feat(sandbox): add encode_base64 opt-in to read_file tool with backward-compatible default

[Copilot]

read_file in CodeInterpreterToolSet lacked a way to safely handle binary files, as it only returned raw text with no encoding control.

Changes

• CodeInterpreterToolSet.read_file: default behavior returns plain text unchanged (backward compatible); response now includes an encoding field ("raw" or "base64")
• Added encode_base64: bool = False parameter — set True to receive base64-encoded content, useful for binary files
• Handles both str and bytes from the underlying API when encoding

Example

# Default — plain text, backward compatible
result = toolset.read_file(path="/tmp/notes.txt")
# {"path": "/tmp/notes.txt", "content": "hello world", "encoding": "raw"}

# Opt-in base64 for binary files
result = toolset.read_file(path="/tmp/output.bin", encode_base64=True)
# {"path": "/tmp/output.bin", "content": "aGVsbG8=", "encoding": "base64"}

import base64
content = base64.b64decode(result["content"])

[OhYee]

@copilot 如果要保持向前兼容，是否应该默认原文，传入 base64 后使用 base64？

[Copilot]

@copilot 如果要保持向前兼容，是否应该默认原文，传入 base64 后使用 base64？

已按此思路调整（commit ed0b462）：

• 默认行为改回返回原始文本（encoding: "raw"），保持向前兼容
• 新增 encode_base64: bool = False 参数，传入 True 时返回 base64 编码内容（encoding: "base64"）

# 默认：原文（向前兼容）
result = toolset.read_file(path="/tmp/notes.txt")
# {"path": "...", "content": "hello world", "encoding": "raw"}

# 按需开启 base64（适合二进制文件）
result = toolset.read_file(path="/tmp/output.bin", encode_base64=True)
# {"path": "...", "content": "aGVsbG8=", "encoding": "base64"}