContext, iteration_context: Option[IterationContext]) None#

Run the skill in the current thread

run_skill(context: ChainContext, executor: ThreadPoolExecutor) None#

Run the skill in a different thread, and await for completion