Merge branch 'develop' into copy_and_paste

This commit is contained in:
misyaguziya
2026-01-09 19:18:58 +09:00
100 changed files with 5548 additions and 1108 deletions

View File

@@ -1,7 +0,0 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.

View File

@@ -1,7 +0,0 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.

View File

@@ -1,7 +0,0 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.

View File

@@ -1,7 +0,0 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.

View File

@@ -4,14 +4,14 @@ from langchain_google_genai import ChatGoogleGenerativeAI
try:
from .translation_languages import translation_lang
from .translation_utils import loadPromptConfig
from .translation_utils import loadTranslatePromptConfig
except Exception:
import sys
from os import path as os_path
print(os_path.dirname(os_path.dirname(os_path.dirname(os_path.abspath(__file__)))))
sys.path.append(os_path.dirname(os_path.dirname(os_path.dirname(os_path.abspath(__file__)))))
from translation_languages import translation_lang
from translation_utils import loadPromptConfig
from translation_utils import loadTranslatePromptConfig
logger = logging.getLogger("langchain_google_genai")
logger.setLevel(logging.ERROR)
@@ -57,9 +57,19 @@ class GeminiClient:
self.model = None
# プロンプト設定をYAMLファイルから読み込む
prompt_config = loadPromptConfig(root_path, "translation_gemini.yml")
prompt_config = loadTranslatePromptConfig(root_path, "translation_gemini.yml")
self.supported_languages = list(translation_lang["Gemini_API"]["source"].keys())
self.prompt_template = prompt_config["system_prompt"]
# history config (optional)
self.history_cfg = prompt_config.get("history", {
"use_history": False,
"sources": [],
"max_messages": 0,
"max_chars": 0,
"header_template": "",
"item_template": "[{source}] {role}: {text}",
})
self._context_history: list[dict] = []
self.gemini_llm = None
@@ -91,6 +101,16 @@ class GeminiClient:
api_key=self.api_key,
)
def setContextHistory(self, history_items: list[dict]) -> None:
"""Set recent conversation history for prompt injection.
Each item should be a dict containing:
- source: "chat" | "mic" | "speaker"
- text: message string
- timestamp: ISO format datetime string
"""
self._context_history = history_items or []
def translate(self, text: str, input_lang: str, output_lang: str) -> str:
system_prompt = self.prompt_template.format(
supported_languages=self.supported_languages,
@@ -98,6 +118,41 @@ class GeminiClient:
output_lang=output_lang
)
# Inject recent conversation history if enabled by YAML config
if self.history_cfg.get("use_history"):
allowed_sources = set(self.history_cfg.get("sources", []))
max_messages = int(self.history_cfg.get("max_messages", 0))
max_chars = int(self.history_cfg.get("max_chars", 0))
item_tmpl = self.history_cfg.get("item_template", "[{source}] {role}: {text}")
header_tmpl = self.history_cfg.get("header_template", "{history}")
filtered = [h for h in self._context_history if h.get("source") in allowed_sources]
recent = filtered[-max_messages:] if max_messages > 0 else filtered
formatted_items = []
for h in recent:
# Format timestamp as HH:MM to save tokens
timestamp_str = ''
if 'timestamp' in h:
from datetime import datetime
try:
ts = datetime.fromisoformat(h['timestamp'])
timestamp_str = ts.strftime('%H:%M')
except:
timestamp_str = ''
formatted_items.append(
item_tmpl.format(
timestamp=timestamp_str,
source=h.get("source", ""),
text=h.get("text", ""),
)
)
history_blob = "\n".join(formatted_items).strip()
if max_chars and len(history_blob) > max_chars:
history_blob = history_blob[-max_chars:]
history_header = header_tmpl.format(max_messages=max_messages, history=history_blob)
if history_header:
system_prompt = f"{system_prompt}\n\n{history_header}"
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": text}

View File

@@ -0,0 +1,201 @@
from openai import OpenAI
from langchain_openai import ChatOpenAI
from pydantic import SecretStr
try:
from .translation_languages import translation_lang
from .translation_utils import loadTranslatePromptConfig
except Exception:
import sys
from os import path as os_path
sys.path.append(os_path.dirname(os_path.dirname(os_path.dirname(os_path.abspath(__file__)))))
from translation_languages import translation_lang, loadTranslationLanguages
from translation_utils import loadTranslatePromptConfig
translation_lang = loadTranslationLanguages(path=".", force=True)
def _authentication_check(api_key: str) -> bool:
"""Check if the provided API key is valid by attempting to list models.
"""
try:
client = OpenAI(
api_key=api_key,
base_url="https://api.groq.com/openai/v1",
)
client.models.list()
return True
except Exception:
return False
def _get_available_text_models(api_key: str) -> list[str]:
"""Extract only Groq models suitable for translation and chat applications.
"""
client = OpenAI(
api_key=api_key,
base_url="https://api.groq.com/openai/v1",
)
res = client.models.list()
allowed_models = []
for model in res.data:
model_id = model.id
# 除外対象のキーワード
exclude_keywords = [
"whisper", # 音声認識
"embedding", # 埋め込み
"image", # 画像生成
"tts", # 音声合成
"audio", # 音声系
"search", # 検索補助モデル
"transcribe", # 音声→文字起こし
"diarize", # 話者分離
"vision" # 画像入力系
]
# 除外キーワードが含まれているモデルをスキップ
if any(kw in model_id.lower() for kw in exclude_keywords):
continue
# テキスト処理用モデルのみ対象
allowed_models.append(model_id)
allowed_models.sort()
return allowed_models
class GroqClient:
"""Groq API Translation wrapper using OpenAI-compatible endpoint.
Groq provides a fast LLM inference platform with an OpenAI-compatible API.
The API endpoint: https://api.groq.com/openai/v1
"""
def __init__(self, root_path: str = None):
self.api_key = None
self.model = None
self.base_url = "https://api.groq.com/openai/v1"
prompt_config = loadTranslatePromptConfig(root_path, "translation_groq.yml")
self.supported_languages = list(translation_lang["Groq_API"]["source"].keys())
self.prompt_template = prompt_config["system_prompt"]
# history config (optional)
self.history_cfg = prompt_config.get("history", {
"use_history": False,
"sources": [],
"max_messages": 0,
"max_chars": 0,
"header_template": "",
"item_template": "[{source}] {role}: {text}",
})
self._context_history: list[dict] = []
self.groq_llm = None
def getModelList(self) -> list[str]:
return _get_available_text_models(self.api_key) if self.api_key else []
def getAuthKey(self) -> str:
return self.api_key
def setAuthKey(self, api_key: str) -> bool:
result = _authentication_check(api_key)
if result:
self.api_key = api_key
return result
def getModel(self) -> str:
return self.model
def setModel(self, model: str) -> bool:
if model in self.getModelList():
self.model = model
return True
else:
return False
def updateClient(self) -> None:
self.groq_llm = ChatOpenAI(
base_url=self.base_url,
model=self.model,
api_key=SecretStr(self.api_key),
streaming=False,
)
def setContextHistory(self, history_items: list[dict]) -> None:
"""Set recent conversation history for prompt injection.
Each item should be a dict containing:
- source: "chat" | "mic" | "speaker"
- text: message string
- timestamp: ISO format datetime string
"""
self._context_history = history_items or []
def translate(self, text: str, input_lang: str, output_lang: str) -> str:
system_prompt = self.prompt_template.format(
supported_languages=self.supported_languages,
input_lang=input_lang,
output_lang=output_lang,
)
# Inject recent conversation history if enabled by YAML config
if self.history_cfg.get("use_history"):
allowed_sources = set(self.history_cfg.get("sources", []))
max_messages = int(self.history_cfg.get("max_messages", 0))
max_chars = int(self.history_cfg.get("max_chars", 0))
item_tmpl = self.history_cfg.get("item_template", "[{source}] {role}: {text}")
header_tmpl = self.history_cfg.get("header_template", "{history}")
filtered = [h for h in self._context_history if h.get("source") in allowed_sources]
recent = filtered[-max_messages:] if max_messages > 0 else filtered
formatted_items = []
for h in recent:
# Format timestamp as HH:MM to save tokens
timestamp_str = ''
if 'timestamp' in h:
from datetime import datetime
try:
ts = datetime.fromisoformat(h['timestamp'])
timestamp_str = ts.strftime('%H:%M')
except:
timestamp_str = ''
formatted_items.append(
item_tmpl.format(
timestamp=timestamp_str,
source=h.get("source", ""),
text=h.get("text", ""),
)
)
history_blob = "\n".join(formatted_items).strip()
if max_chars and len(history_blob) > max_chars:
history_blob = history_blob[-max_chars:]
history_header = header_tmpl.format(max_messages=max_messages, history=history_blob)
if history_header:
system_prompt = f"{system_prompt}\n\n{history_header}"
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": text},
]
resp = self.groq_llm.invoke(messages)
content = ""
if isinstance(resp.content, str):
content = resp.content
elif isinstance(resp.content, list):
for part in resp.content:
if isinstance(part, str):
content += part
elif isinstance(part, dict) and "content" in part and isinstance(part["content"], str):
content += part["content"]
return content.strip()
if __name__ == "__main__":
AUTH_KEY = "GROQ_API_KEY"
client = GroqClient()
client.setAuthKey(AUTH_KEY)
models = client.getModelList()
if models:
print("Available models:", models)
model = input("Select a model: ")
client.setModel(model)
client.updateClient()
print(client.translate("こんにちは世界", "Japanese", "English"))

View File

@@ -41,14 +41,14 @@ def _load_languages(path: str, filename: str) -> str:
Returns:
Absolute path to the resource file
"""
if os.path.exists(os.path.join(path, "_internal", "languages", "languages.yml")):
languages_path = os.path.join(path, "_internal", "languages", "languages.yml")
elif os.path.exists(os.path.join(os.path.dirname(os.path.abspath(__file__)), "models", "translation", "languages", "languages.yml")):
languages_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "models", "translation", "languages", "languages.yml")
elif os.path.exists(os.path.join(os.path.dirname(os.path.abspath(__file__)), "languages", "languages.yml")):
languages_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "languages", "languages.yml")
if os.path.exists(os.path.join(path, "_internal", "translation_settings", "languages", filename)):
languages_path = os.path.join(path, "_internal", "translation_settings", "languages", filename)
elif os.path.exists(os.path.join(os.path.dirname(os.path.abspath(__file__)), "models", "translation", "translation_settings", "languages", filename)):
languages_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "models", "translation", "translation_settings", "languages", filename)
elif os.path.exists(os.path.join(os.path.dirname(os.path.abspath(__file__)), "translation_settings", "languages", filename)):
languages_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "translation_settings", "languages", filename)
else:
raise FileNotFoundError(f"Prompt file not found: {filename}")
raise FileNotFoundError(f"Languages file not found: {filename}")
with open(languages_path, "r", encoding="utf-8") as f:
return yaml.safe_load(f)
@@ -99,7 +99,7 @@ def loadTranslationLanguages(path: str, force: bool = False) -> Dict[str, Any]:
if _loaded and not force:
return translation_lang
data = _load_languages(path, "languages/languages.yml")
data = _load_languages(path, "languages.yml")
if not isinstance(data, dict):
raise ValueError(

View File

@@ -1,40 +1,42 @@
from openai import OpenAI
from langchain_openai import ChatOpenAI
from pydantic import SecretStr
import requests
try:
from .translation_languages import translation_lang
from .translation_utils import loadPromptConfig
from .translation_utils import loadTranslatePromptConfig
except Exception:
import sys
from os import path as os_path
sys.path.append(os_path.dirname(os_path.dirname(os_path.dirname(os_path.abspath(__file__)))))
from translation_languages import translation_lang
from translation_utils import loadPromptConfig
sys.path.append(os_path.dirname(os_path.abspath(__file__)))
from translation_languages import translation_lang, loadTranslationLanguages
from translation_utils import loadTranslatePromptConfig
translation_lang = loadTranslationLanguages(path=".", force=True)
def _authentication_check(api_key: str, base_url: str | None = None) -> bool:
def _authentication_check(base_url: str | None = None) -> bool:
"""Check if the provided API key is valid by attempting to list models.
"""
try:
client = OpenAI(api_key=api_key, base_url=base_url)
client.models.list()
return True
response = requests.get(f"{base_url}/models", timeout=0.2)
if response.status_code == 200:
return True
else:
return False
except Exception:
return False
def _get_available_text_models(api_key: str, base_url: str | None = None) -> list[str]:
def _get_available_text_models(base_url: str | None = None) -> list[str]:
"""Extract the list of available text models from the LM Studio.
"""
try:
client = OpenAI(api_key=api_key, base_url=base_url)
res = client.models.list()
models = res.data
response = requests.get(f"{base_url}/models", timeout=0.2)
models = response.json()["data"]
except Exception:
models = []
allowed_models = []
for model in models:
allowed_models.append(model.id)
allowed_models.append(model["id"])
allowed_models.sort()
return allowed_models
@@ -48,9 +50,19 @@ class LMStudioClient:
self.model = None
self.base_url = base_url # None の場合は公式エンドポイント
prompt_config = loadPromptConfig(root_path, "translation_lmstudio.yml")
prompt_config = loadTranslatePromptConfig(root_path, "translation_lmstudio.yml")
self.supported_languages = list(translation_lang["LMStudio"]["source"].keys())
self.prompt_template = prompt_config["system_prompt"]
# history config (optional)
self.history_cfg = prompt_config.get("history", {
"use_history": False,
"sources": [],
"max_messages": 0,
"max_chars": 0,
"header_template": "",
"item_template": "[{source}] {role}: {text}",
})
self._context_history: list[dict] = []
self.openai_llm = None
@@ -58,13 +70,13 @@ class LMStudioClient:
return self.base_url
def setBaseURL(self, base_url: str | None) -> None:
result = _authentication_check(api_key=self.api_key, base_url=base_url)
result = _authentication_check(base_url=base_url)
if result:
self.base_url = base_url
return result
def getModelList(self) -> list[str]:
return _get_available_text_models(api_key=self.api_key, base_url=self.base_url) if self.base_url else []
return _get_available_text_models(base_url=self.base_url) if self.base_url else []
def getModel(self) -> str:
return self.model
@@ -84,12 +96,58 @@ class LMStudioClient:
streaming=False,
)
def setContextHistory(self, history_items: list[dict]) -> None:
"""Set recent conversation history for prompt injection.
Each item should be a dict containing:
- source: "chat" | "mic" | "speaker"
- text: message string
- timestamp: ISO format datetime string
"""
self._context_history = history_items or []
def translate(self, text: str, input_lang: str, output_lang: str) -> str:
system_prompt = self.prompt_template.format(
supported_languages=self.supported_languages,
input_lang=input_lang,
output_lang=output_lang,
)
# Inject recent conversation history if enabled by YAML config
if self.history_cfg.get("use_history"):
allowed_sources = set(self.history_cfg.get("sources", []))
max_messages = int(self.history_cfg.get("max_messages", 0))
max_chars = int(self.history_cfg.get("max_chars", 0))
item_tmpl = self.history_cfg.get("item_template", "[{source}] {role}: {text}")
header_tmpl = self.history_cfg.get("header_template", "{history}")
filtered = [h for h in self._context_history if h.get("source") in allowed_sources]
recent = filtered[-max_messages:] if max_messages > 0 else filtered
formatted_items = []
for h in recent:
# Format timestamp as HH:MM to save tokens
timestamp_str = ''
if 'timestamp' in h:
from datetime import datetime
try:
ts = datetime.fromisoformat(h['timestamp'])
timestamp_str = ts.strftime('%H:%M')
except:
timestamp_str = ''
formatted_items.append(
item_tmpl.format(
timestamp=timestamp_str,
source=h.get("source", ""),
text=h.get("text", ""),
)
)
history_blob = "\n".join(formatted_items).strip()
if max_chars and len(history_blob) > max_chars:
history_blob = history_blob[-max_chars:]
history_header = header_tmpl.format(max_messages=max_messages, history=history_blob)
if history_header:
system_prompt = f"{system_prompt}\n\n{history_header}"
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": text},
@@ -108,7 +166,7 @@ class LMStudioClient:
return content.strip()
if __name__ == "__main__":
client = LMStudioClient(base_url="http://192.168.68.110:1234/v1")
client = LMStudioClient(base_url="http://127.0.0.1:1234/v1")
models = client.getModelList()
if models:
print("Available models:", models)

View File

@@ -3,19 +3,20 @@ from langchain_ollama import ChatOllama
try:
from .translation_languages import translation_lang
from .translation_utils import loadPromptConfig
from .translation_utils import loadTranslatePromptConfig
except Exception:
import sys
from os import path as os_path
sys.path.append(os_path.dirname(os_path.dirname(os_path.dirname(os_path.abspath(__file__)))))
from translation_languages import translation_lang
from translation_utils import loadPromptConfig
sys.path.append(os_path.dirname(os_path.abspath(__file__)))
from translation_languages import translation_lang, loadTranslationLanguages
from translation_utils import loadTranslatePromptConfig
translation_lang = loadTranslationLanguages(path=".", force=True)
def _authentication_check(base_url: str | None = None) -> bool:
"""Check authentication for Ollama API.
"""
try:
response = requests.get(f"{base_url}")
response = requests.get(f"{base_url}", timeout=0.2)
if response.status_code == 200:
return True
else:
@@ -47,9 +48,19 @@ class OllamaClient:
self.model = None
self.base_url = "http://localhost:11434"
prompt_config = loadPromptConfig(root_path, "translation_ollama.yml")
prompt_config = loadTranslatePromptConfig(root_path, "translation_ollama.yml")
self.supported_languages = list(translation_lang["Ollama"]["source"].keys())
self.prompt_template = prompt_config["system_prompt"]
# history config (optional)
self.history_cfg = prompt_config.get("history", {
"use_history": False,
"sources": [],
"max_messages": 0,
"max_chars": 0,
"header_template": "",
"item_template": "[{source}] {role}: {text}",
})
self._context_history: list[dict] = []
self.openai_llm = None
@@ -78,12 +89,58 @@ class OllamaClient:
streaming=False,
)
def setContextHistory(self, history_items: list[dict]) -> None:
"""Set recent conversation history for prompt injection.
Each item should be a dict containing:
- source: "chat" | "mic" | "speaker"
- text: message string
- timestamp: ISO format datetime string
"""
self._context_history = history_items or []
def translate(self, text: str, input_lang: str, output_lang: str) -> str:
system_prompt = self.prompt_template.format(
supported_languages=self.supported_languages,
input_lang=input_lang,
output_lang=output_lang,
)
# Inject recent conversation history if enabled by YAML config
if self.history_cfg.get("use_history"):
allowed_sources = set(self.history_cfg.get("sources", []))
max_messages = int(self.history_cfg.get("max_messages", 0))
max_chars = int(self.history_cfg.get("max_chars", 0))
item_tmpl = self.history_cfg.get("item_template", "[{source}] {role}: {text}")
header_tmpl = self.history_cfg.get("header_template", "{history}")
filtered = [h for h in self._context_history if h.get("source") in allowed_sources]
recent = filtered[-max_messages:] if max_messages > 0 else filtered
formatted_items = []
for h in recent:
# Format timestamp as HH:MM to save tokens
timestamp_str = ''
if 'timestamp' in h:
from datetime import datetime
try:
ts = datetime.fromisoformat(h['timestamp'])
timestamp_str = ts.strftime('%H:%M')
except:
timestamp_str = ''
formatted_items.append(
item_tmpl.format(
timestamp=timestamp_str,
source=h.get("source", ""),
text=h.get("text", ""),
)
)
history_blob = "\n".join(formatted_items).strip()
if max_chars and len(history_blob) > max_chars:
history_blob = history_blob[-max_chars:]
history_header = header_tmpl.format(max_messages=max_messages, history=history_blob)
if history_header:
system_prompt = f"{system_prompt}\n\n{history_header}"
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": text},

View File

@@ -4,13 +4,14 @@ from pydantic import SecretStr
try:
from .translation_languages import translation_lang
from .translation_utils import loadPromptConfig
from .translation_utils import loadTranslatePromptConfig
except Exception:
import sys
from os import path as os_path
sys.path.append(os_path.dirname(os_path.dirname(os_path.dirname(os_path.abspath(__file__)))))
from translation_languages import translation_lang
from translation_utils import loadPromptConfig
from translation_languages import translation_lang, loadTranslationLanguages
from translation_utils import loadTranslatePromptConfig
translation_lang = loadTranslationLanguages(path=".", force=True)
def _authentication_check(api_key: str, base_url: str | None = None) -> bool:
"""Check if the provided API key is valid by attempting to list models.
@@ -68,9 +69,19 @@ class OpenAIClient:
self.model = None
self.base_url = base_url # None の場合は公式エンドポイント
prompt_config = loadPromptConfig(root_path, "translation_openai.yml")
prompt_config = loadTranslatePromptConfig(root_path, "translation_openai.yml")
self.supported_languages = list(translation_lang["OpenAI_API"]["source"].keys())
self.prompt_template = prompt_config["system_prompt"]
# history config (optional)
self.history_cfg = prompt_config.get("history", {
"use_history": False,
"sources": [],
"max_messages": 0,
"max_chars": 0,
"header_template": "",
"item_template": "[{source}] {role}: {text}",
})
self._context_history: list[dict] = []
self.openai_llm = None
@@ -104,12 +115,62 @@ class OpenAIClient:
streaming=False,
)
def setContextHistory(self, history_items: list[dict]) -> None:
"""Set recent conversation history for prompt injection.
Each item should be a dict containing:
- source: "chat" | "mic" | "speaker"
- text: message string
- timestamp: ISO format datetime string
"""
self._context_history = history_items or []
def translate(self, text: str, input_lang: str, output_lang: str) -> str:
system_prompt = self.prompt_template.format(
supported_languages=self.supported_languages,
input_lang=input_lang,
output_lang=output_lang,
)
# Inject recent conversation history if enabled by YAML config
if self.history_cfg.get("use_history"):
allowed_sources = set(self.history_cfg.get("sources", []))
max_messages = int(self.history_cfg.get("max_messages", 0))
max_chars = int(self.history_cfg.get("max_chars", 0))
item_tmpl = self.history_cfg.get("item_template", "[{source}] {role}: {text}")
header_tmpl = self.history_cfg.get("header_template", "{history}")
# filter by source and take newest N
filtered = [h for h in self._context_history if h.get("source") in allowed_sources]
recent = filtered[-max_messages:] if max_messages > 0 else filtered
# format items
formatted_items = []
for h in recent:
# Format timestamp as HH:MM to save tokens
timestamp_str = ''
if 'timestamp' in h:
from datetime import datetime
try:
ts = datetime.fromisoformat(h['timestamp'])
timestamp_str = ts.strftime('%H:%M')
except:
timestamp_str = ''
formatted_items.append(
item_tmpl.format(
timestamp=timestamp_str,
source=h.get("source", ""),
text=h.get("text", ""),
)
)
history_blob = "\n".join(formatted_items).strip()
# truncate by char limit to mitigate token use
if max_chars and len(history_blob) > max_chars:
history_blob = history_blob[-max_chars:]
# assemble header and append to system prompt
history_header = header_tmpl.format(max_messages=max_messages, history=history_blob)
if history_header:
system_prompt = f"{system_prompt}\n\n{history_header}"
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": text},
@@ -128,7 +189,7 @@ class OpenAIClient:
return content.strip()
if __name__ == "__main__":
AUTH_KEY = "OPENAI_API_KEY"
AUTH_KEY = input("OPENAI_API_KEY: ")
client = OpenAIClient()
client.setAuthKey(AUTH_KEY)
models = client.getModelList()

View File

@@ -0,0 +1,199 @@
import requests
from openai import OpenAI
from langchain_openai import ChatOpenAI
from pydantic import SecretStr
try:
from .translation_languages import translation_lang
from .translation_utils import loadTranslatePromptConfig
except Exception:
import sys
from os import path as os_path
sys.path.append(os_path.dirname(os_path.dirname(os_path.dirname(os_path.abspath(__file__)))))
from translation_languages import translation_lang, loadTranslationLanguages
from translation_utils import loadTranslatePromptConfig
translation_lang = loadTranslationLanguages(path=".", force=True)
def _authentication_check(api_key: str) -> bool:
"""Check if the provided API key is valid by attempting to list models.
"""
url = "https://openrouter.ai/api/v1/auth/key"
headers = {
"Authorization": f"Bearer {api_key}"
}
r = requests.get(url, headers=headers, timeout=10)
return r.status_code == 200
def _get_available_text_models(api_key: str, base_url: str | None = None) -> list[str]:
"""Extract only OpenRouter models suitable for translation and chat applications.
"""
client = OpenAI(api_key=api_key, base_url=base_url)
res = client.models.list()
allowed_models = []
for model in res.data:
model_id = model.id
# 除外対象のキーワード
exclude_keywords = [
"whisper", # 音声認識
"embedding", # 埋め込み
"image", # 画像生成
"tts", # 音声合成
"audio", # 音声系
"search", # 検索補助モデル
"transcribe", # 音声→文字起こし
"diarize", # 話者分離
"vision" # 画像入力系
]
# 除外キーワードが含まれているモデルをスキップ
if any(kw in model_id.lower() for kw in exclude_keywords):
continue
# テキスト処理用モデルのみ対象
allowed_models.append(model_id)
allowed_models.sort()
return allowed_models
class OpenRouterClient:
"""OpenRouter API Translation wrapper using OpenAI-compatible endpoint.
OpenRouter provides access to various LLM models via a unified API.
The API endpoint: https://openrouter.ai/api/v1
"""
def __init__(self, root_path: str = None):
self.api_key = None
self.model = None
self.base_url = "https://openrouter.ai/api/v1"
prompt_config = loadTranslatePromptConfig(root_path, "translation_openrouter.yml")
self.supported_languages = list(translation_lang["OpenRouter_API"]["source"].keys())
self.prompt_template = prompt_config["system_prompt"]
# history config (optional)
self.history_cfg = prompt_config.get("history", {
"use_history": False,
"sources": [],
"max_messages": 0,
"max_chars": 0,
"header_template": "",
"item_template": "[{source}] {role}: {text}",
})
self._context_history: list[dict] = []
self.openrouter_llm = None
def getModelList(self) -> list[str]:
return _get_available_text_models(self.api_key, self.base_url) if self.api_key else []
def getAuthKey(self) -> str:
return self.api_key
def setAuthKey(self, api_key: str) -> bool:
result = _authentication_check(api_key)
if result:
self.api_key = api_key
return result
def getModel(self) -> str:
return self.model
def setModel(self, model: str) -> bool:
if model in self.getModelList():
self.model = model
return True
else:
return False
def updateClient(self) -> None:
self.openrouter_llm = ChatOpenAI(
base_url=self.base_url,
model=self.model,
api_key=SecretStr(self.api_key),
streaming=False,
)
def setContextHistory(self, history_items: list[dict]) -> None:
"""Set recent conversation history for prompt injection.
Each item should be a dict containing:
- source: "chat" | "mic" | "speaker"
- text: message string
- timestamp: ISO format datetime string
"""
self._context_history = history_items or []
def translate(self, text: str, input_lang: str, output_lang: str) -> str:
system_prompt = self.prompt_template.format(
supported_languages=self.supported_languages,
input_lang=input_lang,
output_lang=output_lang,
)
# Inject recent conversation history if enabled by YAML config
if self.history_cfg.get("use_history"):
allowed_sources = set(self.history_cfg.get("sources", []))
max_messages = int(self.history_cfg.get("max_messages", 0))
max_chars = int(self.history_cfg.get("max_chars", 0))
item_tmpl = self.history_cfg.get("item_template", "[{source}] {role}: {text}")
header_tmpl = self.history_cfg.get("header_template", "{history}")
filtered = [h for h in self._context_history if h.get("source") in allowed_sources]
recent = filtered[-max_messages:] if max_messages > 0 else filtered
formatted_items = []
for h in recent:
# Format timestamp as HH:MM to save tokens
timestamp_str = ''
if 'timestamp' in h:
from datetime import datetime
try:
ts = datetime.fromisoformat(h['timestamp'])
timestamp_str = ts.strftime('%H:%M')
except:
timestamp_str = ''
formatted_items.append(
item_tmpl.format(
timestamp=timestamp_str,
source=h.get("source", ""),
text=h.get("text", ""),
)
)
history_blob = "\n".join(formatted_items).strip()
if max_chars and len(history_blob) > max_chars:
history_blob = history_blob[-max_chars:]
history_header = header_tmpl.format(max_messages=max_messages, history=history_blob)
if history_header:
system_prompt = f"{system_prompt}\n\n{history_header}"
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": text},
]
resp = self.openrouter_llm.invoke(messages)
content = ""
if isinstance(resp.content, str):
content = resp.content
elif isinstance(resp.content, list):
for part in resp.content:
if isinstance(part, str):
content += part
elif isinstance(part, dict) and "content" in part and isinstance(part["content"], str):
content += part["content"]
return content.strip()
if __name__ == "__main__":
AUTH_KEY = input("OPENROUTER_API_KEY: ")
client = OpenRouterClient()
client.setAuthKey(AUTH_KEY)
models = client.getModelList()
if models:
print("Available models:", models)
model = input("Select a model: ")
client.setModel(model)
client.updateClient()
print(client.translate("こんにちは世界", "Japanese", "English"))

View File

@@ -4,13 +4,14 @@ from pydantic import SecretStr
try:
from .translation_languages import translation_lang
from .translation_utils import loadPromptConfig
from .translation_utils import loadTranslatePromptConfig
except Exception:
import sys
from os import path as os_path
sys.path.append(os_path.dirname(os_path.dirname(os_path.dirname(os_path.abspath(__file__)))))
from translation_languages import translation_lang
from translation_utils import loadPromptConfig
from translation_languages import translation_lang, loadTranslationLanguages
from translation_utils import loadTranslatePromptConfig
translation_lang = loadTranslationLanguages(path=".", force=True)
BASE_URL = "https://api.platform.preferredai.jp/v1"
@@ -43,9 +44,19 @@ class PlamoClient:
self.base_url = BASE_URL
self.model = None
prompt_config = loadPromptConfig(root_path, "translation_plamo.yml")
prompt_config = loadTranslatePromptConfig(root_path, "translation_plamo.yml")
self.supported_languages = list(translation_lang["Plamo_API"]["source"].keys())
self.prompt_template = prompt_config["system_prompt"]
# history config (optional)
self.history_cfg = prompt_config.get("history", {
"use_history": False,
"sources": [],
"max_messages": 0,
"max_chars": 0,
"header_template": "",
"item_template": "[{source}] {role}: {text}",
})
self._context_history: list[dict] = []
self.plamo_llm = None
@@ -79,12 +90,58 @@ class PlamoClient:
api_key=SecretStr(self.api_key),
)
def setContextHistory(self, history_items: list[dict]) -> None:
"""Set recent conversation history for prompt injection.
Each item should be a dict containing:
- source: "chat" | "mic" | "speaker"
- text: message string
- timestamp: ISO format datetime string
"""
self._context_history = history_items or []
def translate(self, text: str, input_lang: str, output_lang: str) -> str:
system_prompt = self.prompt_template.format(
supported_languages=self.supported_languages,
input_lang=input_lang,
output_lang=output_lang
)
# Inject recent conversation history if enabled by YAML config
if self.history_cfg.get("use_history"):
allowed_sources = set(self.history_cfg.get("sources", []))
max_messages = int(self.history_cfg.get("max_messages", 0))
max_chars = int(self.history_cfg.get("max_chars", 0))
item_tmpl = self.history_cfg.get("item_template", "[{source}] {role}: {text}")
header_tmpl = self.history_cfg.get("header_template", "{history}")
filtered = [h for h in self._context_history if h.get("source") in allowed_sources]
recent = filtered[-max_messages:] if max_messages > 0 else filtered
formatted_items = []
for h in recent:
# Format timestamp as HH:MM to save tokens
timestamp_str = ''
if 'timestamp' in h:
from datetime import datetime
try:
ts = datetime.fromisoformat(h['timestamp'])
timestamp_str = ts.strftime('%H:%M')
except:
timestamp_str = ''
formatted_items.append(
item_tmpl.format(
timestamp=timestamp_str,
source=h.get("source", ""),
text=h.get("text", ""),
)
)
history_blob = "\n".join(formatted_items).strip()
if max_chars and len(history_blob) > max_chars:
history_blob = history_blob[-max_chars:]
history_header = header_tmpl.format(max_messages=max_messages, history=history_blob)
if history_header:
system_prompt = f"{system_prompt}\n\n{history_header}"
messages = [
{"role": "system", "content": system_prompt},
{"role": "user", "content": text},

View File

@@ -769,3 +769,11 @@ LMStudio:
Ollama:
source: *openai_langs
target: *openai_langs
Groq_API:
source: *openai_langs
target: *openai_langs
OpenRouter_API:
source: *openai_langs
target: *openai_langs

View File

@@ -0,0 +1,16 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.
history:
use_history: true
sources: [chat, mic, speaker]
max_messages: 5
max_chars: 4000
header_template: |
Conversation context (recent {max_messages} messages):
{history}
item_template: "[{timestamp}][{source}] {text}"

View File

@@ -0,0 +1,16 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.
history:
use_history: true
sources: [chat, mic, speaker]
max_messages: 5
max_chars: 4000
header_template: |
Conversation context (recent {max_messages} messages):
{history}
item_template: "[{timestamp}][{source}] {text}"

View File

@@ -0,0 +1,16 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.
history:
use_history: true
sources: [chat, mic, speaker]
max_messages: 5
max_chars: 4000
header_template: |
Conversation context (recent {max_messages} messages):
{history}
item_template: "[{timestamp}][{source}] {text}"

View File

@@ -0,0 +1,16 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.
history:
use_history: true
sources: [chat, mic, speaker]
max_messages: 5
max_chars: 4000
header_template: |
Conversation context (recent {max_messages} messages):
{history}
item_template: "[{timestamp}][{source}] {text}"

View File

@@ -0,0 +1,16 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.
history:
use_history: true
sources: [chat, mic, speaker] # 取り込み対象の履歴種別
max_messages: 5 # 注入する履歴件数の上限(新しい順)
max_chars: 4000 # 履歴整形後の最大文字数(超過時は先頭を切り捨て)
header_template: |
Conversation context (recent {max_messages} messages):
{history}
item_template: "[{timestamp}][{source}] {text}"

View File

@@ -0,0 +1,16 @@
system_prompt: |
You are a helpful translation assistant.
Supported languages:
{supported_languages}
Translate the user provided text from {input_lang} to {output_lang}.
Return ONLY the translated text. Do not add quotes or extra commentary.
history:
use_history: true
sources: [chat, mic, speaker]
max_messages: 5
max_chars: 4000
header_template: |
Conversation context (recent {max_messages} messages):
{history}
item_template: "[{timestamp}][{source}] {text}"

View File

@@ -4,4 +4,13 @@ system_prompt: |
{supported_languages}
Translate the following text from {input_lang} to {output_lang}.
output only the translated text without any additional commentary.
output only the translated text without any additional commentary.
history:
use_history: true
sources: [chat, mic, speaker]
max_messages: 5
max_chars: 4000
header_template: |
Conversation context (recent {max_messages} messages):
{history}
item_template: "[{timestamp}][{source}] {text}"

View File

@@ -15,6 +15,8 @@ try:
from .translation_openai import OpenAIClient
from .translation_lmstudio import LMStudioClient
from .translation_ollama import OllamaClient
from .translation_groq import GroqClient
from .translation_openrouter import OpenRouterClient
except Exception:
import sys
sys.path.append(os_path.dirname(os_path.dirname(os_path.dirname(os_path.abspath(__file__)))))
@@ -25,6 +27,8 @@ except Exception:
from translation_openai import OpenAIClient
from translation_lmstudio import LMStudioClient
from translation_ollama import OllamaClient
from translation_groq import GroqClient
from translation_openrouter import OpenRouterClient
import ctranslate2
import transformers
@@ -50,6 +54,8 @@ class Translator:
self.plamo_client: Optional[PlamoClient] = None
self.gemini_client: Optional[GeminiClient] = None
self.openai_client: Optional[OpenAIClient] = None
self.groq_client: Optional[GroqClient] = None
self.openrouter_client: Optional[OpenRouterClient] = None
self.lmstudio_client: LMStudioClient[LMStudioClient] = None
self.ollama_client: OllamaClient[OllamaClient] = None
self.ctranslate2_translator: Any = None
@@ -176,6 +182,84 @@ class Translator:
"""Update the OpenAI client (fetch available models)."""
self.openai_client.updateClient()
def authenticationGroqAuthKey(self, auth_key: str, root_path: str = None) -> bool:
"""Authenticate Groq API with the provided key.
Returns True on success, False on failure.
"""
self.groq_client = GroqClient(root_path=root_path)
if self.groq_client.setAuthKey(auth_key):
return True
else:
self.groq_client = None
return False
def getGroqModelList(self) -> list[str]:
"""Get available Groq models.
Returns a list of model names, or an empty list on failure.
"""
if self.groq_client is None:
return []
return self.groq_client.getModelList()
def setGroqModel(self, model: str) -> bool:
"""Change the Groq model used for translation.
Returns True on success, False on failure.
"""
if self.groq_client is None:
return False
return self.groq_client.setModel(model)
def updateGroqClient(self) -> None:
"""Update the Groq client (fetch available models)."""
self.groq_client.updateClient()
def authenticationOpenRouterAuthKey(self, auth_key: str, root_path: str = None) -> bool:
"""Authenticate OpenRouter API with the provided key.
Returns True on success, False on failure.
"""
self.openrouter_client = OpenRouterClient(root_path=root_path)
if self.openrouter_client.setAuthKey(auth_key):
return True
else:
self.openrouter_client = None
return False
def getOpenRouterModelList(self) -> list[str]:
"""Get available OpenRouter models.
Returns a list of model names, or an empty list on failure.
"""
if self.openrouter_client is None:
return []
return self.openrouter_client.getModelList()
def setOpenRouterModel(self, model: str) -> bool:
"""Change the OpenRouter model used for translation.
Returns True on success, False on failure.
"""
if self.openrouter_client is None:
return False
return self.openrouter_client.setModel(model)
def updateOpenRouterClient(self) -> None:
"""Update the OpenRouter client (fetch available models)."""
self.openrouter_client.updateClient()
def getLMStudioConnected(self) -> bool:
"""Get LM Studio connection status.
Returns True if connected, False otherwise.
"""
if self.lmstudio_client is None:
return False
else:
return True
def setLMStudioClientURL(self, base_url: str | None = None, root_path: str = None) -> bool:
"""Authenticate LM Studio with the provided base URL.
@@ -207,13 +291,26 @@ class Translator:
"""Update the LM Studio client (fetch available models)."""
self.lmstudio_client.updateClient()
def getOllamaConnected(self) -> bool:
"""Get Ollama connection status.
Returns True if connected, False otherwise.
"""
if self.ollama_client is None:
return False
else:
return True
def checkOllamaClient(self, root_path: str = None) -> bool:
"""Check if Ollama client is available.
Returns True if Ollama is reachable, False otherwise.
"""
self.ollama_client = OllamaClient(root_path=root_path)
return self.ollama_client.authenticationCheck()
result = self.ollama_client.authenticationCheck()
if result is False:
self.ollama_client = None
return result
def getOllamaModelList(self, root_path: str = None) -> bool:
"""Initialize Ollama client and fetch available models.
@@ -328,9 +425,18 @@ class Translator:
target_language = translation_lang[translator_name]["target"][target_language]
return source_language, target_language
def translate(self, translator_name: str, weight_type: str, source_language: str, target_language: str, target_country: str, message: str) -> Any:
def translate(self, translator_name: str, weight_type: str, source_language: str, target_language: str, target_country: str, message: str, context_history: Optional[list[dict]] = None) -> Any:
"""Translate `message` using the named translator backend.
Args:
translator_name: Name of the translator backend to use
weight_type: Model weight type for CTranslate2
source_language: Source language name
target_language: Target language name
target_country: Target country for locale-specific translations
message: Text to translate
context_history: Optional conversation context (Chat/Mic/Speaker messages)
Returns translated string on success, or False on failure. When
source_language == target_language the original message is returned.
"""
@@ -363,6 +469,8 @@ class Translator:
if self.plamo_client is None:
result = False
else:
if context_history:
self.plamo_client.setContextHistory(context_history)
result = self.plamo_client.translate(
message,
input_lang=source_language,
@@ -372,6 +480,8 @@ class Translator:
if self.gemini_client is None:
result = False
else:
if context_history:
self.gemini_client.setContextHistory(context_history)
result = self.gemini_client.translate(
message,
input_lang=source_language,
@@ -381,15 +491,41 @@ class Translator:
if self.openai_client is None:
result = False
else:
if context_history:
self.openai_client.setContextHistory(context_history)
result = self.openai_client.translate(
message,
input_lang=source_language,
output_lang=target_language,
)
case "Groq_API":
if self.groq_client is None:
result = False
else:
if context_history:
self.groq_client.setContextHistory(context_history)
result = self.groq_client.translate(
message,
input_lang=source_language,
output_lang=target_language,
)
case "OpenRouter_API":
if self.openrouter_client is None:
result = False
else:
if context_history:
self.openrouter_client.setContextHistory(context_history)
result = self.openrouter_client.translate(
message,
input_lang=source_language,
output_lang=target_language,
)
case "LMStudio":
if self.lmstudio_client is None:
result = False
else:
if context_history:
self.lmstudio_client.setContextHistory(context_history)
result = self.lmstudio_client.translate(
message,
input_lang=source_language,
@@ -399,6 +535,8 @@ class Translator:
if self.ollama_client is None:
result = False
else:
if context_history:
self.ollama_client.setContextHistory(context_history)
result = self.ollama_client.translate(
message,
input_lang=source_language,

View File

@@ -101,16 +101,16 @@ def downloadCTranslate2Tokenizer(path: str, weight_type: str = "m2m100_418M-ct2-
tokenizer_path = os_path.join("./weights", "ctranslate2", directory_name, "tokenizer")
transformers.AutoTokenizer.from_pretrained(tokenizer, cache_dir=tokenizer_path)
def loadPromptConfig(root_path: str | None = None, prompt_filename: str | None = None) -> dict:
def loadTranslatePromptConfig(root_path: str | None = None, prompt_filename: str | None = None) -> dict:
# PyInstaller 展開後
if root_path and prompt_filename and os_path.exists(os_path.join(root_path, "_internal", "prompt", prompt_filename)):
prompt_path = os_path.join(root_path, "_internal", "prompt", prompt_filename)
if root_path and prompt_filename and os_path.exists(os_path.join(root_path, "_internal", "translation_settings", "prompt", prompt_filename)):
prompt_path = os_path.join(root_path, "_internal", "translation_settings", "prompt", prompt_filename)
# src-python 直下実行
elif prompt_filename and os_path.exists(os_path.join(os_path.dirname(__file__), "models", "translation", "prompt", prompt_filename)):
prompt_path = os_path.join(os_path.dirname(__file__), "models", "translation", "prompt", prompt_filename)
elif prompt_filename and os_path.exists(os_path.join(os_path.dirname(__file__), "models", "translation", "translation_settings", "prompt", prompt_filename)):
prompt_path = os_path.join(os_path.dirname(__file__), "models", "translation", "translation_settings", "prompt", prompt_filename)
# translation フォルダ直下実行
elif prompt_filename and os_path.exists(os_path.join(os_path.dirname(__file__), "prompt", prompt_filename)):
prompt_path = os_path.join(os_path.dirname(__file__), "prompt", prompt_filename)
elif prompt_filename and os_path.exists(os_path.join(os_path.dirname(__file__), "translation_settings", "prompt", prompt_filename)):
prompt_path = os_path.join(os_path.dirname(__file__), "translation_settings", "prompt", prompt_filename)
else:
raise FileNotFoundError(f"Prompt file not found: {prompt_filename}")
with open(prompt_path, "r", encoding="utf-8") as f: