Release v2.1.1: Token Handling Improvement and DeepSeek Support

- Completely reworked token handling mechanism
- Removed custom token calculation logic
- Direct max_tokens passing to LLM APIs
- Added support for DeepSeek provider
- Integrated deepseek-chat and deepseek-reasoner models

Thanks to @estiens for reporting token handling issues and providing valuable feedback (https://github.com/smkrv/ha-text-ai/issues/1).

custom_components/ha_text_ai/__init__.py

@@ -39,11 +39,14 @@ from .const import (
     CONF_CONTEXT_MESSAGES,
     API_PROVIDER_OPENAI,
     API_PROVIDER_ANTHROPIC,
+    API_PROVIDER_DEEPSEEK,
     DEFAULT_MODEL,
+    DEFAULT_DEEPSEEK_MODEL,
     DEFAULT_TEMPERATURE,
     DEFAULT_MAX_TOKENS,
     DEFAULT_OPENAI_ENDPOINT,
     DEFAULT_ANTHROPIC_ENDPOINT,
+    DEFAULT_DEEPSEEK_ENDPOINT,
     DEFAULT_REQUEST_INTERVAL,
     DEFAULT_CONTEXT_MESSAGES,
     API_TIMEOUT,
@@ -235,6 +238,8 @@ async def async_check_api(session, endpoint: str, headers: dict, provider: str)
     try:
         if provider == API_PROVIDER_ANTHROPIC:
             check_url = f"{endpoint}/v1/models"
+        elif provider == API_PROVIDER_DEEPSEEK:
+            check_url = f"{endpoint}/models"  # DeepSeek
         else:  # OpenAI
             check_url = f"{endpoint}/models"