ya-agent-environment
Environment abstractions for general agents.
ya-agent-environment provides the shared base interfaces used by YA agents:
EnvironmentFileOperatorShellResourceRegistry- resumable resources
- environment-owned temporary storage via
tmp_dirandresolve_tmp_path()
The Python import package is ya_agent_environment.
Environment and File Backend Contracts
Environment owns temporary-directory configuration and lifecycle. While entered,
env.tmp_dir exposes the agent-facing root (or None) and
env.resolve_tmp_path(relative_path) safely resolves a contained path, rejecting
absolute paths and parent traversal. Temporary paths are handled through the normal
FileOperator methods; there is no separate temporary-file operator or routing API.
Environment cleanup closes registered resources, the shell, and the file operator
before _teardown() releases backend/container and temporary-directory resources.
The same ordering is preserved after partial setup failures and during cancellation. The base class does not assume
a filesystem backing store and never deletes _tmp_dir itself: custom environments
must allocate their agent-facing temporary backend in _setup() and remove only owned
storage in _teardown().
A FileOperator implementation exposes one logical path space by implementing its
public abstract methods directly. read_bytes_stream() returns an AsyncIterator
directly, so consume the returned iterator without awaiting the method call.
Workspace-backed concrete environments allocate owned temporary instances below
.tmp/ya-agent-<id>. Each instance contains a self-ignoring .gitignore, and teardown
removes only that owned instance. Custom environments remain responsible for allocating
and tearing down their own temporary backend.
LocalFileOperator.walk_files() returns paths relative to default_path when possible
and directly reusable absolute paths for explicitly allowed roots outside it. glob and
grep may search those roots explicitly; root="." does not implicitly include
temporary storage.
Bounded Output
Shared output-bounding policy lives in ya_agent_environment.output. It provides
head/tail budget splitting, character and UTF-8 byte truncation, and incremental bounded
text accumulation. Shell ingestion and SDK previews should reuse these helpers instead
of defining independent truncation policies.
Shell Backend Contract
Custom Shell backends implement _create_process() and return an ExecutionHandle with stream, wait, and kill callbacks. Shell.execute() is the final public foreground boundary and must not be overridden: the base class owns process creation, timeout, cancellation, foreground/background registration, and session-reset cleanup. The foreground timeout budget begins before process creation, but shell-owned creation still resolves and terminates any eventual handle before reporting timeout. Class creation raises TypeError when final MRO resolution replaces execute(), so direct and mixin-based legacy overrides fail fast instead of silently bypassing lifecycle ownership; migrate their process creation into _create_process(). A wait callback must not report natural completion while owned descendants remain, and a kill callback must not return until backend execution is confirmed terminated; raising preserves the handle for a later reset retry. Backends must retain a stable execution identity through termination: a POSIX numeric PGID is no longer an ownership handle after its leader is reaped, so residual-group cleanup requires a live guardian or an equivalent platform-owned primitive.
Shell.start() returns a compact model-facing handle such as process-1. The handle is
sequential and unique within the current shell session; a successful
reset_background_processes() starts a fresh sequence. It is the only identifier
accepted by wait_process(), kill_process(), stdin, signal, status, and completion
APIs. It is not an OS PID, process-group ID, container exec identity, UUID, or durable
cross-session key. Those backend ownership identities remain private in
ExecutionHandle and transport-specific implementations. Foreground execution uses a
separate private identity and never consumes the model-facing sequence. A failed reset
retains both ownership state and the current sequence until termination is confirmed,
so a live handle is never reused.
Development
This package is maintained as a workspace member in ya-mono.
uv run python -m pytest packages/ya-agent-environment/tests -vv
uv run python -m pyright
uv build --package ya-agent-environment -o dist
1 maintainer
Metadata
1 maintainer
Release files for ya-agent-environment 2.11.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| ya_agent_environment-2.11.0.tar.gz | 59.0 kB | Without provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| ya_agent_environment-2.11.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 97.6 kB
Release files / ya_agent_environment-2.11.0.tar.gz
| Download URL | ya_agent_environment-2.11.0.tar.gz |
|---|---|
| Size | 59.0 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
e3d58249b26ea2b7c7302fd55e41c9d8171e214bf5ff3e586e2d41c43a0756bb
|
|
BLAKE2b-256 checksum How to use checksums |
ab93e4e672ca6b7006046aec6f46adf1d6f0f4cdae1e0dba4a4d38888db60a96
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.11.14 {"installer":{"name":"uv","version":"0.11.14","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Release files / ya_agent_environment-2.11.0-py3-none-any.whl
| Download URL | ya_agent_environment-2.11.0-py3-none-any.whl |
|---|---|
| Size | 38.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
4a6228aba641c4dd4c3a2e5a407a3dd671c30595c5c7caf5200c0b894f2b5a38
|
|
BLAKE2b-256 checksum How to use checksums |
0c41db1096be88f216a855403a3280cc01939709255cf139a3bdc7d6e64d4afd
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
uv/0.11.14 {"installer":{"name":"uv","version":"0.11.14","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Ubuntu","version":"24.04","id":"noble","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":true}
|
Security for ya-agent-environment 2.11.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.
