Integrate deepseek (#105)

.env.dev

@@ -30,6 +30,10 @@ OPENAI_API_KEY=
 XAI_MODEL=grok-2-latest
 XAI_API_KEY=
 
+DEEPSEEK_API_KEY=
+DEEPSEEK_MODEL=deepseek-chat
+DEEPSEEK_API_BASE_URL=https://api.deepseek.com
+
 # === Third-party services settings ===
 
 # Perplexity

src/core/config.py

@@ -62,6 +62,11 @@ class Settings(BaseSettings):
     XAI_API_KEY: str = ""
     XAI_MODEL: str = "grok-2-latest"
 
+    #: DeepSeek
+    DEEPSEEK_API_KEY: str = ""
+    DEEPSEEK_MODEL: str = "deepseek-reasoner"
+    DEEPSEEK_API_BASE_URL: str = "https://api.deepseek.com"
+
     #: Llama
     LLAMA_PROVIDER: LlamaProviderType = LlamaProviderType.OLLAMA
     LLAMA_MODEL_NAME: str = "llama3-8b-8192"  # Model name is usually unique for each provider

src/core/defs.py

@@ -67,12 +67,13 @@ class LLMProviderType(str, Enum):
     ANTHROPIC = "anthropic"
     XAI = "xai"
     LLAMA = "llama"
+    DEEPSEEK = "deepseek"
 
 
 class LlamaProviderType(str, Enum):
     """Llama provider type."""
 
     OLLAMA = "ollama"
     FIREWORKS = "fireworks"
-    LLAMA_API = "llama-api"
+    LLAMA_API = "llama_api"
     OPENROUTER = "openrouter"

src/llm/llm.py

@@ -7,6 +7,7 @@
 from src.core.defs import LLMProviderType
 from src.core.exceptions import LLMError
 from src.llm.providers.anthropic import call_anthropic
+from src.llm.providers.deepseek import call_deepseek
 from src.llm.providers.llama import call_llama
 from src.llm.providers.oai import call_openai
 from src.llm.providers.xai import call_xai
@@ -20,7 +21,7 @@ class LLM:
     def __init__(self):
         """
         Initialize the LLM class based on the selected provider from settings.
-        Supported providers: 'openai', 'anthropic', 'xai'
+        Supported providers: 'openai', 'anthropic', 'xai', 'llama', 'deepseek'
         """
         self.provider = settings.LLM_PROVIDER
         logger.debug(f"Using LLM provider: {self.provider}")
@@ -52,6 +53,8 @@ async def generate_response(self, messages: List[Dict[str, Any]], **kwargs) -> s
             return await call_xai(messages, **kwargs)
         elif self.provider == LLMProviderType.LLAMA:
             return await call_llama(messages, **kwargs)
+        elif self.provider == LLMProviderType.DEEPSEEK:
+            return await call_deepseek(messages, **kwargs)
         else:
             raise LLMError(f"Unknown LLM provider: {self.provider}")
 

src/llm/providers/deepseek.py

@@ -0,0 +1,44 @@
+from typing import Dict, List
+
+from loguru import logger
+from openai import OpenAI
+
+from src.core.config import settings
+from src.core.exceptions import LLMError
+
+
+async def call_deepseek(messages: List[Dict[str, str]], **kwargs) -> str:
+    """
+    Call the DeepSeek API endpoint.
+
+    Args:
+        messages: A list of dicts with 'role' and 'content'.
+        kwargs: Additional parameters (e.g., model, temperature).
+
+    Returns:
+        str: Response content from DeepSeek.
+    """
+    model = kwargs.get("model", settings.DEEPSEEK_MODEL)
+    temperature = kwargs.get("temperature", 0.7)
+
+    client = OpenAI(api_key=settings.DEEPSEEK_API_KEY, base_url=settings.DEEPSEEK_API_BASE_URL)
+
+    logger.debug(
+        f"Calling DeepSeek with model={model}, temperature={temperature}, messages={messages}"
+    )
+
+    try:
+        response = client.chat.completions.create(
+            model=model,
+            messages=messages,  # type: ignore
+            temperature=temperature,
+        )
+        if not response.choices[0].message.content:
+            raise LLMError("No content in DeepSeek response")
+
+        content = response.choices[0].message.content.strip()
+        logger.debug(f"DeepSeek response: {content}")
+        return content
+    except Exception as e:
+        logger.error(f"DeepSeek call failed: {e}")
+        raise LLMError("Error during DeepSeek API call") from e