Files
esphome-lsm6ds/esphome/components/audio_http/media_source.py
T

60 lines
1.8 KiB
Python

from typing import Any
import esphome.codegen as cg
from esphome.components import audio, esp32, media_source, psram
import esphome.config_validation as cv
from esphome.const import CONF_BUFFER_SIZE, CONF_ID, CONF_TASK_STACK_IN_PSRAM
from esphome.types import ConfigType
CODEOWNERS = ["@kahrendt"]
AUTO_LOAD = ["audio"]
audio_http_ns = cg.esphome_ns.namespace("audio_http")
AudioHTTPMediaSource = audio_http_ns.class_(
"AudioHTTPMediaSource", cg.Component, media_source.MediaSource
)
def _request_micro_decoder(config: ConfigType) -> ConfigType:
audio.request_micro_decoder_support()
return config
def _validate_task_stack_in_psram(value: Any) -> bool:
# Only require the psram component when actually enabling PSRAM stacks; validating
# the boolean first means `false` doesn't trigger the requires_component check.
if value := cv.boolean(value):
return cv.requires_component(psram.DOMAIN)(value)
return value
CONFIG_SCHEMA = cv.All(
media_source.media_source_schema(
AudioHTTPMediaSource,
)
.extend(
{
cv.Optional(CONF_BUFFER_SIZE, default=50000): cv.int_range(
min=5000, max=1000000
),
cv.Optional(CONF_TASK_STACK_IN_PSRAM): _validate_task_stack_in_psram,
}
)
.extend(cv.COMPONENT_SCHEMA),
cv.only_on_esp32,
_request_micro_decoder,
)
async def to_code(config: ConfigType) -> None:
var = cg.new_Pvariable(config[CONF_ID])
await cg.register_component(var, config)
await media_source.register_media_source(var, config)
if config.get(CONF_TASK_STACK_IN_PSRAM):
cg.add(var.set_task_stack_in_psram(True))
esp32.add_idf_sdkconfig_option(
"CONFIG_SPIRAM_ALLOW_STACK_EXTERNAL_MEMORY", True
)
cg.add(var.set_buffer_size(config[CONF_BUFFER_SIZE]))