Metadata-Version: 2.3
Name: encore-dev
Version: 2.0.0b8
Summary: Encore SDK for Python — infrastructure resources
Author: Encore
License: MPL-2.0
Classifier: Programming Language :: Python :: 3
Requires-Dist: opentelemetry-api>=1.27
Requires-Dist: typing-extensions>=4.13
Requires-Dist: aiohttp>=3.14.4 ; python_full_version >= '3.10' and extra == 'aiohttp'
Requires-Dist: connectrpc>=0.10.1,<0.11 ; python_full_version >= '3.11' and extra == 'grpc'
Requires-Dist: pyqwest>=0.12 ; python_full_version >= '3.11' and extra == 'grpc'
Requires-Dist: httpx>=0.27 ; extra == 'httpx'
Requires-Dist: niquests>=3.19 ; extra == 'niquests'
Requires-Dist: requests>=2.34 ; python_full_version >= '3.10' and extra == 'requests'
Requires-Python: >=3.9
Provides-Extra: aiohttp
Provides-Extra: grpc
Provides-Extra: httpx
Provides-Extra: niquests
Provides-Extra: requests
Description-Content-Type: text/markdown

# Encore SDK for Python

The Python SDK for [Encore](https://encore.dev) — **infrastructure resources only**.

Declare infrastructure (SQL databases, pub/sub topics & subscriptions, object
storage buckets, Redis databases, cron jobs, metrics, secrets) as module-level variables.
Encore discovers these declarations, provisions the resources, and connects your
application to them at runtime.

Services are declared by the language-agnostic `encore.service.json` file, not in
Python. Use the SDK with your preferred Python web framework.

## Installation

Requires Python 3.9 or later. Install from PyPI with pip:

```sh
python -m pip install encore-dev
```

Or add it to a uv project:

```sh
uv add encore-dev
```

## Declare resources

```python
from encore import pubsub, secrets
from encore.storage import postgres

db = postgres.Database("orders", migrations="./migrations")
events = pubsub.Topic("order-events", exactly_once=True)
api_key = secrets.Secret[str]("ApiKey")


async def place_order(order):
    await db.exec("INSERT INTO orders (id) VALUES ($1)", order.id)
    await events.publish({"id": order.id})
```

## Secrets

`secrets.Secret[str]("Name")` belongs to the service that declares it.
`secrets.Secret[str]("Name", scope="shared")` is one secret for the whole app,
given to the services that read it, and `scope="global"` one that every service
is given. `value` reads the secret as a string, and raises `KeyError` when it
has no value. A secret's value is resolved once per process: a new value
reaches a service when it restarts.

A shared secret is declared once in the app. Another Python service can't
import the declaring service's modules, so it references the secret by name
instead:

```python
from encore.secrets import Secret

support_email = Secret[str].named("SupportEmail")
```

Reading `value` makes the service one the secret is given to. The declaration
can be in any language. Encore reports a reference to a name that isn't
declared as a shared secret.

## Service URLs

Use `base_url` to look up a service's configured base URL:

```python
from encore.services import base_url

order = f"{base_url('orders')}/orders/123"
```

The returned string has no trailing slash: a call's path goes after it. Unknown
service names raise `ValueError`. Lookup reads the runtime configuration and
makes no network request.

Looking up a service's URL makes the calling service depend on it, which Encore
reads from the code: pass the name as a string literal, call `base_url` directly
rather than passing it around, and call it in a service.

### Calling services

To call a service, use the adapter for your HTTP client. Each adapter re-exports
`encore.services` and adds `client(service)`, a client whose relative URLs are
the service's: the async one where the library has both kinds, with
`sync_client(service)` for the other:

```python
from encore.services import httpx as services

billing = services.client("billing")


async def invoices():
    response = await billing.get("/invoices")
    return response.json()
```

| Adapter | Install | `client` returns | `sync_client` returns |
|---|---|---|---|
| `encore.services.httpx` | `encore-dev[httpx]` | `httpx.AsyncClient` | `httpx.Client` |
| `encore.services.requests` | `encore-dev[requests]` | `requests.Session` | |
| `encore.services.aiohttp` | `encore-dev[aiohttp]` | `aiohttp.ClientSession` | |
| `encore.services.niquests` | `encore-dev[niquests]` | `niquests.AsyncSession` | `niquests.Session` |

The `requests` and `aiohttp` extras need Python 3.10: on 3.9 they install
nothing, as their releases for 3.9 have known vulnerabilities, and the adapters
use the client the app installed.

`client` and `sync_client` pass their other arguments to the client's
constructor, as keywords. A `requests.Session` takes none, so the requests
adapter's `client` sets them on the session instead, as
`client("billing", headers={"x-tenant": "a"}, verify=False)`, and takes a
`timeout` for each request that sets none. The same rules as `base_url` apply
to the service's name. aiohttp joins relative URLs as
links are, so write them without the leading slash there: `get("invoices")`;
and it creates a session only in a running event loop, so call its `client` in
a coroutine, such as a lifespan handler.

The clients send the Request Context of the calls made with them to the app's
services, and never to anything else, redirects included. A client you build
yourself can do the same with the adapter's hook: `Transport` and
`AsyncTransport` for httpx, the requests `Adapter`, the aiohttp `middleware`, or the
niquests `pre_request` hook.

### Generated clients

A service that lists another in its `encore.service.json` `clients` gets
clients for it in `encore_gen/clients/`: a package per service, with a module
per entry of the service's `api`, named after the entry's `name` (`openapi`,
`graphql` or `grpc` for one that sets none). Each operation is an async
function of its module, with nothing to set up first:

```python
from encore_gen.clients import catalog

product = await catalog.openapi.get_product("xyz")
products = await catalog.storefront.products(skus=["xyz"])
stuff = await catalog.grpc.stuff_service.do_stuff(catalog.grpc.DoStuffMessage())
```

A call looks the service's URL up when it is first made, and carries the
current trace and Request Context. It has no time limit unless it is given
one: every call takes a `timeout`, in seconds, for the whole call, and
`headers` to send:

```python
product = await catalog.openapi.get_product("xyz", timeout=2.0)
stuff = await catalog.grpc.stuff_service.do_stuff(
    catalog.grpc.DoStuffMessage(), headers={"x-tenant": "a"}
)
```

To give many calls the same settings, make the module's
`AsyncClient(base_url=None, *, timeout=None, headers=None, http_client=None)`,
for instance at the top of a module of your service. A call's `timeout` takes
the place of its client's, and `None` is no limit, as a long stream may need;
its `headers` are sent over the client's. The clients share their
connections, so one made for a single call costs nothing:

```python
catalog_api = catalog.openapi.AsyncClient(timeout=10.0)

product = await catalog_api.get_product("xyz")
report = await catalog_api.export_products(timeout=120.0)

async with catalog.storefront.AsyncClient(headers={"x-tenant": "a"}) as client:
    products = await client.products(skus=["xyz"])
```

A spec's own parameter that is named `timeout` or `headers` is renamed
`timeout2` or `headers2`.

Closing a client, as `async with` does, closes neither the shared connections
nor an `http_client` it was given, which stays yours to close. An
`httpx.AsyncClient` you give one carries the Request Context when made with
`transport=services.AsyncTransport()`. An OpenAPI or GraphQL client given a
`transport` instead, as a retrying or instrumented one, sends through it under
the SDK's, so its calls carry their Request Context, and closes the
connections it opens.

A call that fails raises `encore.services.CallError`, whatever the kind of API:
its `code` is the `ErrCode` the service answered with, or the one its HTTP
status means; `UNAVAILABLE` when no answer came, and `DEADLINE_EXCEEDED` when
the call took longer than its timeout. `message` says what went wrong, and
`status` is the HTTP status of the service's answer, 0 when there was none. A GraphQL answer with errors raises its subclass
`GraphQLRequestError`, whose `errors` are the answer's.

```python
from encore.errs import ErrCode
from encore.services import CallError

try:
    product = await catalog.openapi.get_product("xyz")
except CallError as error:
    if error.code is not ErrCode.NOT_FOUND:
        raise
    product = None
```

OpenAPI and GraphQL modules need httpx (`encore-dev[httpx]`), and gRPC modules
connectrpc (`encore-dev[grpc]`, on Python 3.11 and later). A module is imported
when first used, so a service needs only what the modules it uses need.

## Request Context

Request Context fields are typed values that travel with a request to every
service it reaches, such as the user it is made for. Declare a field once, at
module scope, with its name and the type of its value:

```python
from dataclasses import dataclass

from encore import reqctx


@dataclass
class User:
    id: str
    admin: bool = False


current_user = reqctx.Field[User]("user")
tenant = reqctx.Field[str]("tenant", trace=True)
```

Read a field while handling a request, and set or clear one for the calls made
in a block:

```python
user = current_user.get()  # None when not set

with current_user.set(User(id="system")):
    await billing.get("/invoices")

with tenant.unset():
    ...
```

Reference a field declared elsewhere, in any language, by name and with its
value's type: `current_user = reqctx.Field[User].named("user")`. Encore checks
that the types agree.

- `pubsub=True` also carries the field on the Pub/Sub messages published with
  it set, into the handlers that receive them.
- `trace=True` puts the value on the request's spans, as
  `encore.ctx.<name>`. Use it for identifiers worth finding traces by, never
  for secrets.

A request keeps every field it carries, including ones the service never
declares, and passes them on to the services it calls through the
`encore.services` clients. Fields are never sent to anything else. A field's
value may encode to at most 4 KiB (1 KiB with `pubsub`), and a request's fields
to at most 8 KiB in all: setting a value beyond that raises `ValueError`.

## Running the service

`encore.run(app)` runs the service: it serves an ASGI app (FastAPI, Starlette,
...) with Encore's server, built on hyper, Rust's HTTP library, and handles its
lifecycle.

```python
import encore
from fastapi import FastAPI

app = FastAPI()

if __name__ == "__main__":
    encore.run(app)
```

- **Startup**: it binds the address Encore assigns the service, runs the app's
  lifespan startup, starts listening, and starts Encore's Pub/Sub
  subscriptions. A failed startup exits with 1.
- **Requests**: HTTP/1.1 and HTTP/2 (h2c). Each request has its Request Context
  and a span continuing its trace. The
  server records the span itself, from the request's arrival to its response,
  named after the route the app matched; the app's code reaches it as
  `trace.get_current_span()`.
- **Shutdown**: on SIGTERM or SIGINT it stops accepting connections and taking
  Pub/Sub messages, lets the requests in flight finish, runs the app's lifespan
  shutdown, and lets Encore clean up the resources the app used, all within
  Encore's shutdown deadline. A request waiting for its client to go away, as a
  stream of events does, is told it did. A second Ctrl-C stops waiting.

A service without an HTTP API, as one that only handles Pub/Sub messages, calls
`encore.run()`: the same, without the server.

- `threads=N` runs N event loops, each on a thread of its own, sharing the app.
  It needs free-threaded Python with the GIL disabled (`python3.14t`), where
  they run in parallel. Each loop runs the app's lifespan with its own `state`,
  so resources bound to an event loop stay on theirs.
- `max_concurrency` caps the requests the app handles at once.
- `read_timeout` (30 seconds) bounds how long a client may take to send a
  request's headers, a body of up to 64 KiB (answered with 408 when it takes
  longer), or each part of a longer one (after which its request reads as its
  client gone).
- `tracing=False` takes only the Request Context.
- WebSockets are not supported yet.

A WSGI app (Flask, Django's WSGI application) runs through a WSGI-to-ASGI
adapter, such as `a2wsgi`'s, which calls it on a thread pool in the request's
context, so its code reads the Request Context and its calls carry it on:

```python
import encore
from a2wsgi import WSGIMiddleware
from flask import Flask

app = Flask(__name__)

if __name__ == "__main__":
    encore.run(WSGIMiddleware(app))
```

Its spans are not named after its routes: WSGI frameworks don't say which one
they matched.

FastAPI can record its own server, dependency, endpoint and serialization
spans when configured with a tracer provider. `run` turns them off for its
requests, preserving one server span. To leave those spans to FastAPI, pass
`server_spans=False` and configure its provider explicitly:

```python
from encore.tracing import tracer_provider

app = FastAPI(telemetry={"tracer_provider": tracer_provider()})
encore.run(app, server_spans=False)
```

`encore.run` and server middleware leave global OpenTelemetry trace, meter and
logger providers unchanged.

An app served by a server of its own adds Encore's middleware instead, for the
same Request Context and tracing:

```python
from encore.server import asgi, wsgi

app.add_middleware(asgi.Middleware)  # FastAPI, Starlette
app = wsgi.Middleware(app)  # Flask, Django's WSGI application
```

The middleware works beside OpenTelemetry's own: inside it, as
`FastAPIInstrumentor` arranges, the request's span is that one's; outside it,
OpenTelemetry's records an internal span under Encore's. The ASGI middleware
names the span after the route FastAPI or Starlette matched; WSGI frameworks
don't say which route they matched.

## Postgres queries and handles

Declare `postgres.Database("orders", migrations="./migrations")` at module scope.
`migrations` is optional and keyword-only. Use `Database.named("orders")` to
reference a database declared elsewhere.

`query(sql, *params)` returns a `Cursor`. Use `async for` to stream rows,
`await query.one()` to fetch exactly one row (raising `EncoreError` on no rows or
more than one), `await query.first()` to fetch the first row or `None`, or
`await query.all()` to collect all rows in a list. Each query can be consumed
once. Reusing it or mixing helpers with iteration raises `EncoreError`, including
after partial iteration or an error. Helpers close their cursors automatically.
For an early exit from streaming, use `async with db.query(...) as query:`.
Entering the context does not execute SQL. Exiting it closes the query.

`exec(sql, *params)` returns an `ExecResult` with `rows_affected`. Rows are
dictionaries keyed by column name. Values are bound with `$1`, `$2`, etc.

Column values have the Python types asyncpg and psycopg use. `numeric` is a
`decimal.Decimal` with the column's scale, so a `numeric(10,2)` holding 7 is
`Decimal('7.00')`; `NaN` and `±Infinity` are `Decimal('NaN')` and
`Decimal('Infinity')`. `timestamptz` is an aware `datetime.datetime` in UTC,
`timestamp` a naive one, and `date` and `time` are `datetime.date` and
`datetime.time`. `bytea` is `bytes`, arrays are lists of their elements' types,
JSON is decoded, and `uuid`, `inet` and `cidr` are strings. Parameters accept
the same types; a naive `datetime` bound to a `timestamptz` is taken as UTC,
and an aware one may use any `tzinfo`, `zoneinfo.ZoneInfo` included.

```python
async def find_order(database: postgres.Queryable, order_id: int):
    return await database.query("SELECT * FROM orders WHERE id = $1", order_id).first()


async def handle_order(order_id: int):
    return await find_order(db.read(), order_id)
```

`DatabaseRef`, `Connection`, and `Transaction` implement the nominal `Queryable`
interface. `Database` exposes the same query methods but deliberately does not
implement `Queryable`. Encore rejects passing or aliasing the declaration;
select `read()`, `write()`, or `admin()` before handing it to another function.
Direct database operations infer writer permissions. Reference creation records
reader, writer, or admin permissions at that call site.

Handles share the database's pool. A `read()` handle runs on
read-only connections, so its writes fail with SQLSTATE 25006 and its
transactions are read-only; it reads from the database's read-only pool on a
replica if the runtime config has one, unless `primary=True` is set.

```python
async def rename_order(order_id: int, name: str):
    async with db.acquire() as conn:
        async with conn.begin() as tx:
            result = await tx.exec(
                "UPDATE orders SET name = $1 WHERE id = $2", name, order_id
            )
        return result.rows_affected
```

Both `begin()` and `acquire()` can be awaited or used as async context managers.
Transaction contexts commit on a clean exit and roll back on exceptions;
connection contexts return the connection to its pool. A connection cannot be
used or closed while its transaction is active. Closed handles reject queries.

Streaming queries borrow their connection or transaction until exhausted or
closed. If you stop early, close the stream explicitly before reusing its owner:

```python
async def first_order():
    async with db.acquire() as conn:
        async with conn.query("SELECT * FROM orders ORDER BY id") as rows:
            async for row in rows:
                return row
```

## Pub/Sub payloads

Use `pubsub.Subscription[T]` to choose how payloads are decoded. Encore validates
the type argument when subscriptions start. Omitting it selects ordinary JSON
decoding. Handler annotations guide type checking but do not select the decoder.

```python
from pydantic import BaseModel


class Order(BaseModel):
    id: str


async def process_order(message: pubsub.Message[Order]) -> None:
    print(message.data.id)


processed = pubsub.Subscription[Order](events, "process-order", handler=process_order)
```

Supported payloads are JSON values, bytes, and top-level Pydantic v2 models
(when Pydantic is installed). `pubsub.JSON` and `pubsub.Payload` describe these
public types. Use `Subscription[pubsub.JSON]` for plain JSON: decoding returns ordinary
Python values without validating or constructing nested types. Use a Pydantic
model when fields need validation or nested models need construction.

`Subscription[bytes]` returns the payload unchanged. Unparameterized `Subscription`
decodes ordinary JSON. Dataclasses, model unions, models nested in plain JSON
containers, and string forward references are not supported; wrap nested models
in a top-level Pydantic model and pass that class directly. Original bytes remain
available as `message.raw_data`. Payloads are decoded once before the handler is
called; decoding errors propagate for normal retry handling.

## Gateway Middleware

A service with a `gateway` section in its `encore.service.json` hosts the app's
gateway, and its program serves it with Middleware: code that authenticates,
limits, logs or meters the public requests the gateway serves. Encore runs it on
the gateway variant of the native runtime.

```python
import asyncio
from dataclasses import dataclass

import encore
from encore import gateway, reqctx
from encore.errs import ErrCode
from encore.gateway import match


@dataclass
class User:
    id: str


current_user = reqctx.Field[User]("user")


class Authenticate(gateway.Middleware):
    match = match.path_prefix("/api")

    async def on_request(self, req: gateway.Request) -> None:
        user = await verify(req.header("authorization"))
        if user is None:
            raise gateway.reject(ErrCode.UNAUTHENTICATED, "missing token")
        req.set(current_user, user)


async def main() -> None:
    await encore.start()
    await gateway.serve(Authenticate())


asyncio.run(main())
```

A Middleware is called at the hooks it defines (`on_request`, `on_request_body`,
`on_response`, `on_response_body`, `on_complete`), only for the requests its
`match` matches: Encore checks that before a request reaches Python. A hook may
read any header, query parameter or cookie; each reaches Python only as the hook
reads it. Encore checks a subclass when it is defined, so a misspelt hook raises
at once.

`on_request` sets the request's Request Context fields with
`req.set(field, value)`: the services the request reaches read them with
`field.get()`, and no client can set them. Every hook runs in the request's
trace and context, so it reads them with `field.get()` too, or `req.get(field)`,
and the calls it makes to services carry both.

`gateway.serve` handles SIGTERM while it serves, so hooks keep running while the
gateway drains, and returns once the shutdown has completed.

## Application-owned shutdown

Without a hook, Encore shuts down non-cooperatively as the interpreter exits:
after the server has drained, it drains Pub/Sub and flushes telemetry. On
import it makes SIGTERM exit the interpreter, unless a handler is already set
or the process runs tests: with 0 once the shutdown completed in time, else 1.

Use cooperative shutdown when application tasks may still use Encore resources.
The following hook accepts a framework-specific callback that stops new
application work and waits for existing work to finish:

```python
from typing import Awaitable, Callable

from encore import shutdown


async def on_shutdown(
    stop_and_drain_application: Callable[[], Awaitable[None]],
) -> int:
    shutdown.begin(cooperative=True)

    await stop_and_drain_application()
    shutdown.mark_application_drained()

    result = await shutdown.completed()
    return result.code()
```

Connect this hook to the host's shutdown lifecycle. Keep its event loop and the
Encore runtime alive until cleanup completes. The host should enforce its own
overall shutdown deadline and decide when to terminate the process.

Call `begin()` before awaiting phase helpers. With cooperative shutdown, signal
application draining before awaiting `cleanup()` or `completed()`; otherwise
Encore waits for that signal until its deadline expires.

If application work has already drained, use non-cooperative shutdown:

```python
from encore import shutdown


async def stop_encore() -> int:
    shutdown.begin()  # cooperative=False
    result = await shutdown.completed()
    return result.code()
```

`begin()` returns immediately. All callers share the same shutdown process, and
the first call chooses the cooperative mode. Repeated calls do not restart
cleanup, change the mode, or reset the deadline. Cancelling an individual waiter
does not cancel background cleanup.

## Phases and observation

| Phase | Meaning |
| --- | --- |
| `NOT_STARTED` | Shutdown has not started. |
| `STARTED` | Pub/Sub intake has been signalled to stop. Encore is waiting for message handlers and, in cooperative mode, application draining. |
| `CLEANUP` | Encore has stopped waiting for draining and is performing resource cleanup. |
| `COMPLETED` | The cleanup task has finished, including if it failed. Check the result or exception. |

`phase()` reads the current phase synchronously. `await started()` and
`await cleanup()` wait until shutdown has reached **or passed** their respective
phases. They return immediately if the phase has already passed.

`await completed()` waits for actual cleanup completion, including after the
deadline. These waits are independent: callers do not need to await every phase
in sequence. Reading `phase()` is valid before `begin()`; waiting for a phase
before `begin()` raises `ShutdownError`.

`CLEANUP` closes SQL, Redis, and object-storage resources concurrently, then
exports final metrics and flushes tracing. Retaining a database, cache client,
bucket, or object handle does not keep its connections usable after cleanup.
New operations fail; interrupted object downloads report an error rather than
returning a partial body as success, and Redis subscriptions end.

Drain application work before signalling `mark_application_drained()`. Forced
cleanup cancels Encore-owned resource I/O even if application tasks are still
running. External clients created from connection URLs are owned by your
application. Cloud SDKs release their HTTP pools when their clients are dropped;
best-effort S3 multipart abort requests may outlive `completed()` and require a
running executor to finish.

## Deadline and forced cleanup

The core deadline defaults to five seconds, configured by
`deployment.graceful_shutdown.total` in the runtime configuration. It starts
when shutdown is first begun and covers both draining and cleanup.

When the deadline expires:

1. Encore records `Timeout` in the shared state.
2. If draining is still in progress, Encore stops waiting for it and enters
   `CLEANUP`, even without an application-drained signal.
3. Cleanup continues. `completed()` returns only after the cleanup task finishes.

The timeout is sticky: finishing later does not change it to success. It is not
a hard upper bound on how long `completed()` can wait. The controller does not
exit the process when its deadline expires.

To skip draining before the deadline, request forced cleanup:

```python
from encore import shutdown


async def force_remaining_cleanup() -> int:
    # Shutdown must already have been started with begin().
    shutdown.force_cleanup()
    await shutdown.cleanup()  # Observe the transition; cleanup is not done yet.
    result = await shutdown.completed()
    return result.code()
```

`force_cleanup()` returns immediately and is idempotent. It raises
`ShutdownError` before `begin()`. Once resource cleanup has started, forcing it
again has no effect. Forcing alone does not record `Timeout`: successful cleanup
before the deadline still returns `SUCCESS`.

Both the deadline and explicit forcing abandon the drain wait; they do not kill
active handlers. Those handlers may continue using resources while cleanup runs.
The host must account for this when choosing when to force cleanup.

## Results and errors

| Python outcome | Meaning |
| --- | --- |
| `ShutdownResult.SUCCESS` | Cleanup finished before the deadline. Explicit forcing may have skipped draining. `code()` returns `0`. |
| `ShutdownResult.TIMEOUT` | The deadline expired before cleanup finished. The returned result still means the cleanup task has now finished. `code()` returns `1`. |
| `ShutdownError` | The core reported cleanup failure or progress could not be observed, including waiting before `begin()`. |

The SDK maps native `Failed` to `ShutdownError`; it does not return the public
`FAILED` enum member from `completed()`. Native progress errors are translated
to `ShutdownError` with the original `EncoreError` retained as their cause.
An unrecognized native enum value raises `ValueError`.

If cleanup panics after the deadline has already recorded `Timeout`, the core
logs the failure and preserves `Timeout` as the reported result. A timeout is
not evidence that all cleanup operations eventually succeeded.

## Resource initialization

Importing the SDK synchronously initializes the Encore runtime. Run applications through Encore so runtime configuration is available
before imports. Initialization failures surface during import, before normal
application startup completes.

Buckets, SQL databases, pub/sub topics, and secrets store their declarations and
resolve native handles on first use, caching the results with thread-safe
initialization. These lookups are synchronous against the already initialized
runtime. Importing application modules requires the configured runtime and native
extension. Use `encore test` to run application tests with Encore-managed
resources.

Initialize independently in each worker process. Forking an already initialized
native runtime is not supported by this startup model.

## Logs

Write structured application logs with a message and fields:

```python
from encore import logs

logs.info("Order placed", order_id=42, items=3)
```

`logs.debug`, `info`, `warning`, `error`, and `critical` take a positional
message and structured keyword fields. `logs.exception` logs the active
exception at error level. Helper fields cannot replace metadata or trace
correlation: `message`, `level`, `time`, `logger`, `trace_id`, `span_id`,
`sampled`, `service.name`, `exception.type`, `exception.message`,
`exception.stacktrace`, `code.function.name`, `code.file.path`,
`code.line.number`, `code.column.number`, and `encore.runtime.target` are
reserved, as are `LogRecord` attributes, methods, and factory-added attributes. Collisions gain `raw.` prefixes until the key
conflicts with neither record attributes nor other supplied fields. For
example, `logs.info("charged", level="fatal")` keeps level `info` and writes
`raw.level="fatal"`; if `raw.level` is also supplied, the `level` field
becomes `raw.raw.level`. Keyword arguments are fields, not Python logging
options. Direct `StructuredLogHandler` usage retains its reserved-key behavior:
valid reserved-key attributes can override the corresponding output fields.

The helpers use the standard `encore.logs` logger. With no application
handler, they write Encore-formatted lines to stderr without changing the
root logger. The fallback uses INFO when root remains WARNING with no handlers;
setting root WARNING alone is indistinguishable from the default. To suppress
INFO, set `encore.logs` to WARNING or install an application handler with the
desired level. Other explicit levels, filters and disabling are respected.
Application handlers replace the fallback, without duplicate output.
To enable debug output:

```python
import logging

logging.getLogger("encore.logs").setLevel(logging.DEBUG)
```

### Python logging

Use `StructuredLogHandler` for structured JSON output from standard Python loggers:

```python
import logging

from encore.logs import StructuredLogHandler

logging.basicConfig(level=logging.INFO, handlers=[StructuredLogHandler()])
```

### OpenTelemetry

Encore collects process output and exports it through an out-of-process
collector as OpenTelemetry logs over OTLP. Encore-formatted logs preserve
structured fields, severity and current trace/span IDs, including those from
OpenTelemetry instrumentation. Other output is collected as unstructured text.
Use the helpers or Python `logging` with `StructuredLogHandler` for structured logs;
global OpenTelemetry logger providers are unchanged.

## Traces

Create spans around application work:

```python
from encore import tracing

with tracing.span("place_order", attributes={"order.id": order_id}) as span:
    span.set_attribute("order.items", 3)
    with tracing.span("validate_order"):
        ...
```

A span becomes current inside its block. Nested spans inherit its trace; spans
end on block exit, and uncaught exceptions are recorded as failures and
propagate. Use regular `with` blocks in async code too.

### OpenTelemetry

Encore spans share OpenTelemetry context. Obtain a tracer directly from Encore's
provider for standard trace API calls:

```python
from encore.tracing import tracer_provider

tracer = tracer_provider().get_tracer(__name__)

with tracer.start_as_current_span("place_order") as span:
    span.set_attribute("order.id", order_id)
```

Optional `opentelemetry-instrumentation-*` packages instrument frameworks and
libraries. Pass the provider directly, for example
`FastAPIInstrumentor().instrument_app(app, tracer_provider=tracer_provider())`,
or opt in once if your app owns global OpenTelemetry setup:

```python
from opentelemetry import trace

trace.set_tracer_provider(tracer_provider())
```

Encore always uses its cached, owned provider for SDK-created spans,
independent of the global provider. `tracing.span`, `encore.run` and server
middleware leave global trace, meter and logger providers unchanged.

## Metrics

Declare metrics at module scope: `Counter` for increasing totals,
`UpDownCounter` for totals that can decrease, `Gauge` for current values, and
`Histogram` for distributions. Each has a `…Group` form whose labels are a
`TypedDict` or a dataclass:

```python
from dataclasses import dataclass

from encore import metrics


@dataclass(frozen=True)
class Labels:
    route: str


latency = metrics.HistogramGroup[Labels]("request_duration", unit="s")

latency.with_labels(Labels("/orders")).record(0.025)
```

A service's own metric, the default scope, is exported as `<service>.<name>`:
`request_duration` of the service `shop` is `shop.request_duration`. Shared
(`scope="shared"`) and global (`scope="global"`) ones keep their name, and are
declared once: other modules import the declaration, and code in another
language binds one by name, as in
`metrics.CounterGroup[Labels].named("responses_total")`. Hand a metric to other
code as a reference, `latency.ref()`.

Metric names and scopes must be string literals. Names start with a letter,
followed by letters, digits, `_`, `.`, `-` or `/`. Label keys are snake_case;
values are strings, ints and bools, anything else a `TypeError`. `unit`,
`description` and a histogram's `buckets` must be literals: Encore's parser reads
them.

### OpenTelemetry

Encore metric declarations use the `encore` meter. Obtain a meter from Encore's
provider for standard metrics API calls:

```python
from encore.metrics import meter_provider

meter = meter_provider().get_meter("orders")
orders = meter.create_counter("orders_placed", unit="{order}")
orders.add(1, attributes={"outcome": "success"})
```

These instruments are exported alongside Encore metrics, with their names
unchanged. The meter's instrumentation scope (`orders` here) is distinct from
an Encore declaration's service/shared/global scope. The process resource
identifies the service recording them. Use the standard API for observable
instruments or attributes beyond Encore's label types.

Encore declarations always use the cached, owned meter provider, independent
of the global provider. Pass `meter_provider()` directly to instrumentation,
or opt in once if your app owns global OpenTelemetry setup:

```python
from opentelemetry import metrics

metrics.set_meter_provider(meter_provider())
```

Encore leaves global trace, meter and logger providers unchanged, including in
`encore.run` and server middleware.

## Object storage

Declare a bucket at module scope, then use its async methods:

```python
from encore.storage import objects

uploads = objects.Bucket("uploads")


async def example() -> None:
    await uploads.upload(
        "hello.txt",
        b"Hello!",
        content_type="text/plain",
        if_not_exists=True,
    )
    attrs = await uploads.attrs("hello.txt")
    present = await uploads.exists("hello.txt")
    part = await uploads.download("hello.txt", offset=0, length=5)
    async for entry in uploads.list(prefix="hello", limit=10):
        print(entry.name, entry.size)
    download_url = await uploads.signed_download_url("hello.txt", ttl=3600)
    await uploads.remove("hello.txt")
```

`attrs`, `exists`, `download`, and `remove` accept a keyword-only `version`.
`signed_upload_url` and `signed_download_url` return URL strings (`str`);
TTL is in seconds, defaults to one hour, and must be between 1 and 604800.
`public_url` is synchronous and requires a public bucket. `Bucket.named(name)`
looks up an existing bucket configured for the service.

Storage failures raise `ObjectsError` (an `EncoreError` subclass), with
`ObjectNotFound`, `PreconditionFailed`, and `InvalidArgument` for specific cases.
`exists` returns False for missing objects and raises on other storage failures.
If leaving a listing early, use `contextlib.aclosing(bucket.list(...))` to close
the native stream promptly.

For large objects, `upload_stream(name, chunks)` consumes an asynchronous iterable
of `bytes` with backpressure. It supports the same `content_type` and
`if_not_exists` options as `upload`. The native pipe is limited to 64 KiB, and
each write across the native boundary is at most 64 KiB. S3 multipart uploads retain at most four 8 MiB part payloads
and apply backpressure while waiting for the provider; completed parts retain
only their metadata. Other providers may use their own upload buffers.
The input iterable owns its chunks, so avoid producing the entire object at once.
Producer exceptions and cancellation abort the upload. The caller owns closing
an input generator that needs cleanup; cancellation cannot undo an upload the
provider has already committed.

`download_stream` yields provider-sized byte chunks and supports `version`,
`offset`, and `length`. It reads on demand. Use `aclosing` when a loop might stop
early, including when piping a download into another upload:

```python
from contextlib import aclosing


async def copy_large_object() -> None:
    async with aclosing(uploads.download_stream("source.bin")) as chunks:
        await uploads.upload_stream(
            "copy.bin", chunks, content_type="application/octet-stream"
        )
```

## Redis databases

Declare a database at module scope and call commands through its families:

```python
from encore.storage.redis import Database, ReadOnlyDatabaseRef

sessions = Database("sessions")


async def save_user():
    await sessions.string.set("user:1", b"hello", ex=60)
    return await sessions.string.get("user:1")  # bytes | None


async def load(reader: ReadOnlyDatabaseRef):
    return await reader.string.get("user:1")


async def handle_request():
    return await load(sessions.read())
```

A database is a cache unless declared `Database(name, persistent=True)`: a
cache may lose what it holds, and evicts the least recently used keys when it
runs out of memory. `Database.named(name)` references a database declared
elsewhere. Creating a database,
selecting a family, and creating a handle defer native lookup until first use.
Direct resource calls must keep the family and command together, as in
`sessions.string.get(...)`, so Encore can determine their permissions. To pass
access to another function or keep a family interface in a variable, obtain a
`read()` or `write()` handle first.

Families are the properties `generic`, `string`, `bitmap`, `geo`, `hash`, `json`,
`list`, `set`, `sorted_set`, `stream`, `hyperloglog`, and `pubsub`. Read handles expose
only read operations, and the native runtime also rejects writes through them.
`read(primary=True)` requests primary access; the current runtime uses the primary
for all reads, including when this flag is false.

A pipeline sends commands in one round trip, and a transaction (MULTI/EXEC)
runs them together. Commands are queued through the same families, each
returning a `Queued` whose `value` is the decoded result once the batch has
run; `execute()` raises `BatchError` if any failed, each `Queued` its own
error:

```python
pipe = sessions.pipeline()
name = pipe.string.get[str]("name:1")
hits = pipe.string.incr("hits")
await pipe.execute()
print(name.value, hits.value)
```

`watch(keys, body)` runs `body(conn, tx)` as an optimistic transaction,
retrying while a watched key changes; `acquire()` takes a connection out of the
pool (`async with sessions.acquire() as conn:`). A `Script` runs by its hash,
falling back to its source; only its `KEYS` get the database's prefix, so it
must take every key it touches through `key()`:

```python
RELEASE = Script(
    'if redis.call("get", KEYS[1]) == ARGV[1] then '
    'return redis.call("del", KEYS[1]) end return 0'
)
released = await sessions.script(RELEASE).key("lock:job").arg(token).invoke()
```

The blocking pops (`list.blpop`, `brpop`, `blmove`, `blmpop`,
`sorted_set.bzpopmin`, `bzpopmax`, `bzmpop`) take a keyword-only `timeout` in
seconds, 0 waiting for ever, and return `None` once it passes. They wait on a
connection of their own, so they hold up no other command; cancelling the task
ends the wait. They can't be queued, nor run on an acquired connection.
`bitmap.bitfield(key)` builds a BITFIELD (`.get()`, `.set()`, `.incrby()`,
`.overflow()`, then `await .invoke()`), and `generic.scan_iter()`,
`hash.hscan_iter`, `set.sscan_iter` and `sorted_set.zscan_iter` follow a scan's
cursor to the end.

`command(name)` sends a command the families lack, by its name, which must be
a string literal: `await sessions.command("OBJECT").arg("FREQ").key("hot").invoke()`.
Before it is sent the runtime checks it against Redis's command table, and
raises `CommandRefused` if a key it takes isn't passed with `key()`, or if it
would reach past the database, change a connection other calls share, or has a
call of its own here (MULTI, SUBSCRIBE, EVAL, SCAN, FLUSHALL, ...). A command
the server rejects raises `ServerError`, whose `code` is the first word of the
server's message (`WRONGTYPE`).

Keys and hash/stream field names are `str` inputs. Stored values, members, and
message payloads accept `Value`: `bytes`, `str`, `int`, `float`, or an implementation of
`RedisValue`. Strings are encoded as UTF-8, integers as decimal text, floats as
ASCII text using `repr()`, and bytes are sent unchanged. Booleans are not
supported as stored values. Floats include `nan` and infinities, which can be
stored and retrieved but may be rejected by Redis numeric commands.

Reads return bytes by default. Select a type with `command[str](...)`,
`command[int](...)`, `command[float](...)`, or a custom value class to decode
stored values. The return type follows the requested type:

```python
await sessions.string.set("visits", 42)
visits = await sessions.string.get[int]("visits")  # int | None

await sessions.string.set("balance", 1.25)
balance = await sessions.string.get[float]("balance")  # float | None

await sessions.list.rpush("names", "Ada", "Grace")
names = await sessions.list.lrange[str]("names", 0, -1)  # list[str]
```

`RedisValue` is a structural protocol: no base class or registration is required.
Implement `encode(self) -> bytes` and a class method `decode(cls, data: bytes)`
that returns an instance of the class:

```python
import json
from dataclasses import dataclass


@dataclass
class User:
    name: str

    def encode(self) -> bytes:
        return json.dumps({"name": self.name}).encode("utf-8")

    @classmethod
    def decode(cls, data: bytes) -> "User":
        return cls(**json.loads(data))


await sessions.string.set("user:1", User("Ada"))
user = await sessions.string.get[User]("user:1")  # User | None
await sessions.hash.hset("profiles", ("first", User("Ada")))
users = await sessions.hash.hgetall[User]("profiles")  # dict[bytes, User]
```

The same conversion applies to list elements, hash values, set/sorted-set/geo
members, stream field values, and pubsub payloads. For pubsub, use `subscribe[User](...)`
or `psubscribe[User](...)`; `Message.payload` then has that type.
Database keys, hash/stream field names, stream IDs, channel names, cursors,
scores, and other metadata retain their existing types. A custom value class
is not recorded in Redis: callers must choose the matching decoder.

Missing values remain `None`, and empty collections remain empty. Malformed
UTF-8, invalid integers, and custom codec errors raise instead of becoming
`None`. Reads that also mutate data (for example `getdel`, `lpop`, or `set` with
`get=True`) execute before decoding: a decoding error does not undo the command.
`set[User](..., get=True)` decodes the previous value; without `get=True`,
`set` only accepts the default bytes response mode.

Multi-value operations accept variadic arguments, with tuples for grouped values,
for example
`db.hash.hset("user", ("name", b"Ada"))` and
`db.string.mget("first", "second")`. Unpack existing collections with `*items`.
Most commands require at least one item; `geo.geohash`, `geo.geopos`, and
`json.get` allow none. `json.mget` requires a keyword-only `path`, and
`sorted_set.zinterstore`/`zunionstore` keep `weights` as a keyword sequence.
Commands with structured, option-dependent
replies (including streams and geo searches) retain the native RESP structure:
lists/dictionaries containing bytes, numbers, and `None`. With `withscores=True`,
sorted-set ranges return `(member: bytes, score: float)` pairs.

`string.set()` returns `None` normally; with `get=True` it returns the previous
value. With `nx=True` or `xx=True` and no `get`, it returns `b"OK"` if the write
happened and `None` otherwise.

RedisJSON methods accept Python JSON values and decode document replies:

```python
await sessions.json.set("profile", "$", {"name": "Ada", "active": True})
profiles = await sessions.json.get("profile", "$")
```

JSONPath replies preserve arrays of matches. JSON requires a server with
RedisJSON support; the local emulator does not currently implement it.

Redis channel subscriptions are separate from Encore's `pubsub.Topic` API. Use
an async context manager to release the dedicated connection on exit:

```python
async def listen():
    async with await sessions.pubsub.subscribe("events") as messages:
        async for message in messages:
            print(message.channel, message.payload)  # both bytes
```

`psubscribe()` accepts glob patterns. `Subscription.receive()` returns a `Message`
or `None` after closure; `close()` interrupts pending receives and is safe to
repeat. Leaving the context also closes the subscription when the body raises or
is cancelled.

`db.connection_string` (also available on write handles) returns the direct server
URL and grants broad read-write access. External clients using that URL do not
apply Encore's key prefixes or tracing. Read-only handles do not expose this URL.
