* Include speech_slots in IntentResponse.as_dict
* Populate speech_slots only if available
* fix typo
* Add test
* test all fields
* Fix another test
---------
Co-authored-by: Paulus Schoutsen <balloob@gmail.com>
* Always do thread safety check when writing state
Refactor the 3 most common places where the thread safety check
for the event loop to be inline to make the check fast enough
that we can keep it long term. While code review catches most
of the thread safety issues in core, some of them still make
it through, and new ones keep getting added. Its not possible
to catch them all with manual code review, so its worth the
tiny overhead to check each time.
Previously the check was limited to custom components
because they were the most common source of thread
safety issues.
* Always do thread safety check when writing state
Refactor the 3 most common places where the thread safety check
for the event loop to be inline to make the check fast enough
that we can keep it long term. While code review catches most
of the thread safety issues in core, some of them still make
it through, and new ones keep getting added. Its not possible
to catch them all with manual code review, so its worth the
tiny overhead to check each time.
Previously the check was limited to custom components
because they were the most common source of thread
safety issues.
* async_fire is more common than expected with ccs
* fix mock
* fix hass mocking
* Automatically fill in slots from LLM context
* Add tests
* Apply suggestions from code review
Co-authored-by: Allen Porter <allen@thebends.org>
---------
Co-authored-by: Allen Porter <allen@thebends.org>
* Ignore deprecated open and close cover intents for LLMs
* Fix LLMs asking which area when there is only one device
* remove unrelated changed
* remove unrelated changes
* Add configuration option positive to DurationSelector
* Rename to allow_negative in conjunction with a deprecation notice
Co-authored-by: Erik Montnemery <erik@montnemery.com>
---------
Co-authored-by: Erik Montnemery <erik@montnemery.com>
* LLM Assist API to ignore timer intents if device doesn't support it
* Refactor to use API instances
* Extract ToolContext class
* Limit exposed intents based on exposed entities
These functions have been stripped down to always return
the same path so there was no longer a need to have a
function for this. This is left-over cleanup from
previous refactoring.
* initial commit
* Undo prompt chenges
* Move format_tool out of the class
* Only catch HomeAssistantError and vol.Invalid
* Add config flow option
* Fix type
* Add translation
* Allow changing API access from options flow
* Allow model picking
* Remove allowing HASS Access in main flow
* Move model to the top in options flow
* Make prompt conditional based on API access
* convert only once to dict
* Reduce debug logging
* Update title
* re-order models
* Address comments
* Move things
* Update labels
* Add tool call tests
* coverage
* Use LLM APIs
* Fixes
* Address comments
* Reinstate the title to not break entity name
---------
Co-authored-by: Paulus Schoutsen <balloob@gmail.com>
* Add llm helper
* break out Tool.specification as class members
* Format state output
* Fix intent tests
* Removed auto initialization of intents - let conversation platforms do that
* Handle DynamicServiceIntentHandler.extra_slots
* Add optional description to IntentTool init
* Add device_id and conversation_id parameters
* intent tests
* Add LLM tools tests
* coverage
* add agent_id parameter
* Apply suggestions from code review
* Apply suggestions from code review
* Apply suggestions from code review
* Apply suggestions from code review
* Apply suggestions from code review
* Fix tests
* Fix intent schema
* Allow a Python function to be registered as am LLM tool
* Add IntentHandler.effective_slot_schema
* Ensure IntentHandler.slot_schema to be vol.Schema
* Raise meaningful error on tool not found
* Move this change to a separate PR
* Update todo integration intent
* Remove Tool constructor
* Move IntentTool to intent helper
* Convert custom serializer into class method
* Remove tool_input from FunctionTool auto arguments to avoid recursion
* Remove conversion into Open API format
* Apply suggestions from code review
* Fix tests
* Use HassKey for helpers (see #117012)
* Add support for functions with typed lists, dicts, and sets as type hints
* Remove FunctionTool
* Added API to get registered intents
* Move IntentTool to the llm library
* Return only handlers in intents.async.get
* Removed llm tool registration from intent library
* Removed tool registration
* Add bind_hass back for now
* removed area and floor resolving
* fix test
* Apply suggestions from code review
* Improve coverage
* Fix intent_type type
* Temporary disable HassClimateGetTemperature intent
* Remove bind_hass
* Fix usage of slot schema
* Fix test
* Revert some test changes
* Don't mutate tool_input
---------
Co-authored-by: Paulus Schoutsen <paulus@home-assistant.io>
Co-authored-by: Paulus Schoutsen <balloob@gmail.com>
* Allow templates for enabling automation triggers
* Test exception for non-limited template
* Use `cv.template` instead of `cv.template_complex`
* skip trigger with invalid enable template
instead of returning and thus not evaluating other triggers
* Import and cache supported feature enum flags only when needed
* Add comment aboud being loaded from executor.
---------
Co-authored-by: J. Nick Koston <nick@koston.org>
* Addition of add filter
This change adds an `add` filter, the addition equivalent of the existing `multiply` filter.
* Test for add filter
* Update test_template.py
* Update tests/helpers/test_template.py
---------
Co-authored-by: Erik Montnemery <erik@montnemery.com>
* Turn on thread safety checks in async_dispatcher_send
We keep seeing issues where async_dispatcher_send is called from
a thread which means we call the callback function on the other
side in the thread as well which usually leads to a crash
* Turn on thread safety checks in async_dispatcher_send
We keep seeing issues where async_dispatcher_send is called from
a thread which means we call the callback function on the other
side in the thread as well which usually leads to a crash
* adjust
* Refactor entity_platform polling to avoid double time fetch
Replace async_track_time_interval with loop.call_later
to avoid the useless time fetch every time the listener
fired since we always throw it away
* fix test
It turns out we have custom components that are writing to the area registry using the async APIs from threads. We now catch it at the point async_fire is called. Instead we should check sooner and use async_fire_internal so we catch the unsafe operation before it can corrupt the registry.
* Move thread safety check in entity_registry sooner
It turns out we have a lot of custom components that are writing
to the entity registry using the async APIs from threads. We now
catch it at the point async_fire is called. Instread we should check
sooner and use async_fire_internal so we catch the unsafe operation
before it can corrupt the registry.
* coverage
* Apply suggestions from code review
It turns out we have custom components that are writing to the device registry using the async APIs from threads. We now catch it at the point async_fire is called. Instead we should check sooner and use async_fire_internal so we catch the unsafe operation before it can corrupt the registry.
* Fix lying docstring for relative_time template function
* Update homeassistant/helpers/template.py
Co-authored-by: Martin Hjelmare <marhje52@gmail.com>
---------
Co-authored-by: Martin Hjelmare <marhje52@gmail.com>
Erik Montnemery
Paulus Schoutsen
tronikos
epenet
Sid
paulusbrand
J. Nick Koston
Marc Mueller
Ron Weikamp
Denis Shulyaka
Michael Hansen
Robert Resch
Franck Nijhof
Peter
Matthias Alphart
Allen Porter
Troon
G Johansson
rlippmann