Oliver/presenton
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

refactor: removes redis service and env variables, fix(fastapi): user config bool env variables issues, parse tool fix for custom llm on tool call structured output

Browse source
This commit is contained in:
sauravniraula 2025-08-09 03:03:13 +05:45
parent dc62eb72d1
commit 3f523f1491
No known key found for this signature in database
GPG key ID: 60FCC1B5A5E83326
20 changed files with 131 additions and 181 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
servers/fastapi/api/main.py
View file

@ -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 = ["*"]

2
servers/fastapi/api/v1/ppt/endpoints/ollama.py
View file

@ -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"

29
servers/fastapi/api/v1/test/router.py
View file

@ -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": ""}

4
servers/fastapi/constants/llm.py
View file

@ -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"

3
servers/fastapi/models/user_config.py
View file

@ -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

14
servers/fastapi/services/llm_client.py
View file

@ -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,
),
)

2
servers/fastapi/services/llm_tool_calls_handler.py
View file

@ -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)

115
servers/fastapi/services/redis_service.py
View file

@ -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

20
servers/fastapi/utils/get_env.py
View file

@ -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")

7
servers/fastapi/utils/llm_calls/generate_presentation_outlines.py
View file

@ -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

4
servers/fastapi/utils/set_env.py
View file

@ -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

33
servers/fastapi/utils/user_config.py
View file

@ -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))

4
servers/nextjs/app/api/user-config/route.ts
View file

@ -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

21
servers/nextjs/components/AnthropicConfig.tsx
View file

@ -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({
</div>
</div>
) : null}
{/* Web Grounding Toggle - at the end, below models dropdown */}
<div>
<div className="flex items-center justify-between mb-4 bg-green-50 p-2 rounded-sm">
<label className="text-sm font-medium text-gray-700">
Enable Web Grounding
</label>
<Switch
checked={!!webGrounding}
onCheckedChange={(checked) => onInputChange(checked, "web_grounding")}
/>
</div>
<p className="mt-2 text-sm text-gray-500 flex items-center gap-2">
<span className="block w-1 h-1 rounded-full bg-gray-400"></span>
If enabled, the model can use web search grounding when available.
</p>
</div>
</div>
);
}

24
servers/nextjs/components/GoogleConfig.tsx
View file

@ -13,16 +13,19 @@ import {
import { Popover, PopoverContent, PopoverTrigger } from "./ui/popover";
import { cn } from "@/lib/utils";
import { toast } from "sonner";
import { Switch } from "./ui/switch";
interface GoogleConfigProps {
googleApiKey: string;
googleModel: string;
onInputChange: (value: string, field: string) => void;
webGrounding?: boolean;
onInputChange: (value: string | boolean, field: string) => void;
}
export default function GoogleConfig({
googleApiKey,
googleModel,
webGrounding,
onInputChange
}: GoogleConfigProps) {
const [openModelSelect, setOpenModelSelect] = useState(false);
@ -61,7 +64,7 @@ export default function GoogleConfig({
const data = await response.json();
setAvailableModels(data);
setModelsChecked(true);
onInputChange("models/gemini-2.0-flash", "google_model");
onInputChange("models/gemini-2.5-flash", "google_model");
} else {
console.error('Failed to fetch models');
setAvailableModels([]);
@ -205,6 +208,23 @@ export default function GoogleConfig({
</div>
</div>
) : null}
{/* Web Grounding Toggle - at the end, below models dropdown */}
<div>
<div className="flex items-center justify-between mb-4 bg-green-50 p-2 rounded-sm">
<label className="text-sm font-medium text-gray-700">
Enable Web Grounding
</label>
<Switch
checked={!!webGrounding}
onCheckedChange={(checked) => onInputChange(checked, "web_grounding")}
/>
</div>
<p className="mt-2 text-sm text-gray-500 flex items-center gap-2">
<span className="block w-1 h-1 rounded-full bg-gray-400"></span>
If enabled, the model can use web search grounding when available.
</p>
</div>
</div>
);
}

3
servers/nextjs/components/LLMSelection.tsx
View file

@ -149,6 +149,7 @@ export default function LLMProviderSelection({
<OpenAIConfig
openaiApiKey={llmConfig.OPENAI_API_KEY || ""}
openaiModel={llmConfig.OPENAI_MODEL || ""}
webGrounding={llmConfig.WEB_GROUNDING || false}
onInputChange={input_field_changed}
/>
</TabsContent>
@ -158,6 +159,7 @@ export default function LLMProviderSelection({
<GoogleConfig
googleApiKey={llmConfig.GOOGLE_API_KEY || ""}
googleModel={llmConfig.GOOGLE_MODEL || ""}
webGrounding={llmConfig.WEB_GROUNDING || false}
onInputChange={input_field_changed}
/>
</TabsContent>
@ -168,6 +170,7 @@ export default function LLMProviderSelection({
anthropicApiKey={llmConfig.ANTHROPIC_API_KEY || ""}
anthropicModel={llmConfig.ANTHROPIC_MODEL || ""}
extendedReasoning={llmConfig.EXTENDED_REASONING || false}
webGrounding={llmConfig.WEB_GROUNDING || false}
onInputChange={input_field_changed}
/>
</TabsContent>

22
servers/nextjs/components/OpenAIConfig.tsx
View file

@ -13,16 +13,19 @@ import {
import { Popover, PopoverContent, PopoverTrigger } from "./ui/popover";
import { cn } from "@/lib/utils";
import { toast } from "sonner";
import { Switch } from "./ui/switch";
interface OpenAIConfigProps {
openaiApiKey: string;
openaiModel: string;
onInputChange: (value: string, field: string) => void;
webGrounding?: boolean;
onInputChange: (value: string | boolean, field: string) => void;
}
export default function OpenAIConfig({
openaiApiKey,
openaiModel,
webGrounding,
onInputChange
}: OpenAIConfigProps) {
const [openModelSelect, setOpenModelSelect] = useState(false);
@ -210,6 +213,23 @@ export default function OpenAIConfig({
</div>
</div>
) : null}
{/* Web Grounding Toggle - show at the end, below models dropdown */}
<div>
<div className="flex items-center justify-between mb-4 bg-green-50 p-2 rounded-sm">
<label className="text-sm font-medium text-gray-700">
Enable Web Grounding
</label>
<Switch
checked={!!webGrounding}
onCheckedChange={(checked) => onInputChange(checked, "web_grounding")}
/>
</div>
<p className="mt-2 text-sm text-gray-500 flex items-center gap-2">
<span className="block w-1 h-1 rounded-full bg-gray-400"></span>
If enabled, the model can use web search grounding when available.
</p>
</div>
</div>
);
}

1
servers/nextjs/types/llm_config.ts
View file

@ -31,6 +31,7 @@ export interface LLMConfig {
TOOL_CALLS?: boolean;
DISABLE_THINKING?: boolean;
EXTENDED_REASONING?: boolean;
WEB_GROUNDING?: boolean;
// Only used in UI settings
USE_CUSTOM_URL?: boolean;

1
servers/nextjs/utils/providerUtils.ts
View file

@ -48,6 +48,7 @@ export const updateLLMConfig = (
tool_calls: "TOOL_CALLS",
disable_thinking: "DISABLE_THINKING",
extended_reasoning: "EXTENDED_REASONING",
web_grounding: "WEB_GROUNDING",
};
const configKey = fieldMappings[field];

1
start.js
View file

@ -81,6 +81,7 @@ const setupUserConfigFromEnv = () => {
TOOL_CALLS: process.env.TOOL_CALLS || existingConfig.TOOL_CALLS,
DISABLE_THINKING: process.env.DISABLE_THINKING || existingConfig.DISABLE_THINKING,
EXTENDED_REASONING: process.env.EXTENDED_REASONING || existingConfig.EXTENDED_REASONING,
WEB_GROUNDING: process.env.WEB_GROUNDING || existingConfig.WEB_GROUNDING,
USE_CUSTOM_URL: process.env.USE_CUSTOM_URL || existingConfig.USE_CUSTOM_URL,
};