Merge pull request #237 from acon96/release/v0.3.7

Release v0.3.7

.github/workflows/create-release.yml

@@ -20,33 +20,33 @@ jobs:
       matrix:
         include:
         # ARM variants
-        - home_assistant_version: "2024.2.1"
+        - home_assistant_version: "2024.12.3"
           arch: "aarch64"
-        - home_assistant_version: "2024.2.1"
+        - home_assistant_version: "2024.12.3"
           arch: "armhf"
 
         # Base x86
-        - home_assistant_version: "2024.2.1"
+        - home_assistant_version: "2024.12.3"
           suffix: "-noavx"
           arch: "amd64"
           extra_defines: "-DGGML_AVX=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF -DGGML_F16C=OFF"
-        - home_assistant_version: "2024.2.1"
+        - home_assistant_version: "2024.12.3.1"
           arch: "i386"
           suffix: "-noavx"
           extra_defines: "-DGGML_AVX=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF -DGGML_F16C=OFF"
 
         # AVX2 and AVX512
-        - home_assistant_version: "2024.2.1"
+        - home_assistant_version: "2024.12.3"
           arch: "amd64"
           extra_defines: "-DGGML_AVX=ON -DGGML_AVX2=ON -DGGML_FMA=ON -DGGML_F16C=ON"
-        - home_assistant_version: "2024.2.1"
+        - home_assistant_version: "2024.12.3.1"
           arch: "amd64"
           suffix: "-avx512"
           extra_defines: "-DGGML_AVX512=ON -DGGML_FMA=ON -DGGML_F16C=ON"
-        - home_assistant_version: "2024.2.1"
+        - home_assistant_version: "2024.12.3"
           arch: "i386"
           extra_defines: "-DGGML_AVX=ON -DGGML_AVX2=ON -DGGML_FMA=ON -DGGML_F16C=ON"
-        - home_assistant_version: "2024.2.1"
+        - home_assistant_version: "2024.12.3"
           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 2024.8.0 or newer**
+**The latest version of this integration requires Home Assistant 2024.12.3 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                                                                                                                                                                                                          |
 |---------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| v0.3.7  | 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                                                                                                                |
 | v0.3.4  | Significantly improved language support including full Polish translation, Update bundled llama-cpp-python to support new models, various bug fixes                                                                  |

custom_components/llama_conversation/config_flow.py

@@ -37,7 +37,7 @@ from homeassistant.helpers.selector import (
 from homeassistant.util.package import is_installed
 from importlib.metadata import version
 
-from .utils import download_model_from_hf, install_llama_cpp_python, format_url, MissingQuantizationException
+from .utils import download_model_from_hf, get_llama_cpp_python_version, install_llama_cpp_python, format_url, MissingQuantizationException
 from .const import (
     CONF_CHAT_MODEL,
     CONF_MAX_TOKENS,
@@ -352,7 +352,9 @@ class ConfigFlow(BaseLlamaConversationConfigFlow, config_entries.ConfigFlow, dom
             local_backend = is_local_backend(user_input[CONF_BACKEND_TYPE])
             self.model_config.update(user_input)
             if local_backend:
-                if is_installed("llama-cpp-python") and version("llama-cpp-python") == EMBEDDED_LLAMA_CPP_PYTHON_VERSION:
+                installed_version = await self.hass.async_add_executor_job(get_llama_cpp_python_version)
+                _LOGGER.debug(f"installed version: {installed_version}")
+                if installed_version == EMBEDDED_LLAMA_CPP_PYTHON_VERSION:
                     return await self.async_step_local_model()
                 else:
                     return await self.async_step_install_local_wheels()

custom_components/llama_conversation/const.py

@@ -383,5 +383,5 @@ OPTIONS_OVERRIDES = {
     },
 }
 
-INTEGRATION_VERSION = "0.3.6"
+INTEGRATION_VERSION = "0.3.7"
-EMBEDDED_LLAMA_CPP_PYTHON_VERSION = "0.2.88"
+EMBEDDED_LLAMA_CPP_PYTHON_VERSION = "0.3.5"

custom_components/llama_conversation/conversation.py

@@ -421,8 +421,10 @@ class LocalLLMAgent(ConversationEntity, AbstractConversationAgent):
                 response=intent_response, conversation_id=conversation_id
             )
 
+        tool_response = None
         # parse response
         to_say = service_call_pattern.sub("", response.strip())
+        tool_response = None
         for block in service_call_pattern.findall(response.strip()):
             parsed_tool_call: dict = json.loads(block)
 
@@ -505,8 +507,11 @@ class LocalLLMAgent(ConversationEntity, AbstractConversationAgent):
                 )
 
         # handle models that generate a function call and wait for the result before providing a response
-        if self.entry.options.get(CONF_TOOL_MULTI_TURN_CHAT, DEFAULT_TOOL_MULTI_TURN_CHAT):
+        if self.entry.options.get(CONF_TOOL_MULTI_TURN_CHAT, DEFAULT_TOOL_MULTI_TURN_CHAT) and tool_response is not None:
-            conversation.append({"role": "tool", "message": json.dumps(tool_response)})
+            try:
+                conversation.append({"role": "tool", "message": json.dumps(tool_response)})
+            except:
+                conversation.append({"role": "tool", "message": "No tools were used in this response."})
 
             # generate a response based on the tool result
             try:
@@ -527,6 +532,7 @@ class LocalLLMAgent(ConversationEntity, AbstractConversationAgent):
                 )
 
             conversation.append({"role": "assistant", "message": response})
+            conversation.append({"role": "assistant", "message": to_say})
         
         # generate intent response to Home Assistant
         intent_response = intent.IntentResponse(language=user_input.language)
@@ -749,12 +755,14 @@ class LocalLLMAgent(ConversationEntity, AbstractConversationAgent):
 
                 value = attributes[attribute_name]
                 if value is not None:
-                    if attribute_name == "temperature":
+                    # try to apply unit if present
-                        value = int(value)
+                    unit_suffix = attributes.get(f"{attribute_name}_unit")
-                        if value > 50:
+                    if unit_suffix:
-                            value = f"{value}F"
+                        value = f"{value} {unit_suffix}"
-                        else:
+                    elif attribute_name == "temperature":
-                            value = f"{value}C"
+                        # try to get unit or guess otherwise
+                        suffix = "F" if value > 50 else "C"
+                        value = F"{int(value)} {suffix}"
                     elif attribute_name == "rgb_color":
                         value = F"{closest_color(value)} {value}"
                     elif attribute_name == "volume_level":

custom_components/llama_conversation/in_context_examples_de.csv

@@ -0,0 +1,24 @@
+type,request,tool,response
+fan,Schalte den <name> ein,HassTurnOn,Ich habe den Ventilator eingeschaltet.
+fan,Schalte die Ventilatoren in der <area> ein.,HassTurnOn,Ich habe den Ventilator in der <area> eingeschaltet.
+fan,Schalte den <name> aus,HassTurnOff,Ich habe den Ventilator ausgeschaltet.
+fan,Schalte den Ventilator <name> um,HassToggle,Ich habe den Ventilator umgeschaltet.
+light,Schalte das <name> Licht an,HassTurnOn,Ich habe das Licht eingeschaltet.
+light,Schalte das <name> Licht aus,HassTurnOff,Ich habe das Licht ausgeschaltet.
+light,Schalte das Licht in <area> aus,HassTurnOff,Ich habe das Licht in <area>  ausgeschaltet.
+light,Schalte das Licht in <area> um,HassToggle,Ich habe das Licht in <area> umgeschaltet.
+light,Schalte die Helligkeit von <name> auf <brightness>,HassLightSet,Ich habe die Helligkeit umgeschaltet.
+light,Schalte die Lichter in <area> auf <color>,HassLightSet,Ich habe die Farbe jetzt geändert.
+media_player,Schalte den <name> ein,HassTurnOn,Ich schalte den Media-Player für dich ein.
+media_player,<name> sollte ausgeschaltet werden,HassTurnOff,Ich habe den Media-Player ausgeschaltet.
+media_player,Drücke den Play-Knopf am <name>,HassMediaUnpause,Starte die Medienwiedergabe.
+media_player,Pausiere den <name>,HassMediaPause,Pausiere die Medienwiedergabe.
+media_player,Spiele das nächste auf <name>,HassMediaNext,Überspringe zum nächsten Track.
+switch,Schalte den <name> ein,HassTurnOn,Ich habe den Schalter eingeschaltet.
+switch,Schalte den <name> aus,HassTurnOff,Ich habe den Schalter ausgeschaltet.
+switch,Schalte die Schalter in der <area> aus,HassTurnOff,Ich habe die Geräte wie gewünscht ausgeschaltet.
+switch,Schalte den Schalter <name> um,HassToggle,Ich habe den Schalter umgeschaltet.
+vacuum,Starte den Staubsauger namens <name>,HassVacuumStart,Ich habe den Staubsauger gestartet.
+vacuum,Stoppe den Staubsauger <name>,HassVacuumReturnToBase,Ich habe den Staubsauger zurück zur Basis geschickt.
+blinds, Schliesse den Rolladen <name>, HassSetPosition, Ich habe den Rolladen geschlossen.
+blinds, Öffne den Rolladen <name>, HassSetPosition, Ich habe den Rolladen geöffnet.

custom_components/llama_conversation/manifest.json

@@ -1,7 +1,7 @@
 {
   "domain": "llama_conversation",
   "name": "Local LLM Conversation",
-  "version": "0.3.6",
+  "version": "0.3.7",
   "codeowners": ["@acon96"],
   "config_flow": true,
   "dependencies": ["conversation"],
@@ -10,7 +10,7 @@
   "integration_type": "service",
   "iot_class": "local_polling",
   "requirements": [
-    "huggingface-hub==0.23.0",
+      "huggingface-hub>=0.23.0",
-    "webcolors<=1.13"
+      "webcolors>=24.8.0"
   ]
 }

custom_components/llama_conversation/utils.py

@@ -1,11 +1,13 @@
 import time
 import os
+import re
 import sys
 import platform
 import logging
 import multiprocessing
 import voluptuous as vol
 import webcolors
+from webcolors import CSS3
 from importlib.metadata import version
 
 from homeassistant.helpers import config_validation as cv
@@ -21,6 +23,12 @@ from .const import (
 
 _LOGGER = logging.getLogger(__name__)
 
+CSS3_NAME_TO_RGB = {
+    name: webcolors.name_to_rgb(name, CSS3)
+    for name
+    in webcolors.names(CSS3)
+}
+
 class MissingQuantizationException(Exception):
     def __init__(self, missing_quant: str, available_quants: list[str]):
         self.missing_quant = missing_quant
@@ -28,8 +36,9 @@ class MissingQuantizationException(Exception):
 
 def closest_color(requested_color):
     min_colors = {}
-    for key, name in webcolors.CSS3_HEX_TO_NAMES.items():
+    
-        r_c, g_c, b_c = webcolors.hex_to_rgb(key)
+    for name, rgb in CSS3_NAME_TO_RGB.items():
+        r_c, g_c, b_c = rgb
         rd = (r_c - requested_color[0]) ** 2
         gd = (g_c - requested_color[1]) ** 2
         bd = (b_c - requested_color[2]) ** 2
@@ -97,10 +106,13 @@ def download_model_from_hf(model_name: str, quantization_type: str, storage_fold
     
     fs = HfFileSystem()
     potential_files = [ f for f in fs.glob(f"{model_name}/*.gguf") ]
-    wanted_file = [f for f in potential_files if (f".{quantization_type.lower()}." in f or f".{quantization_type.upper()}." in f)]
+    wanted_file = [f for f in potential_files if (f"{quantization_type.lower()}.gguf" in f or f"{quantization_type.upper()}.gguf" in f)]
 
     if len(wanted_file) != 1:
-        available_quants = [file.split(".")[-2].upper() for file in potential_files]
+        available_quants = [
+            re.split(r"\.|-", file.removesuffix(".gguf"))[-1].upper()
+            for file in potential_files
+        ]
         raise MissingQuantizationException(quantization_type, available_quants)
     try:
         os.makedirs(storage_folder, exist_ok=True)
@@ -138,6 +150,11 @@ def validate_llama_cpp_python_installation():
     if process.exitcode != 0:
         raise Exception(f"Failed to properly initialize llama-cpp-python. (Exit code {process.exitcode}.)")
 
+def get_llama_cpp_python_version():
+    if not is_installed("llama-cpp-python"):
+        return None
+    return version("llama-cpp-python")
+
 def install_llama_cpp_python(config_dir: str):
 
     installed_wrong_version = False

hacs.json

@@ -1,6 +1,6 @@
 {
     "name": "Local LLM Conversation",    
-    "homeassistant": "2024.8.0",
+    "homeassistant": "2024.12.3",
     "content_in_root": false,
     "render_readme": true
 } 

requirements.txt

@@ -14,9 +14,8 @@ langcodes
 babel==2.15.0
 
 # integration requirements
-requests>=2.31.0
+huggingface-hub>=0.23.0
-huggingface-hub==0.23.0
+webcolors>=24.8.0
-webcolors==1.13
 
 # types from Home Assistant
 homeassistant>=2024.6.1

scripts/run_docker_to_make_wheels.sh

@@ -1,15 +1,21 @@
 #!/bin/bash
 
-VERSION_TO_BUILD="v0.2.88"
+VERSION_TO_BUILD="v0.3.5"
 
-# make python11 wheels
+# make python 11 wheels
 # docker run -it --rm \
 #     --entrypoint bash \
 #     -v $(pwd):/tmp/dist \
 #     homeassistant/home-assistant:2023.12.4 /tmp/dist/make_wheel.sh $VERSION_TO_BUILD
 
 # make python 12 wheels
+# docker run -it --rm \
+#     --entrypoint bash \
+#     -v $(pwd):/tmp/dist \
+#     homeassistant/home-assistant:2024.2.1 /tmp/dist/make_wheel.sh $VERSION_TO_BUILD
+
+# make python 13 wheels
 docker run -it --rm \
     --entrypoint bash \
     -v $(pwd):/tmp/dist \
-    homeassistant/home-assistant:2024.2.1 /tmp/dist/make_wheel.sh $VERSION_TO_BUILD
+    homeassistant/home-assistant:2024.12.3 /tmp/dist/make_wheel.sh $VERSION_TO_BUILD