Files
Javier Martinez f2cffd1ab9 feat: code execution v2 (#2278)
* fix: ensure that the publisher is ready to publish new messages

* fix: dockerfile

* fix: clones

(cherry picked from commit bc0a77e050)

* Potential fix for pull request finding

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* feat: M3 persistent sessions + M4 isolated bash + OpenSandbox provider

Introduces mount-aware sessions with canonical path abstraction:
- ContentBundle / SessionMount abstraction for skill/plugin mounting
- PathTranslator: rewrites commands and scrubs output (canonical ↔ real paths)
- LocalMount + ReadOnlyMount: local FS-backed mounts; read-only cache shared across sessions
- BashExecutor: asyncio subprocess with setsid + setrlimit isolation + killpg on timeout
- SkillLoader: downloads skill files from object storage as ContentBundles
- LocalCodeExecutionProvider: rewritten to use mounts, TTL reaper, BashExecutor
- OpenSandboxCodeExecutionProvider: new Docker/K8s backend via opensandbox SDK
- Async cascade: create_session / get_or_create_session / build_tool all async
- ObjectStorage.list_files() added to ABC + both implementations
- New settings: session_ttl_seconds, bash rlimit fields, OpenSandboxSettings

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* refactor: generic ContentBundle with BundledFile, remove skill_filter from code_execution layer

- Replace ContentBundle/dataclass with pydantic BaseModel; add BundledFile with path, content, permissions
- Update ReadOnlyMount to use list[BundledFile] with per-file chmod
- Remove skill_filter and SkillLoader from create_session() in base, local, and code_execution_component
- Move skill-to-bundle resolution into BashToolBuilder (inject SkillLoader there)
- Update SkillLoader.load() to return list[BundledFile] objects

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* fix: remove opensandbox

* feat: use sandbox inside of code executor

* fix: mypy

* feat: add workspace manager

* feat: improvements

* feat: add env abstraction

* feat: remove leftovers

* feat: improve skill content

* fix: macos

* fix: mypy

* feat: add code executor prompt

* feat: add content bundle in stack

* fix: mypy

* fix: layout

* feat: add skill prompt

* fix: order

* feat: allow to present final files

* feat: add container block

* fix: remove default config

* fix: session pers

* fix: do lazy env

* fix: mounter

* feat: refactor mounter

* fix: container

* feat: add requirements

* fix: add container registry

* fix: move to be lazy

* fix: mypy

* fix: current folder

* fix: stop sandbox in tabular

* fi: ensure to use absolute paths in bash & text editor

* feat: add files router

* feat: simplify

* feat: update present files

* fix: download files

* fix: mypy

* fix: tests

* fix: tests

* fix: bash executor in linux

* Revert "fix: bash executor in linux"

This reverts commit 483e208a96.

* Revert "fix: tests"

This reverts commit 50d9288f5e.

* test: remove test in ci

---------

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-06-29 18:29:29 +02:00

40 lines
1.1 KiB
Python

from typing import TypeVar
T = TypeVar("T")
def truncate_output(text: str, max_bytes: int) -> str:
encoded = text.encode("utf-8")
if len(encoded) <= max_bytes:
return text
return encoded[:max_bytes].decode("utf-8", errors="ignore") + "\n...[truncated]"
def is_list_of(
value: object,
typ: type[T],
) -> bool:
"""Check if the value is a list of the given type."""
return isinstance(value, list) and all(isinstance(item, typ) for item in value)
def find_common_prefix(s1: str, s2: str) -> str:
"""Finds a common prefix that is shared between two strings.
This function is provided as a UTILITY for extracting information from JSON
generated by partial_json_parser, to help in ensuring that the right tokens
are returned in streaming, so that close-quotes, close-brackets and
close-braces are not returned prematurely.
e.g. find_common_prefix('{"fruit": "ap"}', '{"fruit": "apple"}') ->
'{"fruit": "ap'
"""
prefix = ""
min_length = min(len(s1), len(s2))
for i in range(0, min_length):
if s1[i] == s2[i]:
prefix += s1[i]
else:
break
return prefix