Merge pull request #265 from acon96/release/v0.3.8

Release v0.3.8

Author: acon96

.github/workflows/create-release.yml

@@ -20,33 +20,33 @@ jobs:
       matrix:
         include:
         # ARM variants
-        - home_assistant_version: "2024.12.3"
+        - home_assistant_version: "2025.4.1"
           arch: "aarch64"
-        - home_assistant_version: "2024.12.3"
+        - home_assistant_version: "2025.4.1"
           arch: "armhf"
 
         # Base x86
-        - home_assistant_version: "2024.12.3"
+        - home_assistant_version: "2025.4.1"
           suffix: "-noavx"
           arch: "amd64"
           extra_defines: "-DGGML_AVX=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF -DGGML_F16C=OFF"
-        - home_assistant_version: "2024.12.3"
+        - home_assistant_version: "2025.4.1"
           arch: "i386"
           suffix: "-noavx"
           extra_defines: "-DGGML_AVX=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF -DGGML_F16C=OFF"
 
         # AVX2 and AVX512
-        - home_assistant_version: "2024.12.3"
+        - home_assistant_version: "2025.4.1"
           arch: "amd64"
           extra_defines: "-DGGML_AVX=ON -DGGML_AVX2=ON -DGGML_FMA=ON -DGGML_F16C=ON"
-        - home_assistant_version: "2024.12.3"
+        - home_assistant_version: "2025.4.1"
           arch: "amd64"
           suffix: "-avx512"
           extra_defines: "-DGGML_AVX512=ON -DGGML_FMA=ON -DGGML_F16C=ON"
-        - home_assistant_version: "2024.12.3"
+        - home_assistant_version: "2025.4.1"
           arch: "i386"
           extra_defines: "-DGGML_AVX=ON -DGGML_AVX2=ON -DGGML_FMA=ON -DGGML_F16C=ON"
-        - home_assistant_version: "2024.12.3"
+        - home_assistant_version: "2025.4.1"
           arch: "i386"
           suffix: "-avx512"
           extra_defines: "-DGGML_AVX512=ON -DGGML_FMA=ON -DGGML_F16C=ON"

README.md

@@ -5,7 +5,7 @@ This project provides the required "glue" components to control your Home Assist
 Please see the [Setup Guide](./docs/Setup.md) for more information on installation.
 
 ## Local LLM Conversation Integration
-**The latest version of this integration requires Home Assistant 2024.12.3 or newer**
+**The latest version of this integration requires Home Assistant 2025.4.1 or newer**
 
 In order to integrate with Home Assistant, we provide a custom component that exposes the locally running LLM as a "conversation agent".
 
@@ -150,6 +150,7 @@ In order to facilitate running the project entirely on the system where Home Ass
 ## Version History
 | Version | Description                                                                                                                                                                                                          |
 |---------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| v0.3.8  | Update llama.cpp, remove think blocks from "thinking" models, fix wheel detection for some Intel CPUs, Fixes for compatibility with latest Home Assistant version (2025.4), other small bug fixes                    |
 | v0.3.7  | Update llama.cpp version to support newer models, Update minimum Home Assistant version to 2024.12.3, Add German In-Context Learning examples, Fix multi-turn use, Fix an issue with webcolors                       | 
 | v0.3.6  | Small llama.cpp backend fixes                                                                                                                                                                                        |
 | v0.3.5  | Fix for llama.cpp backend installation, Fix for Home LLM v1-3 API parameters, add Polish ICL examples                                                                                                                |

custom_components/llama_conversation/__init__.py

@@ -8,19 +8,31 @@
 from homeassistant.config_entries import ConfigEntry
 from homeassistant.const import ATTR_ENTITY_ID, Platform
 from homeassistant.core import HomeAssistant
-from homeassistant.exceptions import HomeAssistantError
 from homeassistant.helpers import config_validation as cv, llm
 from homeassistant.util.json import JsonObjectType
 
 import voluptuous as vol
 
-
 from .const import (
     ALLOWED_SERVICE_CALL_ARGUMENTS,
     DOMAIN,
     HOME_LLM_API_ID,
     SERVICE_TOOL_NAME,
+    SERVICE_TOOL_ALLOWED_SERVICES,
+    SERVICE_TOOL_ALLOWED_DOMAINS,
+    CONF_BACKEND_TYPE,
+    DEFAULT_BACKEND_TYPE,
+    BACKEND_TYPE_LLAMA_HF,
+    BACKEND_TYPE_LLAMA_EXISTING,
+    BACKEND_TYPE_TEXT_GEN_WEBUI,
+    BACKEND_TYPE_GENERIC_OPENAI,
+    BACKEND_TYPE_LLAMA_CPP_PYTHON_SERVER,
+    BACKEND_TYPE_OLLAMA,
 )
+from .conversation import LlamaCppAgent, GenericOpenAIAPIAgent, TextGenerationWebuiAgent, \
+    LlamaCppPythonAPIAgent, OllamaAPIAgent, LocalLLMAgent
+
+type LocalLLMConfigEntry = ConfigEntry[LocalLLMAgent]
 
 _LOGGER = logging.getLogger(__name__)
 
@@ -29,26 +41,51 @@
 PLATFORMS = (Platform.CONVERSATION,)
 
 
-async def async_setup_entry(hass: HomeAssistant, entry: ConfigEntry) -> bool:
+async def async_setup_entry(hass: HomeAssistant, entry: LocalLLMConfigEntry) -> bool:
 
     # make sure the API is registered
     if not any([x.id == HOME_LLM_API_ID for x in llm.async_get_apis(hass)]):
         llm.async_register_api(hass, HomeLLMAPI(hass))
 
     hass.data.setdefault(DOMAIN, {})[entry.entry_id] = entry
 
+    def create_agent(backend_type):
+        agent_cls = None
+
+        if backend_type in [ BACKEND_TYPE_LLAMA_HF, BACKEND_TYPE_LLAMA_EXISTING ]:
+            agent_cls = LlamaCppAgent
+        elif backend_type == BACKEND_TYPE_GENERIC_OPENAI:
+            agent_cls = GenericOpenAIAPIAgent
+        elif backend_type == BACKEND_TYPE_TEXT_GEN_WEBUI:
+            agent_cls = TextGenerationWebuiAgent
+        elif backend_type == BACKEND_TYPE_LLAMA_CPP_PYTHON_SERVER:
+            agent_cls = LlamaCppPythonAPIAgent
+        elif backend_type == BACKEND_TYPE_OLLAMA:
+            agent_cls = OllamaAPIAgent
+        
+        return agent_cls(hass, entry)
+
+    # create the agent in an executor job because the constructor calls `open()`
+    backend_type = entry.data.get(CONF_BACKEND_TYPE, DEFAULT_BACKEND_TYPE)
+    entry.runtime_data = await hass.async_add_executor_job(create_agent, backend_type)
+
+    # call load model
+    await entry.runtime_data._async_load_model(entry)
+
+    # forward setup to platform to register the entity
     await hass.config_entries.async_forward_entry_setups(entry, PLATFORMS)
+    
     return True
 
 
-async def async_unload_entry(hass: HomeAssistant, entry: ConfigEntry) -> bool:
+async def async_unload_entry(hass: HomeAssistant, entry: LocalLLMConfigEntry) -> bool:
     """Unload Ollama."""
     if not await hass.config_entries.async_unload_platforms(entry, PLATFORMS):
         return False
     hass.data[DOMAIN].pop(entry.entry_id)
     return True
 
-async def async_migrate_entry(hass, config_entry: ConfigEntry):
+async def async_migrate_entry(hass: HomeAssistant, config_entry: LocalLLMConfigEntry):
     """Migrate old entry."""
     _LOGGER.debug("Migrating from version %s", config_entry.version)
 
@@ -82,13 +119,8 @@ class HassServiceTool(llm.Tool):
         vol.Optional('item'): str,
     })
 
-    ALLOWED_SERVICES: Final[list[str]] = [
-        "turn_on", "turn_off", "toggle", "press", "increase_speed", "decrease_speed", "open_cover", "close_cover", "stop_cover",
-        "lock", "unlock", "start", "stop", "return_to_base", "pause", "cancel", "add_item"
-    ]
-    ALLOWED_DOMAINS: Final[list[str]] = [
-        "light", "switch", "button", "fan", "cover", "lock", "media_player", "climate", "vacuum", "todo", "timer", "script",
-    ]
+    ALLOWED_SERVICES: Final[list[str]] = SERVICE_TOOL_ALLOWED_SERVICES
+    ALLOWED_DOMAINS: Final[list[str]] = SERVICE_TOOL_ALLOWED_DOMAINS
 
     async def async_call(
         self, hass: HomeAssistant, tool_input: llm.ToolInput, llm_context: llm.LLMContext

custom_components/llama_conversation/config_flow.py

@@ -687,6 +687,7 @@ async def async_step_model_parameters(
         for key in OPTIONS_OVERRIDES.keys():
             if key in model_name:
                 selected_default_options.update(OPTIONS_OVERRIDES[key])
+                break
 
         persona = PERSONA_PROMPTS.get(self.selected_language, PERSONA_PROMPTS.get("en"))
         current_date = CURRENT_DATE_PROMPT.get(self.selected_language, CURRENT_DATE_PROMPT.get("en"))
@@ -765,15 +766,15 @@ def async_get_options_flow(
         config_entry: config_entries.ConfigEntry,
     ) -> config_entries.OptionsFlow:
         """Create the options flow."""
-        return OptionsFlow(config_entry)
+        return OptionsFlow()
 
 
 class OptionsFlow(config_entries.OptionsFlow):
     """Local LLM config flow options handler."""
 
-    def __init__(self, config_entry: config_entries.ConfigEntry) -> None:
-        """Initialize options flow."""
-        self.config_entry = config_entry
+    @property
+    def config_entry(self):
+        return self.hass.config_entries.async_get_entry(self.handler)
 
     async def async_step_init(
         self, user_input: dict[str, Any] | None = None

custom_components/llama_conversation/const.py

@@ -4,6 +4,8 @@
 DOMAIN = "llama_conversation"
 HOME_LLM_API_ID = "home-llm-service-api"
 SERVICE_TOOL_NAME = "HassCallService"
+SERVICE_TOOL_ALLOWED_SERVICES = ["turn_on", "turn_off", "toggle", "press", "increase_speed", "decrease_speed", "open_cover", "close_cover", "stop_cover", "lock", "unlock", "start", "stop", "return_to_base", "pause", "cancel", "add_item", "set_temperature", "set_humidity", "set_fan_mode", "set_hvac_mode", "set_preset_mode"]
+SERVICE_TOOL_ALLOWED_DOMAINS = ["light", "switch", "button", "fan", "cover", "lock", "media_player", "climate", "vacuum", "todo", "timer", "script"]
 CONF_PROMPT = "prompt"
 PERSONA_PROMPTS = {
     "en": "You are 'Al', a helpful AI Assistant that controls the devices in a house. Complete the following task as instructed with the information provided only.",
@@ -13,7 +15,7 @@
     "pl": "Jeste\u015b 'Al', pomocnym asystentem AI, kt\u00f3ry kontroluje urz\u0105dzenia w domu. Wykonaj poni\u017csze zadanie zgodnie z instrukcj\u0105 lub odpowiedz na poni\u017csze pytanie, korzystaj\u0105c wy\u0142\u0105cznie z podanych informacji."
 }
 CURRENT_DATE_PROMPT = {
-    "en": """The current time and date is {{ (as_timestamp(now()) | timestamp_custom("%I:%M %p on %A %B %d, %Y", "")) }}""",
+    "en": """The current time and date is {{ (as_timestamp(now()) | timestamp_custom("%I:%M %p on %A %B %d, %Y", True, "")) }}""",
     "de": """{% set day_name = ["Montag", "Dienstag", "Mittwoch", "Donnerstag", "Freitag", "Samstag", "Sonntag"] %}{% set month_name = ["Januar", "Februar", "März", "April", "Mai", "Juni", "Juli", "August", "September", "Oktober", "November", "Dezember"] %}Die aktuelle Uhrzeit und das aktuelle Datum sind {{ (as_timestamp(now()) | timestamp_custom("%H:%M", local=True)) }} {{ day_name[now().weekday()] }}, {{ now().day }} {{ month_name[now().month -1]}} {{ now().year }}.""",
     "fr": """{% set day_name = ["lundi", "mardi", "mercredi", "jeudi", "vendredi", "samedi", "dimanche"] %}{% set month_name = ["janvier", "février", "mars", "avril", "mai", "juin", "juillet", "août", "septembre", "octobre", "novembre", "décembre"] %} L'heure et la date actuelles sont {{ (as_timestamp(now()) | timestamp_custom("%H:%M", local=True)) }} {{ day_name[now().weekday()] }}, {{ now().day }} {{ month_name[now().month -1]}} {{ now().year }}.""",
     "es": """{% set day_name = ["lunes", "martes", "miércoles", "jueves", "viernes", "sábado", "domingo"] %}{% set month_name = ["enero", "febrero", "marzo", "abril", "mayo", "junio", "julio", "agosto", "septiembre", "octubre", "noviembre", "diciembre"] %}La hora y fecha actuales son {{ (as_timestamp(now()) | timestamp_custom("%H:%M", local=True)) }} {{ day_name[now().weekday()] }}, {{ now().day }} de {{ month_name[now().month -1]}} de {{ now().year }}.""",
@@ -146,59 +148,69 @@
         "user": { "prefix": "<|im_start|>user\n", "suffix": "<|im_end|>" },
         "assistant": { "prefix": "<|im_start|>assistant\n", "suffix": "<|im_end|>" },
         "tool": { "prefix": "<|im_start|>tool", "suffix": "<|im_end|>" },
+        "chain_of_thought": { "prefix": "<think>", "suffix": "</think>"},
         "generation_prompt": "<|im_start|>assistant"
     },
     PROMPT_TEMPLATE_COMMAND_R: {
         "system": { "prefix": "<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>", "suffix": "<|END_OF_TURN_TOKEN|>" },
         "user": { "prefix": "<|START_OF_TURN_TOKEN|><|USER_TOKEN|>", "suffix": "<|END_OF_TURN_TOKEN|>" },
         "assistant": { "prefix": "<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>", "suffix": "<|END_OF_TURN_TOKEN|>" },
+        "chain_of_thought": { "prefix": "<think>", "suffix": "</think>"},
         "generation_prompt": "<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>"
     },
     PROMPT_TEMPLATE_ALPACA: {
         "system": { "prefix": "", "suffix": "\n" },
         "user": { "prefix": "### Instruction:\n", "suffix": "\n" },
         "assistant": { "prefix": "### Response:\n", "suffix": "\n" },
+        "chain_of_thought": { "prefix": "<think>", "suffix": "</think>"},
         "generation_prompt": "### Response:"
     },
     PROMPT_TEMPLATE_VICUNA: {
         "system": { "prefix": "", "suffix": "\n" },
         "user": { "prefix": "USER: ", "suffix": "" },
         "assistant": { "prefix": "ASSISTANT: ", "suffix": "</s>" },
+        "chain_of_thought": { "prefix": "<think>", "suffix": "</think>"},
         "generation_prompt": "ASSISTANT:"
     },
     PROMPT_TEMPLATE_NONE: {
         "system": { "prefix": "", "suffix": "" },
         "user": { "prefix": "", "suffix": "" },
         "assistant": { "prefix": "", "suffix": "" },
+        "chain_of_thought": { "prefix": "", "suffix": ""},
         "generation_prompt": ""
     },
     PROMPT_TEMPLATE_MISTRAL: {
         "user": { "prefix": "<s>[INST] ", "suffix": " [/INST] " },
         "assistant": { "prefix": "", "suffix": "</s>" },
+        "chain_of_thought": { "prefix": "<think>", "suffix": "</think>"},
         "generation_prompt": ""
     },
     PROMPT_TEMPLATE_ZEPHYR: {
         "system": { "prefix": "<|system|>\n", "suffix": "<|endoftext|>" },
         "user": { "prefix": "<|user|>\n", "suffix": "<|endoftext|>" },
         "assistant": { "prefix": "<|assistant|>\n", "suffix": "<|endoftext|>" },
+        "chain_of_thought": { "prefix": "<think>", "suffix": "</think>"},
         "generation_prompt": "<|assistant|>\n"
     },
     PROMPT_TEMPLATE_ZEPHYR2: {
         "system": { "prefix": "<|system|>\n", "suffix": "</s>" },
         "user": { "prefix": "<|user|>\n", "suffix": "</s>" },
         "assistant": { "prefix": "<|assistant|>\n", "suffix": "</s>" },
+        "chain_of_thought": { "prefix": "<think>", "suffix": "</think>"},
         "generation_prompt": "<|assistant|>\n"
     },
     PROMPT_TEMPLATE_ZEPHYR3: {
         "system": { "prefix": "<|system|>\n", "suffix": "<|end|>" },
         "user": { "prefix": "<|user|>\n", "suffix": "<|end|>" },
         "assistant": { "prefix": "<|assistant|>\n", "suffix": "<|end|>" },
+        "chain_of_thought": { "prefix": "<think>", "suffix": "</think>"},
         "generation_prompt": "<|assistant|>\n"
     },
     PROMPT_TEMPLATE_LLAMA3: {
         "system": { "prefix": "<|start_header_id|>system<|end_header_id|>\n\n", "suffix": "<|eot_id|>"},
         "user": { "prefix": "<|start_header_id|>user<|end_header_id|>\n\n", "suffix": "<|eot_id|>"},
         "assistant": { "prefix": "<|start_header_id|>assistant<|end_header_id|>\n\n", "suffix": "<|eot_id|>"},
+        "chain_of_thought": { "prefix": "<think>", "suffix": "</think>"},
         "generation_prompt": "<|start_header_id|>assistant<|end_header_id|>\n\n"
     }
 }
@@ -297,6 +309,14 @@
 )
 
 OPTIONS_OVERRIDES = {
+    "home-llama-3.2": {
+        CONF_PROMPT: DEFAULT_PROMPT_BASE_LEGACY,
+        CONF_PROMPT_TEMPLATE: PROMPT_TEMPLATE_LLAMA3,
+        CONF_USE_IN_CONTEXT_LEARNING_EXAMPLES: False,
+        CONF_SERVICE_CALL_REGEX: FINE_TUNED_SERVICE_CALL_REGEX,
+        CONF_TOOL_FORMAT: TOOL_FORMAT_MINIMAL,
+        CONF_CONTEXT_LENGTH: 131072,
+    },
     "home-3b-v3": {
         CONF_PROMPT: DEFAULT_PROMPT_BASE_LEGACY,
         CONF_PROMPT_TEMPLATE: PROMPT_TEMPLATE_ZEPHYR,
@@ -383,5 +403,5 @@
     },
 }
 
-INTEGRATION_VERSION = "0.3.7"
+INTEGRATION_VERSION = "0.3.8"
 EMBEDDED_LLAMA_CPP_PYTHON_VERSION = "0.3.5"