[api] Device defined action responses (#12136)

Co-authored-by: J. Nick Koston <nick@home-assistant.io>
Co-authored-by: J. Nick Koston <nick@koston.org>
This commit is contained in:
Jesse Hills
2025-12-07 04:47:57 +13:00
committed by GitHub
parent 75c41b11d1
commit f20aaf3981
46 changed files with 1455 additions and 105 deletions

View File

@@ -27,12 +27,13 @@ from esphome.const import (
CONF_SERVICE,
CONF_SERVICES,
CONF_TAG,
CONF_THEN,
CONF_TRIGGER_ID,
CONF_VARIABLES,
)
from esphome.core import CORE, ID, CoroPriority, coroutine_with_priority
from esphome.cpp_generator import TemplateArgsType
from esphome.types import ConfigType
from esphome.core import CORE, ID, CoroPriority, EsphomeError, coroutine_with_priority
from esphome.cpp_generator import MockObj, TemplateArgsType
from esphome.types import ConfigFragmentType, ConfigType
_LOGGER = logging.getLogger(__name__)
@@ -63,17 +64,21 @@ HomeAssistantActionResponseTrigger = api_ns.class_(
"HomeAssistantActionResponseTrigger", automation.Trigger
)
APIConnectedCondition = api_ns.class_("APIConnectedCondition", Condition)
APIRespondAction = api_ns.class_("APIRespondAction", automation.Action)
APIUnregisterServiceCallAction = api_ns.class_(
"APIUnregisterServiceCallAction", automation.Action
)
UserServiceTrigger = api_ns.class_("UserServiceTrigger", automation.Trigger)
ListEntitiesServicesArgument = api_ns.class_("ListEntitiesServicesArgument")
SERVICE_ARG_NATIVE_TYPES = {
"bool": bool,
SERVICE_ARG_NATIVE_TYPES: dict[str, MockObj] = {
"bool": cg.bool_,
"int": cg.int32,
"float": float,
"float": cg.float_,
"string": cg.std_string,
"bool[]": cg.FixedVector.template(bool).operator("const").operator("ref"),
"bool[]": cg.FixedVector.template(cg.bool_).operator("const").operator("ref"),
"int[]": cg.FixedVector.template(cg.int32).operator("const").operator("ref"),
"float[]": cg.FixedVector.template(float).operator("const").operator("ref"),
"float[]": cg.FixedVector.template(cg.float_).operator("const").operator("ref"),
"string[]": cg.FixedVector.template(cg.std_string)
.operator("const")
.operator("ref"),
@@ -102,6 +107,85 @@ def validate_encryption_key(value):
return value
CONF_SUPPORTS_RESPONSE = "supports_response"
# Enum values in api::enums namespace
enums_ns = api_ns.namespace("enums")
SUPPORTS_RESPONSE_OPTIONS = {
"none": enums_ns.SUPPORTS_RESPONSE_NONE,
"optional": enums_ns.SUPPORTS_RESPONSE_OPTIONAL,
"only": enums_ns.SUPPORTS_RESPONSE_ONLY,
"status": enums_ns.SUPPORTS_RESPONSE_STATUS,
}
def _auto_detect_supports_response(config: ConfigType) -> ConfigType:
"""Auto-detect supports_response based on api.respond usage in the action's then block.
- If api.respond with data found: set to "optional" (unless user explicitly set)
- If api.respond without data found: set to "status" (unless user explicitly set)
- If no api.respond found: set to "none" (unless user explicitly set)
"""
def scan_actions(items: ConfigFragmentType) -> tuple[bool, bool]:
"""Recursively scan actions for api.respond.
Returns: (found, has_data) tuple - has_data is True if ANY api.respond has data
"""
found_any = False
has_data_any = False
if isinstance(items, list):
for item in items:
found, has_data = scan_actions(item)
if found:
found_any = True
has_data_any = has_data_any or has_data
elif isinstance(items, dict):
# Check if this is an api.respond action
if "api.respond" in items:
respond_config = items["api.respond"]
has_data = isinstance(respond_config, dict) and "data" in respond_config
return True, has_data
# Recursively check all values
for value in items.values():
found, has_data = scan_actions(value)
if found:
found_any = True
has_data_any = has_data_any or has_data
return found_any, has_data_any
then = config.get(CONF_THEN, [])
action_name = config.get(CONF_ACTION)
found, has_data = scan_actions(then)
# If user explicitly set supports_response, validate and use that
if CONF_SUPPORTS_RESPONSE in config:
user_value = config[CONF_SUPPORTS_RESPONSE]
# Validate: "only" requires api.respond with data
if user_value == "only" and not has_data:
raise cv.Invalid(
f"Action '{action_name}' has supports_response=only but no api.respond "
"action with 'data:' was found. Use 'status' for responses without data, "
"or add 'data:' to your api.respond action."
)
return config
# Auto-detect based on api.respond usage
if found:
config[CONF_SUPPORTS_RESPONSE] = "optional" if has_data else "status"
else:
config[CONF_SUPPORTS_RESPONSE] = "none"
return config
def _validate_supports_response(value):
"""Validate supports_response after auto-detection has set the value."""
return cv.enum(SUPPORTS_RESPONSE_OPTIONS, lower=True)(value)
ACTIONS_SCHEMA = automation.validate_automation(
{
cv.GenerateID(CONF_TRIGGER_ID): cv.declare_id(UserServiceTrigger),
@@ -112,10 +196,20 @@ ACTIONS_SCHEMA = automation.validate_automation(
cv.validate_id_name: cv.one_of(*SERVICE_ARG_NATIVE_TYPES, lower=True),
}
),
# No default - auto-detected by _auto_detect_supports_response
cv.Optional(CONF_SUPPORTS_RESPONSE): cv.enum(
SUPPORTS_RESPONSE_OPTIONS, lower=True
),
},
cv.All(
cv.has_exactly_one_key(CONF_SERVICE, CONF_ACTION),
cv.rename_key(CONF_SERVICE, CONF_ACTION),
_auto_detect_supports_response,
# Re-validate supports_response after auto-detection sets it
cv.Schema(
{cv.Required(CONF_SUPPORTS_RESPONSE): _validate_supports_response},
extra=cv.ALLOW_EXTRA,
),
),
)
@@ -242,7 +336,7 @@ CONFIG_SCHEMA = cv.All(
@coroutine_with_priority(CoroPriority.WEB)
async def to_code(config):
async def to_code(config: ConfigType) -> None:
var = cg.new_Pvariable(config[CONF_ID])
await cg.register_component(var, config)
@@ -279,20 +373,61 @@ async def to_code(config):
# Collect all triggers first, then register all at once with initializer_list
triggers: list[cg.Pvariable] = []
for conf in actions:
template_args = []
func_args = []
service_arg_names = []
func_args: list[tuple[MockObj, str]] = []
service_template_args: list[MockObj] = [] # User service argument types
# Determine supports_response mode
# cv.enum returns the key with enum_value attribute containing the MockObj
supports_response_key = conf[CONF_SUPPORTS_RESPONSE]
supports_response = supports_response_key.enum_value
is_none = supports_response_key == "none"
is_optional = supports_response_key == "optional"
# Add call_id and return_response based on supports_response mode
# These must match the C++ Trigger template arguments
# - none: no extra args
# - status: call_id only (for reporting success/error without data)
# - only: call_id only (response always expected with data)
# - optional: call_id + return_response (client decides)
if not is_none:
# call_id is present for "optional", "only", and "status"
func_args.append((cg.uint32, "call_id"))
# return_response only present for "optional"
if is_optional:
func_args.append((cg.bool_, "return_response"))
service_arg_names: list[str] = []
for name, var_ in conf[CONF_VARIABLES].items():
native = SERVICE_ARG_NATIVE_TYPES[var_]
template_args.append(native)
service_template_args.append(native)
func_args.append((native, name))
service_arg_names.append(name)
templ = cg.TemplateArguments(*template_args)
# Template args: supports_response mode, then user service arg types
templ = cg.TemplateArguments(supports_response, *service_template_args)
trigger = cg.new_Pvariable(
conf[CONF_TRIGGER_ID], templ, conf[CONF_ACTION], service_arg_names
conf[CONF_TRIGGER_ID],
templ,
conf[CONF_ACTION],
service_arg_names,
)
triggers.append(trigger)
await automation.build_automation(trigger, func_args, conf)
auto = await automation.build_automation(trigger, func_args, conf)
# For non-none response modes, automatically append unregister action
# This ensures the call is unregistered after all actions complete (including async ones)
if not is_none:
arg_types = [arg[0] for arg in func_args]
action_templ = cg.TemplateArguments(*arg_types)
unregister_id = ID(
f"{conf[CONF_TRIGGER_ID]}__unregister",
is_declaration=True,
type=APIUnregisterServiceCallAction.template(action_templ),
)
unregister_action = cg.new_Pvariable(
unregister_id,
var,
)
cg.add(auto.add_actions([unregister_action]))
# Register all services at once - single allocation, no reallocations
cg.add(var.initialize_user_services(triggers))
@@ -538,6 +673,80 @@ async def homeassistant_tag_scanned_to_code(config, action_id, template_arg, arg
return var
CONF_SUCCESS = "success"
CONF_ERROR_MESSAGE = "error_message"
def _validate_api_respond_data(config):
"""Set flag during validation so AUTO_LOAD can include json component."""
if CONF_DATA in config:
CORE.data.setdefault(DOMAIN, {})[CONF_CAPTURE_RESPONSE] = True
return config
API_RESPOND_ACTION_SCHEMA = cv.All(
cv.Schema(
{
cv.GenerateID(): cv.use_id(APIServer),
cv.Optional(CONF_SUCCESS, default=True): cv.templatable(cv.boolean),
cv.Optional(CONF_ERROR_MESSAGE, default=""): cv.templatable(cv.string),
cv.Optional(CONF_DATA): cv.lambda_,
}
),
_validate_api_respond_data,
)
@automation.register_action(
"api.respond",
APIRespondAction,
API_RESPOND_ACTION_SCHEMA,
)
async def api_respond_to_code(
config: ConfigType,
action_id: ID,
template_arg: cg.TemplateArguments,
args: TemplateArgsType,
) -> MockObj:
# Validate that api.respond is used inside an API action context.
# We can't easily validate this at config time since the schema validation
# doesn't have access to the parent action context. Validating here in to_code
# is still much better than a cryptic C++ compile error.
has_call_id = any(name == "call_id" for _, name in args)
if not has_call_id:
raise EsphomeError(
"api.respond can only be used inside an API action's 'then:' block. "
"The 'call_id' variable is required to send a response."
)
cg.add_define("USE_API_USER_DEFINED_ACTION_RESPONSES")
serv = await cg.get_variable(config[CONF_ID])
var = cg.new_Pvariable(action_id, template_arg, serv)
# Check if we're in optional mode (has return_response arg)
is_optional = any(name == "return_response" for _, name in args)
if is_optional:
cg.add(var.set_is_optional_mode(True))
templ = await cg.templatable(config[CONF_SUCCESS], args, cg.bool_)
cg.add(var.set_success(templ))
templ = await cg.templatable(config[CONF_ERROR_MESSAGE], args, cg.std_string)
cg.add(var.set_error_message(templ))
if CONF_DATA in config:
cg.add_define("USE_API_USER_DEFINED_ACTION_RESPONSES_JSON")
# Lambda populates the JsonObject root - no return value needed
lambda_ = await cg.process_lambda(
config[CONF_DATA],
args + [(cg.JsonObject, "root")],
return_type=cg.void,
)
cg.add(var.set_data(lambda_))
return var
API_CONNECTED_CONDITION_SCHEMA = cv.Schema(
{
cv.GenerateID(): cv.use_id(APIServer),

View File

@@ -855,6 +855,14 @@ enum ServiceArgType {
SERVICE_ARG_TYPE_FLOAT_ARRAY = 6;
SERVICE_ARG_TYPE_STRING_ARRAY = 7;
}
enum SupportsResponseType {
SUPPORTS_RESPONSE_NONE = 0;
SUPPORTS_RESPONSE_OPTIONAL = 1;
SUPPORTS_RESPONSE_ONLY = 2;
// Status-only response - reports success/error without data payload
// Value is higher to avoid conflicts with future Home Assistant values
SUPPORTS_RESPONSE_STATUS = 100;
}
message ListEntitiesServicesArgument {
option (ifdef) = "USE_API_USER_DEFINED_ACTIONS";
string name = 1;
@@ -868,6 +876,7 @@ message ListEntitiesServicesResponse {
string name = 1;
fixed32 key = 2;
repeated ListEntitiesServicesArgument args = 3 [(fixed_vector) = true];
SupportsResponseType supports_response = 4;
}
message ExecuteServiceArgument {
option (ifdef) = "USE_API_USER_DEFINED_ACTIONS";
@@ -890,6 +899,21 @@ message ExecuteServiceRequest {
fixed32 key = 1;
repeated ExecuteServiceArgument args = 2 [(fixed_vector) = true];
uint32 call_id = 3 [(field_ifdef) = "USE_API_USER_DEFINED_ACTION_RESPONSES"];
bool return_response = 4 [(field_ifdef) = "USE_API_USER_DEFINED_ACTION_RESPONSES"];
}
// Message sent by ESPHome to Home Assistant with service execution response data
message ExecuteServiceResponse {
option (id) = 131;
option (source) = SOURCE_SERVER;
option (no_delay) = true;
option (ifdef) = "USE_API_USER_DEFINED_ACTION_RESPONSES";
uint32 call_id = 1; // Matches the call_id from ExecuteServiceRequest
bool success = 2; // Whether the service execution succeeded
string error_message = 3; // Error message if success = false
bytes response_data = 4 [(pointer_to_buffer) = true, (field_ifdef) = "USE_API_USER_DEFINED_ACTION_RESPONSES_JSON"];
}
// ==================== CAMERA ====================

View File

@@ -6,6 +6,9 @@
#ifdef USE_API_PLAINTEXT
#include "api_frame_helper_plaintext.h"
#endif
#ifdef USE_API_USER_DEFINED_ACTIONS
#include "user_services.h"
#endif
#include <cerrno>
#include <cinttypes>
#include <functional>
@@ -1554,15 +1557,54 @@ void APIConnection::on_home_assistant_state_response(const HomeAssistantStateRes
#ifdef USE_API_USER_DEFINED_ACTIONS
void APIConnection::execute_service(const ExecuteServiceRequest &msg) {
bool found = false;
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
// Register the call and get a unique server-generated action_call_id
// This avoids collisions when multiple clients use the same call_id
uint32_t action_call_id = 0;
if (msg.call_id != 0) {
action_call_id = this->parent_->register_active_action_call(msg.call_id, this);
}
// Use the overload that passes action_call_id separately (avoids copying msg)
for (auto *service : this->parent_->get_user_services()) {
if (service->execute_service(msg, action_call_id)) {
found = true;
}
}
#else
for (auto *service : this->parent_->get_user_services()) {
if (service->execute_service(msg)) {
found = true;
}
}
#endif
if (!found) {
ESP_LOGV(TAG, "Could not find service");
}
// Note: For services with supports_response != none, the call is unregistered
// by an automatically appended APIUnregisterServiceCallAction at the end of
// the action list. This ensures async actions (delays, waits) complete first.
}
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
void APIConnection::send_execute_service_response(uint32_t call_id, bool success, const std::string &error_message) {
ExecuteServiceResponse resp;
resp.call_id = call_id;
resp.success = success;
resp.set_error_message(StringRef(error_message));
this->send_message(resp, ExecuteServiceResponse::MESSAGE_TYPE);
}
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
void APIConnection::send_execute_service_response(uint32_t call_id, bool success, const std::string &error_message,
const uint8_t *response_data, size_t response_data_len) {
ExecuteServiceResponse resp;
resp.call_id = call_id;
resp.success = success;
resp.set_error_message(StringRef(error_message));
resp.response_data = response_data;
resp.response_data_len = response_data_len;
this->send_message(resp, ExecuteServiceResponse::MESSAGE_TYPE);
}
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES
#endif
#ifdef USE_API_HOMEASSISTANT_ACTION_RESPONSES

View File

@@ -223,6 +223,13 @@ class APIConnection final : public APIServerConnection {
#endif
#ifdef USE_API_USER_DEFINED_ACTIONS
void execute_service(const ExecuteServiceRequest &msg) override;
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
void send_execute_service_response(uint32_t call_id, bool success, const std::string &error_message);
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
void send_execute_service_response(uint32_t call_id, bool success, const std::string &error_message,
const uint8_t *response_data, size_t response_data_len);
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES
#endif
#ifdef USE_API_NOISE
bool send_noise_encryption_set_key_response(const NoiseEncryptionSetKeyRequest &msg) override;

View File

@@ -1010,11 +1010,13 @@ void ListEntitiesServicesResponse::encode(ProtoWriteBuffer buffer) const {
for (auto &it : this->args) {
buffer.encode_message(3, it, true);
}
buffer.encode_uint32(4, static_cast<uint32_t>(this->supports_response));
}
void ListEntitiesServicesResponse::calculate_size(ProtoSize &size) const {
size.add_length(1, this->name_ref_.size());
size.add_fixed32(1, this->key);
size.add_repeated_message(1, this->args);
size.add_uint32(1, static_cast<uint32_t>(this->supports_response));
}
bool ExecuteServiceArgument::decode_varint(uint32_t field_id, ProtoVarInt value) {
switch (field_id) {
@@ -1075,6 +1077,23 @@ void ExecuteServiceArgument::decode(const uint8_t *buffer, size_t length) {
this->string_array.init(count_string_array);
ProtoDecodableMessage::decode(buffer, length);
}
bool ExecuteServiceRequest::decode_varint(uint32_t field_id, ProtoVarInt value) {
switch (field_id) {
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
case 3:
this->call_id = value.as_uint32();
break;
#endif
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
case 4:
this->return_response = value.as_bool();
break;
#endif
default:
return false;
}
return true;
}
bool ExecuteServiceRequest::decode_length(uint32_t field_id, ProtoLengthDelimited value) {
switch (field_id) {
case 2:
@@ -1102,6 +1121,24 @@ void ExecuteServiceRequest::decode(const uint8_t *buffer, size_t length) {
ProtoDecodableMessage::decode(buffer, length);
}
#endif
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
void ExecuteServiceResponse::encode(ProtoWriteBuffer buffer) const {
buffer.encode_uint32(1, this->call_id);
buffer.encode_bool(2, this->success);
buffer.encode_string(3, this->error_message_ref_);
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
buffer.encode_bytes(4, this->response_data, this->response_data_len);
#endif
}
void ExecuteServiceResponse::calculate_size(ProtoSize &size) const {
size.add_uint32(1, this->call_id);
size.add_bool(1, this->success);
size.add_length(1, this->error_message_ref_.size());
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
size.add_length(4, this->response_data_len);
#endif
}
#endif
#ifdef USE_CAMERA
void ListEntitiesCameraResponse::encode(ProtoWriteBuffer buffer) const {
buffer.encode_string(1, this->object_id_ref_);

View File

@@ -75,6 +75,12 @@ enum ServiceArgType : uint32_t {
SERVICE_ARG_TYPE_FLOAT_ARRAY = 6,
SERVICE_ARG_TYPE_STRING_ARRAY = 7,
};
enum SupportsResponseType : uint32_t {
SUPPORTS_RESPONSE_NONE = 0,
SUPPORTS_RESPONSE_OPTIONAL = 1,
SUPPORTS_RESPONSE_ONLY = 2,
SUPPORTS_RESPONSE_STATUS = 100,
};
#endif
#ifdef USE_CLIMATE
enum ClimateMode : uint32_t {
@@ -1257,7 +1263,7 @@ class ListEntitiesServicesArgument final : public ProtoMessage {
class ListEntitiesServicesResponse final : public ProtoMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 41;
static constexpr uint8_t ESTIMATED_SIZE = 48;
static constexpr uint8_t ESTIMATED_SIZE = 50;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "list_entities_services_response"; }
#endif
@@ -1265,6 +1271,7 @@ class ListEntitiesServicesResponse final : public ProtoMessage {
void set_name(const StringRef &ref) { this->name_ref_ = ref; }
uint32_t key{0};
FixedVector<ListEntitiesServicesArgument> args{};
enums::SupportsResponseType supports_response{};
void encode(ProtoWriteBuffer buffer) const override;
void calculate_size(ProtoSize &size) const override;
#ifdef HAS_PROTO_MESSAGE_DUMP
@@ -1297,12 +1304,18 @@ class ExecuteServiceArgument final : public ProtoDecodableMessage {
class ExecuteServiceRequest final : public ProtoDecodableMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 42;
static constexpr uint8_t ESTIMATED_SIZE = 39;
static constexpr uint8_t ESTIMATED_SIZE = 45;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "execute_service_request"; }
#endif
uint32_t key{0};
FixedVector<ExecuteServiceArgument> args{};
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
uint32_t call_id{0};
#endif
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
bool return_response{false};
#endif
void decode(const uint8_t *buffer, size_t length) override;
#ifdef HAS_PROTO_MESSAGE_DUMP
void dump_to(std::string &out) const override;
@@ -1311,6 +1324,32 @@ class ExecuteServiceRequest final : public ProtoDecodableMessage {
protected:
bool decode_32bit(uint32_t field_id, Proto32Bit value) override;
bool decode_length(uint32_t field_id, ProtoLengthDelimited value) override;
bool decode_varint(uint32_t field_id, ProtoVarInt value) override;
};
#endif
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
class ExecuteServiceResponse final : public ProtoMessage {
public:
static constexpr uint8_t MESSAGE_TYPE = 131;
static constexpr uint8_t ESTIMATED_SIZE = 34;
#ifdef HAS_PROTO_MESSAGE_DUMP
const char *message_name() const override { return "execute_service_response"; }
#endif
uint32_t call_id{0};
bool success{false};
StringRef error_message_ref_{};
void set_error_message(const StringRef &ref) { this->error_message_ref_ = ref; }
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
const uint8_t *response_data{nullptr};
uint16_t response_data_len{0};
#endif
void encode(ProtoWriteBuffer buffer) const override;
void calculate_size(ProtoSize &size) const override;
#ifdef HAS_PROTO_MESSAGE_DUMP
void dump_to(std::string &out) const override;
#endif
protected:
};
#endif
#ifdef USE_CAMERA

View File

@@ -231,6 +231,20 @@ template<> const char *proto_enum_to_string<enums::ServiceArgType>(enums::Servic
return "UNKNOWN";
}
}
template<> const char *proto_enum_to_string<enums::SupportsResponseType>(enums::SupportsResponseType value) {
switch (value) {
case enums::SUPPORTS_RESPONSE_NONE:
return "SUPPORTS_RESPONSE_NONE";
case enums::SUPPORTS_RESPONSE_OPTIONAL:
return "SUPPORTS_RESPONSE_OPTIONAL";
case enums::SUPPORTS_RESPONSE_ONLY:
return "SUPPORTS_RESPONSE_ONLY";
case enums::SUPPORTS_RESPONSE_STATUS:
return "SUPPORTS_RESPONSE_STATUS";
default:
return "UNKNOWN";
}
}
#endif
#ifdef USE_CLIMATE
template<> const char *proto_enum_to_string<enums::ClimateMode>(enums::ClimateMode value) {
@@ -1194,6 +1208,7 @@ void ListEntitiesServicesResponse::dump_to(std::string &out) const {
it.dump_to(out);
out.append("\n");
}
dump_field(out, "supports_response", static_cast<enums::SupportsResponseType>(this->supports_response));
}
void ExecuteServiceArgument::dump_to(std::string &out) const {
MessageDumpHelper helper(out, "ExecuteServiceArgument");
@@ -1223,6 +1238,25 @@ void ExecuteServiceRequest::dump_to(std::string &out) const {
it.dump_to(out);
out.append("\n");
}
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
dump_field(out, "call_id", this->call_id);
#endif
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
dump_field(out, "return_response", this->return_response);
#endif
}
#endif
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
void ExecuteServiceResponse::dump_to(std::string &out) const {
MessageDumpHelper helper(out, "ExecuteServiceResponse");
dump_field(out, "call_id", this->call_id);
dump_field(out, "success", this->success);
dump_field(out, "error_message", this->error_message_ref_);
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
out.append(" response_data: ");
out.append(format_hex_pretty(this->response_data, this->response_data_len));
out.append("\n");
#endif
}
#endif
#ifdef USE_CAMERA

View File

@@ -4,8 +4,8 @@
#include "api_connection.h"
#include "esphome/components/network/util.h"
#include "esphome/core/application.h"
#include "esphome/core/defines.h"
#include "esphome/core/controller_registry.h"
#include "esphome/core/defines.h"
#include "esphome/core/hal.h"
#include "esphome/core/log.h"
#include "esphome/core/util.h"
@@ -186,6 +186,9 @@ void APIServer::loop() {
// Rare case: handle disconnection
#ifdef USE_API_CLIENT_DISCONNECTED_TRIGGER
this->client_disconnected_trigger_->trigger(client->client_info_.name, client->client_info_.peername);
#endif
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
this->unregister_active_action_calls_for_connection(client.get());
#endif
ESP_LOGV(TAG, "Remove connection %s", client->client_info_.name.c_str());
@@ -585,5 +588,84 @@ bool APIServer::teardown() {
return this->clients_.empty();
}
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
// Timeout for action calls - matches aioesphomeapi client timeout (default 30s)
// Can be overridden via USE_API_ACTION_CALL_TIMEOUT_MS define for testing
#ifndef USE_API_ACTION_CALL_TIMEOUT_MS
#define USE_API_ACTION_CALL_TIMEOUT_MS 30000 // NOLINT
#endif
uint32_t APIServer::register_active_action_call(uint32_t client_call_id, APIConnection *conn) {
uint32_t action_call_id = this->next_action_call_id_++;
// Handle wraparound (skip 0 as it means "no call")
if (this->next_action_call_id_ == 0) {
this->next_action_call_id_ = 1;
}
this->active_action_calls_.push_back({action_call_id, client_call_id, conn});
// Schedule automatic cleanup after timeout (client will have given up by then)
this->set_timeout(str_sprintf("action_call_%u", action_call_id), USE_API_ACTION_CALL_TIMEOUT_MS,
[this, action_call_id]() {
ESP_LOGD(TAG, "Action call %u timed out", action_call_id);
this->unregister_active_action_call(action_call_id);
});
return action_call_id;
}
void APIServer::unregister_active_action_call(uint32_t action_call_id) {
// Cancel the timeout for this action call
this->cancel_timeout(str_sprintf("action_call_%u", action_call_id));
// Swap-and-pop is more efficient than remove_if for unordered vectors
for (size_t i = 0; i < this->active_action_calls_.size(); i++) {
if (this->active_action_calls_[i].action_call_id == action_call_id) {
std::swap(this->active_action_calls_[i], this->active_action_calls_.back());
this->active_action_calls_.pop_back();
return;
}
}
}
void APIServer::unregister_active_action_calls_for_connection(APIConnection *conn) {
// Remove all active action calls for disconnected connection using swap-and-pop
for (size_t i = 0; i < this->active_action_calls_.size();) {
if (this->active_action_calls_[i].connection == conn) {
// Cancel the timeout for this action call
this->cancel_timeout(str_sprintf("action_call_%u", this->active_action_calls_[i].action_call_id));
std::swap(this->active_action_calls_[i], this->active_action_calls_.back());
this->active_action_calls_.pop_back();
// Don't increment i - need to check the swapped element
} else {
i++;
}
}
}
void APIServer::send_action_response(uint32_t action_call_id, bool success, const std::string &error_message) {
for (auto &call : this->active_action_calls_) {
if (call.action_call_id == action_call_id) {
call.connection->send_execute_service_response(call.client_call_id, success, error_message);
return;
}
}
ESP_LOGW(TAG, "Cannot send response: no active call found for action_call_id %u", action_call_id);
}
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
void APIServer::send_action_response(uint32_t action_call_id, bool success, const std::string &error_message,
const uint8_t *response_data, size_t response_data_len) {
for (auto &call : this->active_action_calls_) {
if (call.action_call_id == action_call_id) {
call.connection->send_execute_service_response(call.client_call_id, success, error_message, response_data,
response_data_len);
return;
}
}
ESP_LOGW(TAG, "Cannot send response: no active call found for action_call_id %u", action_call_id);
}
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES
} // namespace esphome::api
#endif

View File

@@ -12,9 +12,6 @@
#include "esphome/core/log.h"
#include "list_entities.h"
#include "subscribe_state.h"
#ifdef USE_API_USER_DEFINED_ACTIONS
#include "user_services.h"
#endif
#ifdef USE_LOGGER
#include "esphome/components/logger/logger.h"
#endif
@@ -22,11 +19,15 @@
#include "esphome/components/camera/camera.h"
#endif
#include <map>
#include <vector>
namespace esphome::api {
#ifdef USE_API_USER_DEFINED_ACTIONS
// Forward declaration - full definition in user_services.h
class UserServiceDescriptor;
#endif
#ifdef USE_API_NOISE
struct SavedNoisePsk {
psk_t psk;
@@ -154,6 +155,19 @@ class APIServer : public Component,
// Only compile push_back method when custom_services: true (external components)
void register_user_service(UserServiceDescriptor *descriptor) { this->user_services_.push_back(descriptor); }
#endif
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
// Action call context management - supports concurrent calls from multiple clients
// Returns server-generated action_call_id to avoid collisions when clients use same call_id
uint32_t register_active_action_call(uint32_t client_call_id, APIConnection *conn);
void unregister_active_action_call(uint32_t action_call_id);
void unregister_active_action_calls_for_connection(APIConnection *conn);
// Send response for a specific action call (uses action_call_id, sends client_call_id in response)
void send_action_response(uint32_t action_call_id, bool success, const std::string &error_message);
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
void send_action_response(uint32_t action_call_id, bool success, const std::string &error_message,
const uint8_t *response_data, size_t response_data_len);
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES
#endif
#ifdef USE_HOMEASSISTANT_TIME
void request_time();
@@ -230,6 +244,17 @@ class APIServer : public Component,
#endif
#ifdef USE_API_USER_DEFINED_ACTIONS
std::vector<UserServiceDescriptor *> user_services_;
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
// Active action calls - supports concurrent calls from multiple clients
// Uses server-generated action_call_id to avoid collisions when multiple clients use same call_id
struct ActiveActionCall {
uint32_t action_call_id; // Server-generated unique ID (passed to actions)
uint32_t client_call_id; // Client's original call_id (used in response)
APIConnection *connection;
};
std::vector<ActiveActionCall> active_action_calls_;
uint32_t next_action_call_id_{1}; // Counter for generating unique action_call_ids
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES
#endif
#ifdef USE_API_HOMEASSISTANT_ACTION_RESPONSES
struct PendingActionResponse {

View File

@@ -16,7 +16,10 @@ template<typename T, typename... Ts> class CustomAPIDeviceService : public UserS
: UserServiceDynamic<Ts...>(name, arg_names), obj_(obj), callback_(callback) {}
protected:
void execute(Ts... x) override { (this->obj_->*this->callback_)(x...); } // NOLINT
// CustomAPIDevice services don't support action responses - ignore call_id and return_response
void execute(uint32_t /*call_id*/, bool /*return_response*/, Ts... x) override {
(this->obj_->*this->callback_)(x...); // NOLINT
}
T *obj_;
void (T::*callback_)(Ts...);

View File

@@ -5,6 +5,9 @@
#include "esphome/core/application.h"
#include "esphome/core/log.h"
#include "esphome/core/util.h"
#ifdef USE_API_USER_DEFINED_ACTIONS
#include "user_services.h"
#endif
namespace esphome::api {

View File

@@ -1,20 +1,31 @@
#pragma once
#include <tuple>
#include <utility>
#include <vector>
#include "esphome/core/component.h"
#include "esphome/core/automation.h"
#include "api_pb2.h"
#include "esphome/core/automation.h"
#include "esphome/core/component.h"
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
#include "esphome/components/json/json_util.h"
#endif
#ifdef USE_API_USER_DEFINED_ACTIONS
namespace esphome::api {
// Forward declaration - full definition in api_server.h
class APIServer;
class UserServiceDescriptor {
public:
virtual ListEntitiesServicesResponse encode_list_service_response() = 0;
virtual bool execute_service(const ExecuteServiceRequest &req) = 0;
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
// Overload that accepts server-generated action_call_id (avoids client call_id collisions)
virtual bool execute_service(const ExecuteServiceRequest &req, uint32_t action_call_id) = 0;
#endif
bool is_internal() { return false; }
};
@@ -27,8 +38,9 @@ template<typename T> enums::ServiceArgType to_service_arg_type();
// Stores only pointers to string literals in flash - no heap allocation
template<typename... Ts> class UserServiceBase : public UserServiceDescriptor {
public:
UserServiceBase(const char *name, const std::array<const char *, sizeof...(Ts)> &arg_names)
: name_(name), arg_names_(arg_names) {
UserServiceBase(const char *name, const std::array<const char *, sizeof...(Ts)> &arg_names,
enums::SupportsResponseType supports_response = enums::SUPPORTS_RESPONSE_NONE)
: name_(name), arg_names_(arg_names), supports_response_(supports_response) {
this->key_ = fnv1_hash(name);
}
@@ -36,6 +48,7 @@ template<typename... Ts> class UserServiceBase : public UserServiceDescriptor {
ListEntitiesServicesResponse msg;
msg.set_name(StringRef(this->name_));
msg.key = this->key_;
msg.supports_response = this->supports_response_;
std::array<enums::ServiceArgType, sizeof...(Ts)> arg_types = {to_service_arg_type<Ts>()...};
msg.args.init(sizeof...(Ts));
for (size_t i = 0; i < sizeof...(Ts); i++) {
@@ -51,21 +64,37 @@ template<typename... Ts> class UserServiceBase : public UserServiceDescriptor {
return false;
if (req.args.size() != sizeof...(Ts))
return false;
this->execute_(req.args, std::make_index_sequence<sizeof...(Ts)>{});
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
this->execute_(req.args, req.call_id, req.return_response, std::make_index_sequence<sizeof...(Ts)>{});
#else
this->execute_(req.args, 0, false, std::make_index_sequence<sizeof...(Ts)>{});
#endif
return true;
}
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
bool execute_service(const ExecuteServiceRequest &req, uint32_t action_call_id) override {
if (req.key != this->key_)
return false;
if (req.args.size() != sizeof...(Ts))
return false;
this->execute_(req.args, action_call_id, req.return_response, std::make_index_sequence<sizeof...(Ts)>{});
return true;
}
#endif
protected:
virtual void execute(Ts... x) = 0;
virtual void execute(uint32_t call_id, bool return_response, Ts... x) = 0;
template<typename ArgsContainer, size_t... S>
void execute_(const ArgsContainer &args, std::index_sequence<S...> type) {
this->execute((get_execute_arg_value<Ts>(args[S]))...);
void execute_(const ArgsContainer &args, uint32_t call_id, bool return_response, std::index_sequence<S...> /*type*/) {
this->execute(call_id, return_response, (get_execute_arg_value<Ts>(args[S]))...);
}
// Pointers to string literals in flash - no heap allocation
const char *name_;
std::array<const char *, sizeof...(Ts)> arg_names_;
uint32_t key_{0};
enums::SupportsResponseType supports_response_{enums::SUPPORTS_RESPONSE_NONE};
};
// Separate class for custom_api_device services (rare case)
@@ -81,6 +110,7 @@ template<typename... Ts> class UserServiceDynamic : public UserServiceDescriptor
ListEntitiesServicesResponse msg;
msg.set_name(StringRef(this->name_));
msg.key = this->key_;
msg.supports_response = enums::SUPPORTS_RESPONSE_NONE; // Dynamic services don't support responses yet
std::array<enums::ServiceArgType, sizeof...(Ts)> arg_types = {to_service_arg_type<Ts>()...};
msg.args.init(sizeof...(Ts));
for (size_t i = 0; i < sizeof...(Ts); i++) {
@@ -96,15 +126,31 @@ template<typename... Ts> class UserServiceDynamic : public UserServiceDescriptor
return false;
if (req.args.size() != sizeof...(Ts))
return false;
this->execute_(req.args, std::make_index_sequence<sizeof...(Ts)>{});
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
this->execute_(req.args, req.call_id, req.return_response, std::make_index_sequence<sizeof...(Ts)>{});
#else
this->execute_(req.args, 0, false, std::make_index_sequence<sizeof...(Ts)>{});
#endif
return true;
}
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
// Dynamic services don't support responses yet, but need to implement the interface
bool execute_service(const ExecuteServiceRequest &req, uint32_t action_call_id) override {
if (req.key != this->key_)
return false;
if (req.args.size() != sizeof...(Ts))
return false;
this->execute_(req.args, action_call_id, req.return_response, std::make_index_sequence<sizeof...(Ts)>{});
return true;
}
#endif
protected:
virtual void execute(Ts... x) = 0;
virtual void execute(uint32_t call_id, bool return_response, Ts... x) = 0;
template<typename ArgsContainer, size_t... S>
void execute_(const ArgsContainer &args, std::index_sequence<S...> type) {
this->execute((get_execute_arg_value<Ts>(args[S]))...);
void execute_(const ArgsContainer &args, uint32_t call_id, bool return_response, std::index_sequence<S...> /*type*/) {
this->execute(call_id, return_response, (get_execute_arg_value<Ts>(args[S]))...);
}
// Heap-allocated strings for runtime-generated names
@@ -113,15 +159,149 @@ template<typename... Ts> class UserServiceDynamic : public UserServiceDescriptor
uint32_t key_{0};
};
template<typename... Ts> class UserServiceTrigger : public UserServiceBase<Ts...>, public Trigger<Ts...> {
// Primary template declaration
template<enums::SupportsResponseType Mode, typename... Ts> class UserServiceTrigger;
// Specialization for NONE - no extra trigger arguments
template<typename... Ts>
class UserServiceTrigger<enums::SUPPORTS_RESPONSE_NONE, Ts...> : public UserServiceBase<Ts...>, public Trigger<Ts...> {
public:
// Constructor for static names (YAML-defined services - used by code generator)
UserServiceTrigger(const char *name, const std::array<const char *, sizeof...(Ts)> &arg_names)
: UserServiceBase<Ts...>(name, arg_names) {}
: UserServiceBase<Ts...>(name, arg_names, enums::SUPPORTS_RESPONSE_NONE) {}
protected:
void execute(Ts... x) override { this->trigger(x...); } // NOLINT
void execute(uint32_t /*call_id*/, bool /*return_response*/, Ts... x) override { this->trigger(x...); }
};
// Specialization for OPTIONAL - call_id and return_response trigger arguments
template<typename... Ts>
class UserServiceTrigger<enums::SUPPORTS_RESPONSE_OPTIONAL, Ts...> : public UserServiceBase<Ts...>,
public Trigger<uint32_t, bool, Ts...> {
public:
UserServiceTrigger(const char *name, const std::array<const char *, sizeof...(Ts)> &arg_names)
: UserServiceBase<Ts...>(name, arg_names, enums::SUPPORTS_RESPONSE_OPTIONAL) {}
protected:
void execute(uint32_t call_id, bool return_response, Ts... x) override {
this->trigger(call_id, return_response, x...);
}
};
// Specialization for ONLY - just call_id trigger argument
template<typename... Ts>
class UserServiceTrigger<enums::SUPPORTS_RESPONSE_ONLY, Ts...> : public UserServiceBase<Ts...>,
public Trigger<uint32_t, Ts...> {
public:
UserServiceTrigger(const char *name, const std::array<const char *, sizeof...(Ts)> &arg_names)
: UserServiceBase<Ts...>(name, arg_names, enums::SUPPORTS_RESPONSE_ONLY) {}
protected:
void execute(uint32_t call_id, bool /*return_response*/, Ts... x) override { this->trigger(call_id, x...); }
};
// Specialization for STATUS - just call_id trigger argument (reports success/error without data)
template<typename... Ts>
class UserServiceTrigger<enums::SUPPORTS_RESPONSE_STATUS, Ts...> : public UserServiceBase<Ts...>,
public Trigger<uint32_t, Ts...> {
public:
UserServiceTrigger(const char *name, const std::array<const char *, sizeof...(Ts)> &arg_names)
: UserServiceBase<Ts...>(name, arg_names, enums::SUPPORTS_RESPONSE_STATUS) {}
protected:
void execute(uint32_t call_id, bool /*return_response*/, Ts... x) override { this->trigger(call_id, x...); }
};
} // namespace esphome::api
#endif // USE_API_USER_DEFINED_ACTIONS
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES
// Include full definition of APIServer for template implementation
// Must be outside namespace to avoid including STL headers inside namespace
#include "api_server.h"
namespace esphome::api {
template<typename... Ts> class APIRespondAction : public Action<Ts...> {
public:
explicit APIRespondAction(APIServer *parent) : parent_(parent) {}
template<typename V> void set_success(V success) { this->success_ = success; }
template<typename V> void set_error_message(V error) { this->error_message_ = error; }
void set_is_optional_mode(bool is_optional) { this->is_optional_mode_ = is_optional; }
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
void set_data(std::function<void(Ts..., JsonObject)> func) {
this->json_builder_ = std::move(func);
this->has_data_ = true;
}
#endif
void play(const Ts &...x) override {
// Extract call_id from first argument - it's always first for optional/only/status modes
auto args = std::make_tuple(x...);
uint32_t call_id = std::get<0>(args);
bool success = this->success_.value(x...);
std::string error_message = this->error_message_.value(x...);
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
if (this->has_data_) {
// For optional mode, check return_response (second arg) to decide if client wants data
// Use nested if constexpr to avoid compile error when tuple doesn't have enough elements
// (std::tuple_element_t is evaluated before the && short-circuit, so we must nest)
if constexpr (sizeof...(Ts) >= 2) {
if constexpr (std::is_same_v<std::tuple_element_t<1, std::tuple<Ts...>>, bool>) {
if (this->is_optional_mode_) {
bool return_response = std::get<1>(args);
if (!return_response) {
// Client doesn't want response data, just send success/error
this->parent_->send_action_response(call_id, success, error_message);
return;
}
}
}
}
// Build and send JSON response
json::JsonBuilder builder;
this->json_builder_(x..., builder.root());
std::string json_str = builder.serialize();
this->parent_->send_action_response(call_id, success, error_message,
reinterpret_cast<const uint8_t *>(json_str.data()), json_str.size());
return;
}
#endif
this->parent_->send_action_response(call_id, success, error_message);
}
protected:
APIServer *parent_;
TemplatableValue<bool, Ts...> success_{true};
TemplatableValue<std::string, Ts...> error_message_{""};
#ifdef USE_API_USER_DEFINED_ACTION_RESPONSES_JSON
std::function<void(Ts..., JsonObject)> json_builder_;
bool has_data_{false};
#endif
bool is_optional_mode_{false};
};
// Action to unregister a service call after execution completes
// Automatically appended to the end of action lists for non-none response modes
template<typename... Ts> class APIUnregisterServiceCallAction : public Action<Ts...> {
public:
explicit APIUnregisterServiceCallAction(APIServer *parent) : parent_(parent) {}
void play(const Ts &...x) override {
// Extract call_id from first argument - same convention as APIRespondAction
auto args = std::make_tuple(x...);
uint32_t call_id = std::get<0>(args);
if (call_id != 0) {
this->parent_->unregister_active_action_call(call_id);
}
}
protected:
APIServer *parent_;
};
} // namespace esphome::api
#endif // USE_API_USER_DEFINED_ACTION_RESPONSES