feat: add Novita AI as LLM provider

Add Novita AI integration with OpenAI-compatible API support:
- Add NovitaChat class using OpenAI SDK
- Add resolve_novita_base_url and resolve_novita_api_key settings
- Add 'novita' to CLI --llm choices (ask and react commands)
- Add provider tests following minimax pattern

Default model: moonshotai/kimi-k2.5
Base URL: https://api.novita.ai/openai
API key env var: NOVITA_API_KEY

Author: Alex-yang00

packages/leann-core/src/leann/chat.py

@@ -17,6 +17,8 @@
     resolve_anthropic_base_url,
     resolve_minimax_api_key,
     resolve_minimax_base_url,
+    resolve_novita_api_key,
+    resolve_novita_base_url,
     resolve_ollama_host,
     resolve_openai_api_key,
     resolve_openai_base_url,
@@ -1010,6 +1012,73 @@ def ask(self, prompt: str, **kwargs) -> str:
             return f"Error: Could not get a response from MiniMax. Details: {e}"
 
 
+class NovitaChat(LLMInterface):
+    """LLM interface for Novita AI models via the OpenAI-compatible API.
+
+    Supported models (from Novita AI catalog):
+        - moonshotai/kimi-k2.5 (default): 262K context, MoE architecture
+        - zai-org/glm-5: 202K context, MoE with function calling
+        - minimax/minimax-m2.5: 204K context, MoE with reasoning
+
+    All models support function calling, structured output, and reasoning.
+    """
+
+    def __init__(
+        self,
+        model: str = "moonshotai/kimi-k2.5",
+        api_key: Optional[str] = None,
+        base_url: Optional[str] = None,
+    ):
+        self.model = model
+        self.base_url = resolve_novita_base_url(base_url)
+        self.api_key = resolve_novita_api_key(api_key)
+
+        if not self.api_key:
+            raise ValueError(
+                "Novita AI API key is required. Set NOVITA_API_KEY environment variable or pass api_key parameter."
+            )
+
+        logger.info(
+            "Initializing Novita Chat with model='%s' and base_url='%s'",
+            model,
+            self.base_url,
+        )
+
+        try:
+            import openai
+
+            self.client = openai.OpenAI(api_key=self.api_key, base_url=self.base_url)
+        except ImportError:
+            raise ImportError(
+                "The 'openai' library is required for Novita AI models. Please install it with 'pip install openai'."
+            )
+
+    def ask(self, prompt: str, **kwargs) -> str:
+        params = {
+            "model": self.model,
+            "messages": [{"role": "user", "content": prompt}],
+            "temperature": kwargs.get("temperature", 0.7),
+            "max_tokens": kwargs.get("max_tokens", 1000),
+        }
+
+        if "top_p" in kwargs:
+            params["top_p"] = kwargs["top_p"]
+
+        logger.info(f"Sending request to Novita AI with model {self.model}")
+
+        try:
+            response = cast(Any, self.client.chat.completions).create(**params)
+            print(
+                f"Total tokens = {response.usage.total_tokens}, prompt tokens = {response.usage.prompt_tokens}, completion tokens = {response.usage.completion_tokens}"
+            )
+            if response.choices[0].finish_reason == "length":
+                print("The query is exceeding the maximum allowed number of tokens")
+            return response.choices[0].message.content.strip()
+        except Exception as e:
+            logger.error(f"Error communicating with Novita AI: {e}")
+            return f"Error: Could not get a response from Novita AI. Details: {e}"
+
+
 class SimulatedChat(LLMInterface):
     """A simple simulated chat for testing and development."""
 
@@ -1074,6 +1143,12 @@ def get_llm(llm_config: Optional[dict[str, Any]] = None) -> LLMInterface:
             api_key=llm_config.get("api_key"),
             base_url=llm_config.get("base_url"),
         )
+    elif llm_type == "novita":
+        return NovitaChat(
+            model=model or "moonshotai/kimi-k2.5",
+            api_key=llm_config.get("api_key"),
+            base_url=llm_config.get("base_url"),
+        )
     elif llm_type == "simulated":
         return SimulatedChat()
     else:

packages/leann-core/src/leann/cli.py

@@ -24,6 +24,8 @@
     resolve_anthropic_base_url,
     resolve_minimax_api_key,
     resolve_minimax_base_url,
+    resolve_novita_api_key,
+    resolve_novita_base_url,
     resolve_ollama_host,
     resolve_openai_api_key,
     resolve_openai_base_url,
@@ -495,7 +497,7 @@ def create_parser(self) -> argparse.ArgumentParser:
             "--llm",
             type=str,
             default="ollama",
-            choices=["simulated", "ollama", "hf", "openai", "anthropic", "minimax"],
+            choices=["simulated", "ollama", "hf", "openai", "anthropic", "minimax", "novita"],
             help="LLM provider (default: ollama)",
         )
         ask_parser.add_argument(
@@ -558,7 +560,7 @@ def create_parser(self) -> argparse.ArgumentParser:
             "--llm",
             type=str,
             default="ollama",
-            choices=["simulated", "ollama", "hf", "openai", "anthropic", "minimax"],
+            choices=["simulated", "ollama", "hf", "openai", "anthropic", "minimax", "novita"],
             help="LLM provider (default: ollama)",
         )
         react_parser.add_argument(
@@ -2719,6 +2721,11 @@ async def ask_questions(self, args):
             resolved_api_key = resolve_minimax_api_key(args.api_key)
             if resolved_api_key:
                 llm_config["api_key"] = resolved_api_key
+        elif args.llm == "novita":
+            llm_config["base_url"] = resolve_novita_base_url(args.api_base)
+            resolved_api_key = resolve_novita_api_key(args.api_key)
+            if resolved_api_key:
+                llm_config["api_key"] = resolved_api_key
 
         chat = LeannChat(index_path=index_path, llm_config=llm_config)
 
@@ -2816,6 +2823,11 @@ async def react_agent(self, args):
             resolved_api_key = resolve_minimax_api_key(args.api_key)
             if resolved_api_key:
                 llm_config["api_key"] = resolved_api_key
+        elif args.llm == "novita":
+            llm_config["base_url"] = resolve_novita_base_url(args.api_base)
+            resolved_api_key = resolve_novita_api_key(args.api_key)
+            if resolved_api_key:
+                llm_config["api_key"] = resolved_api_key
 
         from .react_agent import create_react_agent
 

packages/leann-core/src/leann/settings.py

@@ -11,6 +11,7 @@
 _DEFAULT_OPENAI_BASE_URL = "https://api.openai.com/v1"
 _DEFAULT_ANTHROPIC_BASE_URL = "https://api.anthropic.com"
 _DEFAULT_MINIMAX_BASE_URL = "https://api.minimax.io/v1"
+_DEFAULT_NOVITA_BASE_URL = "https://api.novita.ai/openai"
 
 
 def _clean_url(value: str) -> str:
@@ -114,6 +115,33 @@ def resolve_minimax_api_key(explicit: str | None = None) -> str | None:
     return os.getenv("MINIMAX_API_KEY")
 
 
+def resolve_novita_base_url(explicit: str | None = None) -> str:
+    """Resolve the base URL for Novita AI services (OpenAI-compatible)."""
+
+    candidates = (
+        explicit,
+        os.getenv("LEANN_NOVITA_BASE_URL"),
+        os.getenv("NOVITA_BASE_URL"),
+        os.getenv("OPENAI_BASE_URL"),
+        os.getenv("OPENAI_API_BASE"),
+    )
+
+    for candidate in candidates:
+        if candidate:
+            return _clean_url(candidate)
+
+    return _clean_url(_DEFAULT_NOVITA_BASE_URL)
+
+
+def resolve_novita_api_key(explicit: str | None = None) -> str | None:
+    """Resolve the API key for Novita AI services."""
+
+    if explicit:
+        return explicit
+
+    return os.getenv("NOVITA_API_KEY")
+
+
 def encode_provider_options(options: dict[str, Any] | None) -> str | None:
     """Serialize provider options for child processes."""