feat: add support for gemini llm via langchain

flowsettings.py

@@ -184,14 +184,17 @@
     },
     "default": False,
 }
-# KH_LLMS["gemini"] = {
-#     "spec": {
-#         "__type__": "kotaemon.llms.chats.LCGeminiChat",
-#         "model_name": "gemini-1.5-pro",
-#         "api_key": "your-key",
-#     },
-#     "default": False,
-# }
+
+KH_LLMS["gemini"] = {
+    "spec": {
+        "__type__": "kotaemon.llms.chats.LCChatGemini",
+        "model": "models/gemini-1.5-flash",
+        "google_api_key": "your-key",
+        "temperature": 0.7,
+    },
+    "default": False,
+}
+
 KH_LLMS["groq"] = {
     "spec": {
         "__type__": "kotaemon.llms.ChatOpenAI",
@@ -211,6 +214,7 @@
     },
     "default": False,
 }
+
 # KH_EMBEDDINGS["huggingface"] = {
 #     "spec": {
 #         "__type__": "kotaemon.embeddings.LCHuggingFaceEmbeddings",

libs/kotaemon/kotaemon/llms/__init__.py

@@ -9,8 +9,8 @@
     EndpointChatLLM,
     LCAnthropicChat,
     LCAzureChatOpenAI,
+    LCChatGemini,
     LCChatOpenAI,
-    LCGeminiChat,
     LlamaCppChat,
 )
 from .completions import LLM, AzureOpenAI, LlamaCpp, OpenAI
@@ -30,7 +30,7 @@
     "AzureChatOpenAI",
     "ChatOpenAI",
     "LCAnthropicChat",
-    "LCGeminiChat",
+    "LCChatGemini",
     "LCAzureChatOpenAI",
     "LCChatOpenAI",
     "LlamaCppChat",

libs/kotaemon/kotaemon/llms/chats/__init__.py

@@ -3,9 +3,9 @@
 from .langchain_based import (
     LCAnthropicChat,
     LCAzureChatOpenAI,
+    LCChatGemini,
     LCChatMixin,
     LCChatOpenAI,
-    LCGeminiChat,
 )
 from .llamacpp import LlamaCppChat
 from .openai import AzureChatOpenAI, ChatOpenAI
@@ -17,7 +17,7 @@
     "EndpointChatLLM",
     "ChatOpenAI",
     "LCAnthropicChat",
-    "LCGeminiChat",
+    "LCChatGemini",
     "LCChatOpenAI",
     "LCAzureChatOpenAI",
     "LCChatMixin",

libs/kotaemon/kotaemon/llms/chats/langchain_based.py

@@ -4,6 +4,7 @@
 from typing import AsyncGenerator, Iterator
 
 from kotaemon.base import BaseMessage, HumanMessage, LLMInterface
+from kotaemon.platforms.gemini import BaseGeminiChatModel
 
 from .base import ChatLLM
 
@@ -247,18 +248,23 @@ def _get_lc_class(self):
         return ChatAnthropic
 
 
-class LCGeminiChat(LCChatMixin, ChatLLM):  # type: ignore
+class LCChatGemini(LCChatMixin, ChatLLM, BaseGeminiChatModel):  # type: ignore
+    """Gemini Chat Model
+
+    https://python.langchain.com/v0.2/api_reference/google_genai/chat_models/langchain_google_genai.chat_models.ChatGoogleGenerativeAI.html
+    """
+
     def __init__(
         self,
-        api_key: str | None = None,
-        model_name: str | None = None,
+        google_api_key: str | None = None,
+        model: str | None = None,
         temperature: float = 0.7,
         **params,
     ):
         super().__init__(
-            google_api_key=api_key,
-            model=model_name,
-            temperature=temperature,
+            google_api_key=google_api_key or self.google_api_key,
+            model=model or self.model,
+            temperature=temperature or self.temperature,
             **params,
         )
 

libs/kotaemon/kotaemon/platforms/gemini.py

@@ -0,0 +1,61 @@
+from typing import Optional
+
+from kotaemon.base import Param
+
+
+def get_model(type):
+    types = {"embedding": "text-embedding-004", "chat": "gemini-1.5-flash"}
+    return Param(
+        f"models/{types.get(type)}",
+        help=(
+            f"The name of the required {type} model "
+            "(https://ai.google.dev/gemini-api/docs/models/gemini). "
+            "Must be in the format 'model/...'."
+        ),
+    )
+
+
+class BaseGeminiModel:
+    google_api_key: str = Param("", help="The API key received from Gemini.")
+
+
+class BaseGeminiChatModel(BaseGeminiModel):
+    model: str = get_model("chat")
+
+    temperature: Optional[float] = Param(
+        0.7,
+        help="Run inference with this temperature. Must be between 0.0 and 1.0.",
+    )
+
+    top_k: Optional[float] = Param(
+        None,
+        help=(
+            "Decode using top-k sampling by considering the set of "
+            "top_k most probable tokens. Must be larger than 0."
+        ),
+    )
+
+    top_p: Optional[float] = Param(
+        None,
+        help=(
+            "Decode using nucleus samplingby considering the smallest set of "
+            "tokens whose probability sum is at least top_p. "
+            "Must be between 0.0 and 1.0."
+        ),
+    )
+
+    max_output_tokens: Optional[float] = Param(
+        None,
+        help=(
+            "Maximum number of tokens to include in a candidate. "
+            "Must be greater than zero. If unset, will default to 64."
+        ),
+    )
+
+    max_retries: Optional[float] = Param(
+        None, help="The maximum number of retries to make when generating."
+    )
+
+    timeout: Optional[float] = Param(
+        None, help="The maximum number of seconds to wait for a response."
+    )

libs/ktem/ktem/llms/manager.py

@@ -58,15 +58,15 @@ def load_vendors(self):
             AzureChatOpenAI,
             ChatOpenAI,
             LCAnthropicChat,
-            LCGeminiChat,
+            LCChatGemini,
             LlamaCppChat,
         )
 
         self._vendors = [
             ChatOpenAI,
             AzureChatOpenAI,
             LCAnthropicChat,
-            LCGeminiChat,
+            LCChatGemini,
             LlamaCppChat,
         ]