pyproject.toml

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

[project]
name = "agentenv-framework"
version = "0.9.1389"
description = "SDK and CLI for building, deploying and running agent environments and the tasks that grade agents in them."
readme = "README.md"
license = "Apache-2.0"
license-files = ["LICENSE", "NOTICE", "THIRD_PARTY_NOTICES.md"]
authors = [{ name = "Scale AI" }]
requires-python = ">=3.11.4"
keywords = ["reinforcement-learning", "rl-environments", "ai-agents", "llm-agents", "agent-evaluation", "mcp", "model-context-protocol", "a2a", "sandbox"]
classifiers = [
    "Development Status :: 4 - Beta",
    "Environment :: Console",
    "Intended Audience :: Developers",
    "Intended Audience :: Science/Research",
    "Operating System :: MacOS",
    "Operating System :: POSIX :: Linux",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3 :: Only",
    "Programming Language :: Python :: 3.11",
    "Programming Language :: Python :: 3.12",
    "Programming Language :: Python :: 3.13",
    "Topic :: Scientific/Engineering :: Artificial Intelligence",
    "Topic :: Software Development :: Libraries :: Application Frameworks",
    "Typing :: Typed",
]
dependencies = [
    "agentenv-framework-protocol[computer]>=0.1.0",  # a computer_driver env's put reads the driver class with it
    "certifi>=2024.2.2",  # the public roots the local transfer CA's trust bundle carries
    "click>=8.0.0",
    "cryptography>=42",  # the local transfer CA
    "httpx>=0.27.0",
    "jsonschema>=4.20",  # checks an app's settings against its Settings' schema before a deploy
    "litellm>=1.96.0,<1.97",
    "mcp>=1.25,<2",  # same floor as env/gateway/requirements.txt and agentenv-protocol; 2.0.0 removed mcp.server.fastmcp, imported at module scope by env/gateway
    "packaging>=22",  # parses a bundle.toml's requires; evaluates a plugin's requirement markers with no extra
    "a2a-sdk==0.3.26",
    "pydantic>=2.12",  # model_validate(extra=...) reads agent responses tolerantly
    "pymongo>=4.7",
    "python-multipart>=0.0.13",  # the local grant server's upload forms
    "pyyaml>=6.0",
    "starlette>=0.40",  # the local grant server
    "uvicorn>=0.47",  # 0.47 adds ssl_context_factory, how the local grant server gets its TLS context
    "modal>=1.5.2.dev22",  # dev22 adds _experimental_create(tags=); dev2 added VM-mode V2 (vm_runtime)
    "e2b~=2.46.4",  # AsyncSandbox/AsyncTemplate APIs validated against the 2.46 line
    "fastapi>=0.110",  # the explorer, agent-env up
]

[project.optional-dependencies]
# agent_env.openenv: bundle tasks served as OpenEnv envs. No cap: OpenEnv ships a minor release a week.
openenv = [
    "openenv>=0.9",
]
# agent_env.training.trl: TRL's trainers on a bundle served as an OpenEnv env. TRL itself is the trainer's to pin.
trl = [
    "agentenv-framework[openenv]",
    "datasets>=3",
    "transformers>=5.2",  # renders the tools as TRL shows them to a model; 5.2 is what TRL's environment_factory needs
]
# The AWS store backends (S3, Secrets Manager, DynamoDB, ECR), selected by impl pointer in config.toml.
aws = [
    "boto3>=1.35.0",
    "botocore>=1.35.0",
]
# The Google Cloud store backends, selected by impl pointer in config.toml.
gcp = [
    "google-api-core>=2.24",
    "google-auth>=2.40",
    "google-cloud-secret-manager>=2.20",
    "google-cloud-storage>=3.0",
    "requests>=2.31",
]
# The sail_vm sandbox provider (Sail Research Sailboxes), selected with --sandbox sail_vm.
sail = [
    "sail~=0.12.8",  # Sailbox exec/fs/listener/egress APIs validated against the 0.12 line
]
vercel = [
    "anyio>=4,<5",
    "httpx2>=2.13.1,<3",
    "vercel-sandbox>=0.7.0,<0.8",  # provides the vercel.sandbox namespace without the full vercel bundle
]
# The Tensorlake sandbox provider, selected as the built-in `tensorlake` backend.
tensorlake = [
    "tensorlake~=0.5.140",  # AsyncSandbox APIs validated against 0.5.140 to 0.5.149
]
# The Tenki Sandbox VM provider, selected as the built-in `tenki` backend.
tenki = [
    "tenki[async]~=1.4.1",  # AsyncClient/AsyncSandbox exec, stdin, ports and egress APIs validated against the 1.4 line
]
# The boxd machine provider, selected as the built-in `boxd` backend.
boxd = [
    "boxd~=0.2.17",  # AsyncBoxd machines, exec, proxies, egress allowlist and delete-after validated against 0.2.17
]
# The Daytona sandbox provider, selected as the built-in `daytona` backend.
daytona = [
    "daytona~=0.224.0",  # AsyncDaytona sandboxes, sessions, file API, preview links and egress lists validated against 0.224
]
# The boat.dev sandbox provider, selected as the built-in `boat` backend.
boat = [
    "boat-sdk~=1.13.0",  # sandboxes, labels, hosted ports, detached commands and their output stream validated against 1.13
    "urllib3>=2.1.0,<3",  # the SDK's transport, whose errors are what a dropped connection raises
]
dev = [
    "agentenv-framework[aws,boat,boxd,daytona,gcp,openenv,sail,tensorlake,tenki,trl,vercel]",  # exercise portable optional providers and stores in CI
    "fastmcp>=3.4,<4",  # the standalone-fastmcp mount path of the protocol; <4 keeps mcp 1.x
    "Pillow>=10.4",  # the protocol's computer extra, whose tests the unit tier runs
    "av>=17",  # the protocol's computer extra, whose tests the unit tier runs and which the recording tests decode with
    "griffelib==2.3.0",  # the plugin API check; exact, since what it reports as a break changes between releases
    "tomli-w==1.2.0",  # the Codex agent's config writer, which only its image installs; its tests import it
    "moto>=5.0.0",
    "psycopg2-binary>=2.9.0",  # the gateway server module's driver; outside its container only the tests import it
    "pytest>=9.0",
    "pytest-asyncio>=0.23.0",
    "pytest-dependency>=0.6.0",
    "pytest-socket>=0.7.0",
    "pytest-timeout>=2.3.0",
    "pytest-xdist>=3.0.0",
    "ruff==0.16.10",  # the lint gate; exact, since what it reports changes between releases
]
smol = [
    "smolmachines>=1.27.0",
]

[project.scripts]
agent-env = "agent_env.cli:cli"

[project.entry-points."agent_env.bundles"]
agents = "agent_env"
computers = "agent_env"
drivers = "agent_env"
desktop = "agent_env.examples"
hello = "agent_env.examples"
echo-mcp = "agent_env.examples"
mail-mcp = "agent_env.examples"
next-env-event = "agent_env.examples"

[project.urls]
Homepage = "https://www.agentenvframework.com"
Documentation = "https://www.agentenvframework.com/docs"
Repository = "https://github.com/scaleapi/agentenv-framework"
Issues = "https://github.com/scaleapi/agentenv-framework/issues"
Changelog = "https://github.com/scaleapi/agentenv-framework/releases"

[tool.hatch.build.targets.wheel]
packages = ["src/agent_env"]
# The explorer UI *source* lives at src/agent_env/explorer/ui/ (a Next.js project).
# Keep it out of the wheel — only its built output ships, copied to explorer/static/
# by the release workflow. node_modules/out/.next are already .gitignore-excluded.
exclude = ["src/agent_env/explorer/ui"]
# The release workflow writes these and git ignores them; a build ships them when they're there.
artifacts = ["src/agent_env/env/published_images.json", "src/agent_env/explorer/static"]

[tool.hatch.build.targets.sdist]
only-include = ["src", "tst", "packages/agentenv-protocol", ".github/scripts"]
artifacts = ["src/agent_env/env/published_images.json", "src/agent_env/explorer/static"]

[tool.uv.workspace]
members = ["packages/agentenv-protocol"]

[tool.uv.sources]
agentenv-framework-protocol = { workspace = true }

[tool.pytest.ini_options]
asyncio_mode = "strict"
asyncio_default_test_loop_scope = "session"
asyncio_default_fixture_loop_scope = "session"
filterwarnings = [
    "ignore:Pydantic serializer warnings:UserWarning",
]
markers = [
    "integration: marks tests as integration tests (may require external services)",
    "installer: runs the real pip, uv and pipx offline against locally built wheels (tst/installer; `make installer-test`)",
    "int_test_slow: marks integration tests that take minutes per test (real Docker build / sandbox VM / Modal / gateway / agent). Excluded from the fast tier; opt in via `make int-test-slow`.",
]
log_cli = true
log_cli_level = "INFO"
log_cli_format = "%(asctime)s %(levelname)s %(name)s: %(message)s"

[tool.ruff.lint]
select = ["F"]