Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
Python agentagentic-aiagentic-frameworkagentic-workflowaiautomationclaudedeepseekgenaigptllmlow-codemcpnextjsno-codeopenaipythonskillsworkflow
"""Helpers for formatting shell output into bounded prompt-safe text.
The shell layer uses byte budgets, not Python character counts, because
shellctl output windows are byte-limited. These helpers keep UTF-8 boundaries
valid while producing a compact head/tail rendering for model-visible output.
"""
from __future__ import annotations
def utf8_prefix(text: str, max_bytes: int) -> str:
"""Return the longest UTF-8-safe prefix that fits within ``max_bytes``."""
if max_bytes <= 0:
return ""
return text.encode("utf-8")[:max_bytes].decode("utf-8", errors="ignore")
def utf8_suffix(text: str, max_bytes: int) -> str:
"""Return the longest UTF-8-safe suffix that fits within ``max_bytes``."""
if max_bytes <= 0:
return ""
return text.encode("utf-8")[-max_bytes:].decode("utf-8", errors="ignore")
def normalized_output_text(
head: str,
*,
tail: str | None,
output_path: str | None,
max_output_size_bytes: int,
truncated_in_middle: bool | None = None,
truncation_message: str | None = None,
) -> str:
"""Format bounded shell output with an optional truncation marker and log path."""
if truncated_in_middle is None:
truncated_in_middle = tail is not None and tail != head
if not truncated_in_middle:
return head
if truncation_message is None:
truncation_message = (
f"truncated in middle because the max output size is limited to {max_output_size_bytes} bytes"
)
parts = [
head,
f"... ({truncation_message}) ...",
]
if tail is not None:
parts.append(tail)
if output_path:
parts.append(f"(check the {output_path} for full output)")
return "\n".join(parts)
__all__ = ["normalized_output_text", "utf8_prefix", "utf8_suffix"]