Leolivier/issue3

.env.example

@@ -30,6 +30,8 @@ CONTEXT_LANGUAGE=French
 OPENAI_API_KEY=[your API key goes here]
 # for Claude models, set:
 ANTHROPIC_API_KEY=[your API key goes here]
+# for Grok models, set:
+XAI_API_KEY=[your API key goes here]
 # for Gemini models, set:
 GEMINI_API_KEY=[your API key goes here]
 

pyproject.toml

@@ -18,6 +18,7 @@ dependencies=[
   "langcodes>=3.4.0",
   "openai>=1.12.0",
   "anthropic>=0.34.1",
+  "xai-sdk>=0.3.0",
   "google-generativeai>=0.7.2",
 ]
 classifiers = [

src/auto_po_lyglot/clients/grok_client.py

@@ -0,0 +1,21 @@
+import xai_sdk
+import asyncio
+from .client_base import TranspoClient
+import logging
+
+logger = logging.getLogger(__name__)
+
+
+class GrokClient(TranspoClient):
+  def __init__(self, params, target_language=None):
+    params.model = params.model or ""  # default model given by Grok itself if not provided
+    super().__init__(params, target_language)
+    self.client = xai_sdk.Client(api_key=params.xai_api_key) if hasattr(params, 'xai_api_key') else xai_sdk.Client()
+
+  async def async_get_translation(self, system_prompt, user_prompt):
+    conversation = self.client.chat.create_conversation()
+    response = await conversation.add_response_no_stream(f'{system_prompt}\n{user_prompt}\n')
+    return response.message
+
+  def get_translation(self, system_prompt, user_prompt):
+    return asyncio.run(self.async_get_translation(system_prompt, user_prompt))

src/auto_po_lyglot/getenv.py

@@ -60,7 +60,7 @@ def parse_args(self, additional_args=None):
                         help='Le type of LLM you want to use. Can be openai, ollama, claude or claude_cached. '
                              'For openai or claude[_cached], you need to set the api key in the environment. '
                              'Supersedes LLM_CLIENT in .env. Default is ollama',
-                        choices=['openai', 'ollama', 'claude', 'claude_cached', 'gemini'])
+                        choices=['openai', 'ollama', 'claude', 'claude_cached', 'gemini', 'grok'])
     parser.add_argument('-m', '--model',
                         type=str,
                         help='the name of the model to use. Supersedes LLM_MODEL in .env. If not provided at all, '
@@ -184,6 +184,8 @@ def get_client(self):
           from .clients.claude_client import CachedClaudeClient as LLMClient
         case 'gemini':
           from .clients.gemini_client import GeminiClient as LLMClient
+        case 'grok':
+          from .clients.grok_client import GrokClient as LLMClient
         case _:
           raise Exception(
             f"LLM_CLIENT must be one of 'ollama', 'openai', 'claude' or 'claude_cached', not '{self.llm_client}'"

src/auto_po_lyglot/po_main.py

@@ -118,7 +118,7 @@ def main():
             sleep(1.0)  # Sleep for 1 second to avoid rate limiting
             nb_translations += 1
       except Exception as e:
-        logger.info(f"Error: {e}")
+        logger.error(f"Error: {e}")
       # Save the new .po file even if there was an error to not lose what was translated
       po.save(output_file)
       percent_translated = round(nb_translations / len(po) * 100, 2)