diff --git a/servers/fastapi/api/main.py b/servers/fastapi/api/main.py
index f5f4a8d3..80eea709 100644
--- a/servers/fastapi/api/main.py
+++ b/servers/fastapi/api/main.py
@@ -3,7 +3,6 @@ from fastapi.middleware.cors import CORSMiddleware
from api.lifespan import app_lifespan
from api.middlewares import UserConfigEnvUpdateMiddleware
from api.v1.ppt.router import API_V1_PPT_ROUTER
-from api.v1.test.router import API_V1_TEST_ROUTER
app = FastAPI(lifespan=app_lifespan)
@@ -11,7 +10,6 @@ app = FastAPI(lifespan=app_lifespan)
# Routers
app.include_router(API_V1_PPT_ROUTER)
-app.include_router(API_V1_TEST_ROUTER)
# Middlewares
origins = ["*"]
diff --git a/servers/fastapi/api/v1/ppt/endpoints/ollama.py b/servers/fastapi/api/v1/ppt/endpoints/ollama.py
index adde8669..0dafa3e1 100644
--- a/servers/fastapi/api/v1/ppt/endpoints/ollama.py
+++ b/servers/fastapi/api/v1/ppt/endpoints/ollama.py
@@ -64,7 +64,7 @@ async def pull_model(
# If the model is being pulled, return the model
if saved_model_status:
# If the model is being pulled, return the model
- # ? If the model status is pulled in redis but was not found while listing pulled models,
+ # ? If the model status is pulled in database but was not found while listing pulled models,
# ? it means the model was deleted and we need to pull it again
if (
saved_model_status["status"] == "error"
diff --git a/servers/fastapi/api/v1/test/router.py b/servers/fastapi/api/v1/test/router.py
deleted file mode 100644
index 71b77d2b..00000000
--- a/servers/fastapi/api/v1/test/router.py
+++ /dev/null
@@ -1,29 +0,0 @@
-from fastapi import APIRouter
-from pydantic import BaseModel, Field
-
-from models.llm_message import LLMUserMessage
-from models.llm_tools import GetCurrentDatetimeTool, SearchWebTool
-from services.llm_client import LLMClient
-from utils.llm_calls.generate_presentation_outlines import generate_ppt_outline
-from utils.llm_provider import get_model
-
-API_V1_TEST_ROUTER = APIRouter(prefix="/api/v1/test", tags=["test"])
-
-
-class ResponseContent(BaseModel):
- trending_ai_tool: str = Field(
- description="The summary of the trending AI tool in about 50 words",
- min_length=50,
- max_length=100,
- )
- current_date_time: str
-
-
-@API_V1_TEST_ROUTER.get("")
-async def test():
- client = LLMClient()
-
- response = await client._search_anthropic("Trending AI tool now")
- # print(response)
-
- return {"data": ""}
diff --git a/servers/fastapi/constants/llm.py b/servers/fastapi/constants/llm.py
index ac4bd527..7d374f30 100644
--- a/servers/fastapi/constants/llm.py
+++ b/servers/fastapi/constants/llm.py
@@ -2,5 +2,5 @@ OPENAI_URL = "https://api.openai.com/v1"
# Default models
DEFAULT_OPENAI_MODEL = "gpt-4.1"
-DEFAULT_GOOGLE_MODEL = "models/gemini-2.0-flash"
-DEFAULT_ANTHROPIC_MODEL = "claude-3-5-sonnet-20240620"
+DEFAULT_GOOGLE_MODEL = "models/gemini-2.5-flash"
+DEFAULT_ANTHROPIC_MODEL = "claude-sonnet-4-20250514"
diff --git a/servers/fastapi/models/user_config.py b/servers/fastapi/models/user_config.py
index 50783544..c040d22c 100644
--- a/servers/fastapi/models/user_config.py
+++ b/servers/fastapi/models/user_config.py
@@ -35,3 +35,6 @@ class UserConfig(BaseModel):
TOOL_CALLS: Optional[bool] = None
DISABLE_THINKING: Optional[bool] = None
EXTENDED_REASONING: Optional[bool] = None
+
+ # Web Search
+ WEB_GROUNDING: Optional[bool] = None
diff --git a/servers/fastapi/services/llm_client.py b/servers/fastapi/services/llm_client.py
index e220f577..3e8b35f2 100644
--- a/servers/fastapi/services/llm_client.py
+++ b/servers/fastapi/services/llm_client.py
@@ -44,10 +44,11 @@ from utils.get_env import (
get_ollama_url_env,
get_openai_api_key_env,
get_tool_calls_env,
+ get_web_grounding_env,
)
from utils.llm_provider import get_llm_provider, get_model
from utils.parsers import parse_bool_or_none
-from utils.schema_utils import ensure_strict_json_schema, flatten_json_schema
+from utils.schema_utils import ensure_strict_json_schema
class LLMClient:
@@ -62,6 +63,15 @@ class LLMClient:
return False
return parse_bool_or_none(get_tool_calls_env()) or False
+ # ? Web Grounding
+ def enable_web_grounding(self) -> bool:
+ if (
+ self.llm_provider == LLMProvider.OLLAMA
+ or self.llm_provider == LLMProvider.CUSTOM
+ ):
+ return False
+ return parse_bool_or_none(get_web_grounding_env()) or False
+
# ? Disable thinking
def disable_thinking(self) -> bool:
return parse_bool_or_none(get_disable_thinking_env()) or False
@@ -569,7 +579,7 @@ class LLMClient:
tools=google_tools,
system_instruction=self._get_system_prompt(messages),
response_mime_type="application/json" if not tools else None,
- response_schema=response_format if not tools else None,
+ response_json_schema=response_format if not tools else None,
max_output_tokens=max_tokens,
),
)
diff --git a/servers/fastapi/services/llm_tool_calls_handler.py b/servers/fastapi/services/llm_tool_calls_handler.py
index 1d8ffec4..ed0d51ee 100644
--- a/servers/fastapi/services/llm_tool_calls_handler.py
+++ b/servers/fastapi/services/llm_tool_calls_handler.py
@@ -51,7 +51,7 @@ class LLMToolCallsHandler:
self.dynamic_tools.append(tool)
match self.client.llm_provider:
- case LLMProvider.OPENAI:
+ case LLMProvider.OPENAI | LLMProvider.OLLAMA | LLMProvider.CUSTOM:
return self.parse_tool_openai(tool, strict)
case LLMProvider.ANTHROPIC:
return self.parse_tool_anthropic(tool)
diff --git a/servers/fastapi/services/redis_service.py b/servers/fastapi/services/redis_service.py
deleted file mode 100644
index f2e3d8c9..00000000
--- a/servers/fastapi/services/redis_service.py
+++ /dev/null
@@ -1,115 +0,0 @@
-from typing import Any, Optional
-import redis
-from redis.exceptions import RedisError
-
-from utils.get_env import (
- get_redis_db_env,
- get_redis_host_env,
- get_redis_password_env,
- get_redis_port_env,
-)
-
-
-class RedisService:
- def __init__(self):
- self.redis_host = get_redis_host_env() or "localhost"
- self.redis_port = int(get_redis_port_env() or "6379")
- self.redis_db = int(get_redis_db_env() or "0")
- self.redis_password = get_redis_password_env() or None
- self.client = self._create_client()
-
- def _create_client(self) -> redis.Redis:
- return redis.Redis(
- host=self.redis_host,
- port=self.redis_port,
- db=self.redis_db,
- password=self.redis_password,
- decode_responses=True,
- )
-
- def set(self, key: str, value: Any, expire: Optional[int] = None) -> bool:
- try:
- return self.client.set(key, value, ex=expire)
- except RedisError:
- return False
-
- def get(self, key: str) -> Optional[str]:
- try:
- return self.client.get(key)
- except RedisError:
- return None
-
- def delete(self, key: str) -> bool:
- try:
- return bool(self.client.delete(key))
- except RedisError:
- return False
-
- def exists(self, key: str) -> bool:
- try:
- return bool(self.client.exists(key))
- except RedisError:
- return False
-
- def set_hash(self, name: str, mapping: dict) -> bool:
- try:
- return self.client.hmset(name, mapping)
- except RedisError:
- return False
-
- def get_hash(self, name: str) -> Optional[dict]:
- try:
- return self.client.hgetall(name)
- except RedisError:
- return None
-
- def delete_hash(self, name: str, *fields: str) -> int:
- try:
- return self.client.hdel(name, *fields)
- except RedisError:
- return 0
-
- def set_list(self, name: str, values: list) -> bool:
- try:
- self.client.delete(name)
- if values:
- self.client.rpush(name, *values)
- return True
- except RedisError:
- return False
-
- def get_list(self, name: str, start: int = 0, end: int = -1) -> Optional[list]:
- try:
- return self.client.lrange(name, start, end)
- except RedisError:
- return None
-
- def add_to_set(self, name: str, *values: str) -> int:
- try:
- return self.client.sadd(name, *values)
- except RedisError:
- return 0
-
- def get_set(self, name: str) -> Optional[set]:
- try:
- return self.client.smembers(name)
- except RedisError:
- return None
-
- def remove_from_set(self, name: str, *values: str) -> int:
- try:
- return self.client.srem(name, *values)
- except RedisError:
- return 0
-
- def clear(self) -> bool:
- try:
- return self.client.flushdb()
- except RedisError:
- return False
-
- def close(self):
- try:
- self.client.close()
- except RedisError:
- pass
diff --git a/servers/fastapi/utils/get_env.py b/servers/fastapi/utils/get_env.py
index c2c72efd..fa80b2a2 100644
--- a/servers/fastapi/utils/get_env.py
+++ b/servers/fastapi/utils/get_env.py
@@ -81,22 +81,6 @@ def get_pixabay_api_key_env():
return os.getenv("PIXABAY_API_KEY")
-def get_redis_host_env():
- return os.getenv("REDIS_HOST")
-
-
-def get_redis_port_env():
- return os.getenv("REDIS_PORT")
-
-
-def get_redis_db_env():
- return os.getenv("REDIS_DB")
-
-
-def get_redis_password_env():
- return os.getenv("REDIS_PASSWORD")
-
-
def get_tool_calls_env():
return os.getenv("TOOL_CALLS")
@@ -107,3 +91,7 @@ def get_disable_thinking_env():
def get_extended_reasoning_env():
return os.getenv("EXTENDED_REASONING")
+
+
+def get_web_grounding_env():
+ return os.getenv("WEB_GROUNDING")
diff --git a/servers/fastapi/utils/llm_calls/generate_presentation_outlines.py b/servers/fastapi/utils/llm_calls/generate_presentation_outlines.py
index 6c0ad512..892b9cff 100644
--- a/servers/fastapi/utils/llm_calls/generate_presentation_outlines.py
+++ b/servers/fastapi/utils/llm_calls/generate_presentation_outlines.py
@@ -4,7 +4,10 @@ from models.llm_message import LLMSystemMessage, LLMUserMessage
from models.llm_tools import GetCurrentDatetimeTool, SearchWebTool
from services.llm_client import LLMClient
from utils.get_dynamic_models import get_presentation_outline_model_with_n_slides
+from utils.get_env import get_web_grounding_env
from utils.llm_provider import get_model
+from utils.parsers import parse_bool_or_none
+from utils.user_config import get_user_config
system_prompt = """
You are an expert presentation creator. Generate structured presentations based on user requirements and format them according to the specified JSON schema with markdown content.
@@ -49,11 +52,13 @@ async def generate_ppt_outline(
client = LLMClient()
+ tools = [SearchWebTool, GetCurrentDatetimeTool]
+
async for chunk in client.stream_structured(
model,
get_messages(prompt, n_slides, language, content),
response_model.model_json_schema(),
strict=True,
- tools=[SearchWebTool, GetCurrentDatetimeTool],
+ tools=tools if client.enable_web_grounding() else None,
):
yield chunk
diff --git a/servers/fastapi/utils/set_env.py b/servers/fastapi/utils/set_env.py
index 7ac0e335..ea3758f3 100644
--- a/servers/fastapi/utils/set_env.py
+++ b/servers/fastapi/utils/set_env.py
@@ -79,3 +79,7 @@ def set_disable_thinking_env(value):
def set_extended_reasoning_env(value):
os.environ["EXTENDED_REASONING"] = value
+
+
+def set_web_grounding_env(value):
+ os.environ["WEB_GROUNDING"] = value
\ No newline at end of file
diff --git a/servers/fastapi/utils/user_config.py b/servers/fastapi/utils/user_config.py
index 06235d5a..49fd1722 100644
--- a/servers/fastapi/utils/user_config.py
+++ b/servers/fastapi/utils/user_config.py
@@ -22,6 +22,7 @@ from utils.get_env import (
get_image_provider_env,
get_pixabay_api_key_env,
get_extended_reasoning_env,
+ get_web_grounding_env,
)
from utils.parsers import parse_bool_or_none
from utils.set_env import (
@@ -43,6 +44,7 @@ from utils.set_env import (
set_image_provider_env,
set_pixabay_api_key_env,
set_tool_calls_env,
+ set_web_grounding_env,
)
@@ -76,12 +78,26 @@ def get_user_config():
IMAGE_PROVIDER=existing_config.IMAGE_PROVIDER or get_image_provider_env(),
PIXABAY_API_KEY=existing_config.PIXABAY_API_KEY or get_pixabay_api_key_env(),
PEXELS_API_KEY=existing_config.PEXELS_API_KEY or get_pexels_api_key_env(),
- TOOL_CALLS=existing_config.TOOL_CALLS
- or parse_bool_or_none(get_tool_calls_env()),
- DISABLE_THINKING=existing_config.DISABLE_THINKING
- or parse_bool_or_none(get_disable_thinking_env()),
- EXTENDED_REASONING=existing_config.EXTENDED_REASONING
- or parse_bool_or_none(get_extended_reasoning_env()),
+ TOOL_CALLS=(
+ existing_config.TOOL_CALLS
+ if existing_config.TOOL_CALLS is not None
+ else (parse_bool_or_none(get_tool_calls_env()) or False)
+ ),
+ DISABLE_THINKING=(
+ existing_config.DISABLE_THINKING
+ if existing_config.DISABLE_THINKING is not None
+ else (parse_bool_or_none(get_disable_thinking_env()) or False)
+ ),
+ EXTENDED_REASONING=(
+ existing_config.EXTENDED_REASONING
+ if existing_config.EXTENDED_REASONING is not None
+ else (parse_bool_or_none(get_extended_reasoning_env()) or False)
+ ),
+ WEB_GROUNDING=(
+ existing_config.WEB_GROUNDING
+ if existing_config.WEB_GROUNDING is not None
+ else (parse_bool_or_none(get_web_grounding_env()) or False)
+ ),
)
@@ -122,5 +138,6 @@ def update_env_with_user_config():
if user_config.DISABLE_THINKING:
set_disable_thinking_env(str(user_config.DISABLE_THINKING))
if user_config.EXTENDED_REASONING:
- if user_config.EXTENDED_REASONING:
- set_extended_reasoning_env(str(user_config.EXTENDED_REASONING))
+ set_extended_reasoning_env(str(user_config.EXTENDED_REASONING))
+ if user_config.WEB_GROUNDING:
+ set_web_grounding_env(str(user_config.WEB_GROUNDING))
diff --git a/servers/nextjs/app/api/user-config/route.ts b/servers/nextjs/app/api/user-config/route.ts
index ff3c643a..03b801e3 100644
--- a/servers/nextjs/app/api/user-config/route.ts
+++ b/servers/nextjs/app/api/user-config/route.ts
@@ -57,6 +57,10 @@ export async function POST(request: Request) {
userConfig.EXTENDED_REASONING === undefined
? existingConfig.EXTENDED_REASONING
: userConfig.EXTENDED_REASONING,
+ WEB_GROUNDING:
+ userConfig.WEB_GROUNDING === undefined
+ ? existingConfig.WEB_GROUNDING
+ : userConfig.WEB_GROUNDING,
USE_CUSTOM_URL:
userConfig.USE_CUSTOM_URL === undefined
? existingConfig.USE_CUSTOM_URL
diff --git a/servers/nextjs/components/AnthropicConfig.tsx b/servers/nextjs/components/AnthropicConfig.tsx
index 567846b5..4b61bb65 100644
--- a/servers/nextjs/components/AnthropicConfig.tsx
+++ b/servers/nextjs/components/AnthropicConfig.tsx
@@ -19,6 +19,7 @@ interface AnthropicConfigProps {
anthropicApiKey: string;
anthropicModel: string;
extendedReasoning: boolean;
+ webGrounding?: boolean;
onInputChange: (value: string | boolean, field: string) => void;
}
@@ -27,6 +28,7 @@ export default function AnthropicConfig({
anthropicApiKey,
anthropicModel,
extendedReasoning,
+ webGrounding,
onInputChange,
}: AnthropicConfigProps) {
const [openModelSelect, setOpenModelSelect] = useState(false);
@@ -65,7 +67,7 @@ export default function AnthropicConfig({
const data = await response.json();
setAvailableModels(data);
setModelsChecked(true);
- onInputChange("claude-3-5-sonnet-20241022", "anthropic_model");
+ onInputChange("claude-sonnet-4-20250514", "anthropic_model");
} else {
console.error('Failed to fetch models');
setAvailableModels([]);
@@ -226,6 +228,23 @@ export default function AnthropicConfig({
) : null}
+
+ {/* Web Grounding Toggle - at the end, below models dropdown */}
+
+
+
+ onInputChange(checked, "web_grounding")}
+ />
+
+
+
+ If enabled, the model can use web search grounding when available.
+
+
+
+
+ onInputChange(checked, "web_grounding")}
+ />
+
+
+
+ If enabled, the model can use web search grounding when available.
+
+
+
+
+ onInputChange(checked, "web_grounding")}
+ />
+
+
+
+ If enabled, the model can use web search grounding when available.
+
+