lokiai / main.py
ParthSadaria's picture
Update main.py
96a881b verified
raw
history blame
5.71 kB
import os
from dotenv import load_dotenv
from fastapi import FastAPI, HTTPException, Request
from fastapi.responses import StreamingResponse, HTMLResponse, JSONResponse
from pydantic import BaseModel
import httpx
import requests
import re
import json
from typing import Optional
load_dotenv()
app = FastAPI()
# Get API keys and secret endpoint from environment variables
api_keys_str = os.getenv('API_KEYS')
valid_api_keys = api_keys_str.split(',') if api_keys_str else []
secret_api_endpoint = os.getenv('SECRET_API_ENDPOINT')
secret_api_endpoint_2 = os.getenv('SECRET_API_ENDPOINT_2')
secret_api_endpoint_3 = os.getenv('SECRET_API_ENDPOINT_3') # New endpoint for searchgpt
# Validate if the main secret API endpoints are set
if not secret_api_endpoint or not secret_api_endpoint_2 or not secret_api_endpoint_3:
raise HTTPException(status_code=500, detail="API endpoint(s) are not configured in environment variables.")
# Define models that should use the secondary endpoint
alternate_models = {"gpt-4o-mini", "claude-3-haiku", "llama-3.1-70b", "mixtral-8x7b"}
class Payload(BaseModel):
model: str
messages: list
stream: bool
def generate_search(query: str, stream: bool = True) -> str:
headers = {"User-Agent": ""}
prompt = [
{"role": "user", "content": query},
]
# Insert the system prompt at the beginning of the conversation history
prompt.insert(0, {"content": "Be Helpful and Friendly", "role": "system"})
payload = {
"is_vscode_extension": True,
"message_history": prompt,
"requested_model": "searchgpt",
"user_input": prompt[-1]["content"],
}
# Use the newly added SECRET_API_ENDPOINT_3 for the search API call
chat_endpoint = secret_api_endpoint_3
response = requests.post(chat_endpoint, headers=headers, json=payload, stream=True)
# Collect streamed text content
streaming_text = ""
for value in response.iter_lines(decode_unicode=True, chunk_size=12):
modified_value = re.sub("data:", "", value)
if modified_value:
try:
json_modified_value = json.loads(modified_value)
content = json_modified_value["choices"][0]["delta"]["content"]
if stream:
yield f"data: {content}\n\n"
streaming_text += content
except:
continue
if not stream:
yield streaming_text
@app.get("/searchgpt")
async def search_gpt(q: str, stream: Optional[bool] = False):
if not q:
raise HTTPException(status_code=400, detail="Query parameter 'q' is required")
if stream:
return StreamingResponse(
generate_search(q, stream=True),
media_type="text/event-stream"
)
else:
# For non-streaming response, collect all content and return as JSON
response_text = "".join([chunk for chunk in generate_search(q, stream=False)])
return JSONResponse(content={"response": response_text})
@app.get("/", response_class=HTMLResponse)
async def root():
# Open and read the content of index.html (in the same folder as the app)
file_path = "index.html"
try:
with open(file_path, "r") as file:
html_content = file.read()
return HTMLResponse(content=html_content)
except FileNotFoundError:
return HTMLResponse(content="<h1>File not found</h1>", status_code=404)
async def get_models():
async with httpx.AsyncClient() as client:
try:
response = await client.get(f"{secret_api_endpoint}/v1/models", timeout=3)
response.raise_for_status()
return response.json()
except httpx.RequestError as e:
raise HTTPException(status_code=500, detail=f"Request failed: {e}")
@app.get("/models")
async def fetch_models():
return await get_models()
@app.post(["/chat/completions", "/v1/chat/completions"])
async def get_completion(payload: Payload, request: Request):
# Use the correct endpoint depending on the model type (no authentication now πŸ˜‰)
endpoint = secret_api_endpoint_2 if payload.model in alternate_models else secret_api_endpoint
# Use the payload directly as it includes stream and other user data
payload_dict = payload.dict()
print(payload_dict) # coz i m curious af heheheh :)
#data is kept to me only so dont worry
async def stream_generator(payload_dict):
async with httpx.AsyncClient() as client:
try:
async with client.stream("POST", f"{endpoint}/v1/chat/completions", json=payload_dict, timeout=10) as response:
response.raise_for_status()
async for line in response.aiter_lines():
if line:
yield f"{line}\n"
except httpx.HTTPStatusError as status_err:
raise HTTPException(status_code=status_err.response.status_code, detail=f"HTTP error: {status_err}")
except httpx.RequestError as req_err:
raise HTTPException(status_code=500, detail=f"Streaming failed: {req_err}")
except Exception as e:
raise HTTPException(status_code=500, detail=f"An unexpected error occurred: {e}")
return StreamingResponse(stream_generator(openai_payload), media_type="application/json")
@app.on_event("startup")
async def startup_event():
print("API endpoints:")
print("GET /")
print("GET /models")
print("GET /searchgpt") # We now have the new search API
print("POST /chat/completions")
if __name__ == "__main__":
import uvicorn
uvicorn.run(app, host="0.0.0.0", port=8000)