Install the Python SDK, run a command on a computer and look up the client, computers, exec, files and errors
The Python SDK runs computers from Python code. It has two clients with the
same resources and methods. Nevia blocks and AsyncNevia awaits. The SDK is
in early access, so names and behaviour can change.
The quick start creates a computer, runs a command on it, writes and reads a file and deletes the computer. The reference after it covers the client, computers, exec, files and errors.
Prerequisites
You need:
- Python 3.11 or later.
- A Nevia workspace where you have the admin or editor role.
- The
neviaCLI, signed in withnevia login. The Quickstart shows how to install it.
Install the SDK
Invited testers only
The SDKs are in early access for invited testers. To ask for access, see Request access to Nevia.
From your project's root, download the SDK bundle. It unpacks to
vendor/nevia-sdk-0.1.0/:
curl -fsSL --oauth2-bearer "$(nevia auth access-token)" https://sdk.nevia.cloud/0.1.0.tar.gz | tar -xzThen install the SDK from it:
uv add ./vendor/nevia-sdk-0.1.0/python/nevia-0.1.0-py3-none-any.whlAuthenticate
The SDK uses your nevia login session, so a signed-in machine needs nothing
more. A server or CI job uses a service-account key instead:
export NEVIA_KEY_FILE="$HOME/path/to/key.json"Use an automation key file shows how to
create one. If you belong to more than one workspace, set NEVIA_WORKSPACE to
the workspace ID too.
Create a computer and run a command
Save this as quickstart.py:
from nevia import Nevia
with Nevia() as nevia:
with nevia.computers.create(image='docker.io/library/python:3.13-slim') as computer:
print(f'Computer: {computer.id}')
result = computer.exec('python3 --version')
print(f'Exit code: {result.exit_code}, output: {result.stdout.strip()}')
computer.files.write('/tmp/hello.txt', 'Hello from Nevia!\n')
print(computer.files.read_text('/tmp/hello.txt'), end='')Run it:
python quickstart.pyThe computer ID and Python version in the output vary:
Computer: 01d69b2m9d26k14b244xjwmp9mb
Exit code: 0, output: Python 3.13.18
Hello from Nevia!computers.create() returns once the computer is ready, so the first command
runs at once. A command that exits nonzero still returns a result. Check
result.exit_code or result.ok to see whether it succeeded.
Clean up
The with block deletes the computer when it ends, also when the code inside
it raises. nevia ls confirms the computer is gone.
To keep a computer after the program ends, create it without with and save
its ID:
computer = nevia.computers.create(image='docker.io/library/python:3.13-slim')
print(computer.id)A later program reconnects with nevia.computer(saved_id) and calls
computer.delete() when it is done.
Reference
Sync and async clients
Nevia and AsyncNevia have the same resources, methods and arguments. The
tables on this page list each method once and apply to both. Pick one client
per program: Nevia for scripts and blocking code, AsyncNevia for code that
already runs an event loop, such as an async web server or agent framework.
Where the two differ:
Nevia (sync) | AsyncNevia (async) | |
|---|---|---|
| Calling a method | computer.exec(...) | await computer.exec(...) |
| Scoped client | with Nevia() as nevia: | async with AsyncNevia() as nevia: |
| Closing the client | nevia.close() | await nevia.aclose() |
| Deleting a computer at the end of a block | with nevia.computers.create(...) as computer: | async with nevia.computers.create(...) as computer: |
When create() creates | At the call | When awaited or entered |
| Deleting a branch at the end of a block | Not available, call delete() | async with computer.branch() as branch: |
| Listing | for computer in nevia.computers.list(): | async for computer in nevia.computers.list(): |
| Workspace ID | nevia.workspace() returns a str | await nevia.workspace() |
credentials= callable | A plain function | An async function |
| Deadlines | with nevia.fail_after(30): | with anyio.fail_after(30): |
| Resource classes | Computer, Files, Process, Database | AsyncComputer, AsyncFiles, AsyncProcess, AsyncDatabase |
Results and errors are the same classes in both: ExecResult, ComputerInfo,
FileStat, DirEntry and NeviaError.
The deadline helpers bound every blocking call the SDK makes. fail_after
raises TimeoutError when the time is up, move_on_after ends the block
without raising and shield() lets cleanup run past either one:
import nevia
with nevia.fail_after(30):
computer.exec('npm test')Client
Nevia(
*,
base_url=None,
credentials=None,
key_file=None,
key=None,
workspace=None,
)| Argument | Description |
|---|---|
credentials | Token string, or a callable that returns one |
key_file | Path to a service-account key file |
key | Service-account key as JSON text or a dict |
workspace | Workspace ID (defaults to your only workspace) |
base_url | API base URL (defaults to production) |
The client looks for credentials in this order:
- The
credentials,key_fileorkeyargument. NEVIA_TOKEN.NEVIA_KEY_FILE, orNEVIA_KEYwith the key's JSON.- Your
nevia loginsession.
The workspace is the workspace argument, then NEVIA_WORKSPACE, then the
only workspace you belong to.
Use the client as a context manager, or call close() (aclose() on
AsyncNevia) when you are done. Closing the client leaves your computers
running.
| Member | Description |
|---|---|
nevia.computers | Create, get, find and list computers |
nevia.computer(id) | Handle to an existing computer, without a request |
nevia.workspace() | ID of the selected workspace |
nevia.workspace(id) | Client scoped to another workspace, on the same connections |
Computers
| Method | Description |
|---|---|
computers.create(*, image, name=None, env=None, protected=None, wait=True) | Create a computer and wait until it is ready |
computers.get(id) | Get a computer by ID |
computers.find(name) | First computer whose name is exactly name, or None |
computers.list() | Iterate over every computer in the workspace |
create() returns the computer. Inside a with block the SDK deletes the
computer when the block ends. If the computer fails to become ready, or the
wait times out or is interrupted, create() deletes it before raising.
On a computer:
| Member | Description |
|---|---|
id, name, state | Identity and lifecycle state |
info | Full record as a ComputerInfo |
refresh() | Fetch the latest record |
sleep() | Sleep the computer and wait until it is sleeping |
wake() | Wake the computer and wait until it is ready |
branch(*, name=None) | Create a branch of the computer and return it |
delete() | Delete the computer (succeeds if it is already gone) |
A branch is a new computer. AsyncNevia deletes it at the end of an
async with block. With Nevia, call delete() on the branch when you are
done.
Exec
computer.exec(command, *, cwd=None, env=None, stdin=None, timeout=None, transport='api', max_output_bytes=65536)exec() runs a command to completion and returns an ExecResult.
| Argument | Description |
|---|---|
command | Shell string (runs under sh -c), or a list of arguments |
cwd | Working directory on the computer |
env | Extra environment variables for this command |
stdin | Text sent to the command's standard input |
timeout | Seconds before the command is stopped |
transport | 'api' (default) or 'direct' for direct exec |
max_output_bytes | Bytes kept from the end of each stream with direct exec |
When timeout passes, the SDK stops the command's process group and returns
the output so far with timed_out=True.
Exec runs through the API by default, which ends a request after about 50
seconds. Direct exec sends the command straight to the computer instead, so it
suits longer commands. It keeps the last 64 KiB of each output stream and sets
truncated=True when it drops output. See Limits and workarounds.
ExecResult field | Description |
|---|---|
exit_code | Exit code of the command |
signal | Signal that ended the command (0 when none) |
stdout, stderr | Output of the command |
timed_out | Whether timeout stopped the command |
truncated | Whether direct exec dropped output |
ok | True when exit_code and signal are 0 and the command did not time out |
Files
computer.files reads and writes files on the computer. Every path is a path
on the computer.
| Method | Description |
|---|---|
read(path) | File content as bytes, or None when there is no file |
read_text(path, encoding='utf-8') | File content as str, or None when there is no file |
write(path, data) | Write a whole file from str or bytes, creating parent directories |
stat(path) | FileStat with type, size and times, or None when nothing is there |
list(path, *, recursive=False, max_depth=1) | Entries in a directory as DirEntry values, sorted by name |
remove(path, *, recursive=False, force=False) | Remove a file, or a directory tree with recursive |
write() replaces the file in one step, so a reader never sees it half
written. list() raises FileNotFoundError for a missing path and
NotADirectoryError for a file.
Errors
The SDK raises every SDK and API error as a NeviaError. A command that exits
nonzero is a result, not an error.
from nevia import Nevia, NeviaError
with Nevia() as nevia:
try:
computer = nevia.computers.get('01d4cnjk3xg90y576w8y6t4gdwm')
except NeviaError as error:
if error.code != 'not_found':
raise
print('No such computer')| Field | Description |
|---|---|
status | HTTP status (0 when the SDK refused the call before sending it) |
code | Error code, such as not_found or invalid_state |
message | Message for a person to read |
details | Extra facts about the error, as a dict |
request_id | API request ID to quote when you report a problem |
retry_after | Seconds to wait before retrying, when the API sent one |
resource | What the call had created before it failed, so you can delete it |
New codes can appear, so treat a code you don't recognise as a generic
failure. is_nevia_error(error, code) checks the type and the code in one
call.