File size: 14,859 Bytes
475b0b9 a506b86 475b0b9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 |
import argparse
import io
import os
from time import time
from typing import List
import tempfile
import uvicorn
from fastapi import Depends, FastAPI, File, HTTPException, Query, Request, UploadFile, Body, Form
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import JSONResponse, RedirectResponse, StreamingResponse
from PIL import Image
from pydantic import BaseModel, field_validator
from pydantic_settings import BaseSettings
from slowapi import Limiter
from slowapi.util import get_remote_address
import requests
from logging_config import logger
from tts_config import SPEED, ResponseFormat, config as tts_config
from gemma_llm import LLMManager
from auth import get_api_key, settings as auth_settings
# Supported language codes
SUPPORTED_LANGUAGES = {
"asm_Beng", "kas_Arab", "pan_Guru", "ben_Beng", "kas_Deva", "san_Deva",
"brx_Deva", "mai_Deva", "sat_Olck", "doi_Deva", "mal_Mlym", "snd_Arab",
"eng_Latn", "mar_Deva", "snd_Deva", "gom_Deva", "mni_Beng", "tam_Taml",
"guj_Gujr", "mni_Mtei", "tel_Telu", "hin_Deva", "npi_Deva", "urd_Arab",
"kan_Knda", "ory_Orya"
}
class Settings(BaseSettings):
llm_model_name: str = "google/gemma-3-4b-it"
max_tokens: int = 512
host: str = "0.0.0.0"
port: int = 7860
chat_rate_limit: str = "100/minute"
speech_rate_limit: str = "5/minute"
@field_validator("chat_rate_limit", "speech_rate_limit")
def validate_rate_limit(cls, v):
if not v.count("/") == 1 or not v.split("/")[0].isdigit():
raise ValueError("Rate limit must be in format 'number/period' (e.g., '5/minute')")
return v
class Config:
env_file = ".env"
settings = Settings()
app = FastAPI(
title="Dhwani API",
description="AI Chat API supporting Indian languages",
version="1.0.0",
redirect_slashes=False,
)
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=False,
allow_methods=["*"],
allow_headers=["*"],
)
limiter = Limiter(key_func=get_remote_address)
app.state.limiter = limiter
llm_manager = LLMManager(settings.llm_model_name)
class ChatRequest(BaseModel):
prompt: str
src_lang: str = "kan_Knda" # Default to Kannada
tgt_lang: str = "kan_Knda" # Default to Kannada
@field_validator("prompt")
def prompt_must_be_valid(cls, v):
if len(v) > 1000:
raise ValueError("Prompt cannot exceed 1000 characters")
return v.strip()
@field_validator("src_lang", "tgt_lang")
def validate_language(cls, v):
if v not in SUPPORTED_LANGUAGES:
raise ValueError(f"Unsupported language code: {v}. Supported codes: {', '.join(SUPPORTED_LANGUAGES)}")
return v
class ChatResponse(BaseModel):
response: str
class TranslationRequest(BaseModel):
sentences: List[str]
src_lang: str
tgt_lang: str
@field_validator("src_lang", "tgt_lang")
def validate_language(cls, v):
if v not in SUPPORTED_LANGUAGES:
raise ValueError(f"Unsupported language code: {v}. Supported codes: {', '.join(SUPPORTED_LANGUAGES)}")
return v
class TranslationResponse(BaseModel):
translations: List[str]
async def call_external_translation(sentences: List[str], src_lang: str, tgt_lang: str) -> List[str]:
external_url = "https://gaganyatri-dhwani-server.hf.space/v1/translate"
payload = {
"sentences": sentences,
"src_lang": src_lang,
"tgt_lang": tgt_lang
}
try:
response = requests.post(
external_url,
json=payload,
headers={
"accept": "application/json",
"Content-Type": "application/json"
},
timeout=10
)
response.raise_for_status()
translations = response.json().get("translations", [])
if not translations or len(translations) != len(sentences):
logger.warning(f"Unexpected response format: {response.json()}")
raise ValueError("Invalid response from translation service")
return translations
except requests.Timeout:
logger.error("Translation request timed out")
raise HTTPException(status_code=504, detail="Translation service timeout")
except requests.RequestException as e:
logger.error(f"Error during translation: {str(e)}")
raise HTTPException(status_code=500, detail=f"Translation failed: {str(e)}")
except ValueError as e:
logger.error(f"Invalid response: {str(e)}")
raise HTTPException(status_code=500, detail=str(e))
@app.get("/v1/health")
async def health_check():
return {"status": "healthy", "model": settings.llm_model_name}
@app.get("/")
async def home():
return RedirectResponse(url="/docs")
@app.post("/v1/unload_all_models")
async def unload_all_models(api_key: str = Depends(get_api_key)):
try:
logger.info("Starting to unload all models...")
llm_manager.unload()
logger.info("All models unloaded successfully")
return {"status": "success", "message": "All models unloaded"}
except Exception as e:
logger.error(f"Error unloading models: {str(e)}")
raise HTTPException(status_code=500, detail=f"Failed to unload models: {str(e)}")
@app.post("/v1/load_all_models")
async def load_all_models(api_key: str = Depends(get_api_key)):
try:
logger.info("Starting to load all models...")
llm_manager.load()
logger.info("All models loaded successfully")
return {"status": "success", "message": "All models loaded"}
except Exception as e:
logger.error(f"Error loading models: {str(e)}")
raise HTTPException(status_code=500, detail=f"Failed to load models: {str(e)}")
@app.post("/v1/translate", response_model=TranslationResponse)
async def translate(request: TranslationRequest):
logger.info(f"Received translation request: {request.dict()}")
try:
translations = await call_external_translation(
sentences=request.sentences,
src_lang=request.src_lang,
tgt_lang=request.tgt_lang
)
logger.info(f"Translation successful: {translations}")
return TranslationResponse(translations=translations)
except HTTPException as e:
raise e
except Exception as e:
logger.error(f"Unexpected error during translation: {str(e)}")
raise HTTPException(status_code=500, detail=f"Translation failed: {str(e)}")
@app.post("/v1/chat", response_model=ChatResponse)
@limiter.limit(settings.chat_rate_limit)
async def chat(request: Request, chat_request: ChatRequest, api_key: str = Depends(get_api_key)):
if not chat_request.prompt:
raise HTTPException(status_code=400, detail="Prompt cannot be empty")
logger.info(f"Received prompt: {chat_request.prompt}, src_lang: {chat_request.src_lang}, tgt_lang: {chat_request.tgt_lang}")
try:
# Translate prompt to English if src_lang is not English
if chat_request.src_lang != "eng_Latn":
translated_prompt = await call_external_translation(
sentences=[chat_request.prompt],
src_lang=chat_request.src_lang,
tgt_lang="eng_Latn"
)
prompt_to_process = translated_prompt[0]
logger.info(f"Translated prompt to English: {prompt_to_process}")
else:
prompt_to_process = chat_request.prompt
logger.info("Prompt already in English, no translation needed")
# Generate response in English
response = await llm_manager.generate(prompt_to_process, settings.max_tokens)
logger.info(f"Generated English response: {response}")
# Translate response to target language if tgt_lang is not English
if chat_request.tgt_lang != "eng_Latn":
translated_response = await call_external_translation(
sentences=[response],
src_lang="eng_Latn",
tgt_lang=chat_request.tgt_lang
)
final_response = translated_response[0]
logger.info(f"Translated response to {chat_request.tgt_lang}: {final_response}")
else:
final_response = response
logger.info("Response kept in English, no translation needed")
return ChatResponse(response=final_response)
except Exception as e:
logger.error(f"Error processing request: {str(e)}")
raise HTTPException(status_code=500, detail=f"An error occurred: {str(e)}")
@app.post("/v1/visual_query/")
async def visual_query(
file: UploadFile = File(...),
query: str = Body(...),
src_lang: str = Query("kan_Knda", enum=list(SUPPORTED_LANGUAGES)),
tgt_lang: str = Query("kan_Knda", enum=list(SUPPORTED_LANGUAGES)),
api_key: str = Depends(get_api_key)
):
try:
image = Image.open(file.file)
if image.size == (0, 0):
raise HTTPException(status_code=400, detail="Uploaded image is empty or invalid")
# Translate query to English if src_lang is not English
if src_lang != "eng_Latn":
translated_query = await call_external_translation(
sentences=[query],
src_lang=src_lang,
tgt_lang="eng_Latn"
)
query_to_process = translated_query[0]
logger.info(f"Translated query to English: {query_to_process}")
else:
query_to_process = query
logger.info("Query already in English, no translation needed")
# Generate response in English
answer = await llm_manager.vision_query(image, query_to_process)
logger.info(f"Generated English answer: {answer}")
# Translate answer to target language if tgt_lang is not English
if tgt_lang != "eng_Latn":
translated_answer = await call_external_translation(
sentences=[answer],
src_lang="eng_Latn",
tgt_lang=tgt_lang
)
final_answer = translated_answer[0]
logger.info(f"Translated answer to {tgt_lang}: {final_answer}")
else:
final_answer = answer
logger.info("Answer kept in English, no translation needed")
return {"answer": final_answer}
except Exception as e:
logger.error(f"Error processing request: {str(e)}")
raise HTTPException(status_code=500, detail=f"An error occurred: {str(e)}")
@app.post("/v1/chat_v2", response_model=ChatResponse)
@limiter.limit(settings.chat_rate_limit)
async def chat_v2(
request: Request,
prompt: str = Form(...),
image: UploadFile = File(default=None),
src_lang: str = Form("kan_Knda"),
tgt_lang: str = Form("kan_Knda"),
api_key: str = Depends(get_api_key)
):
if not prompt:
raise HTTPException(status_code=400, detail="Prompt cannot be empty")
if src_lang not in SUPPORTED_LANGUAGES or tgt_lang not in SUPPORTED_LANGUAGES:
raise HTTPException(status_code=400, detail=f"Unsupported language code. Supported codes: {', '.join(SUPPORTED_LANGUAGES)}")
logger.info(f"Received prompt: {prompt}, src_lang: {src_lang}, tgt_lang: {tgt_lang}, Image provided: {image is not None}")
try:
if image:
image_data = await image.read()
if not image_data:
raise HTTPException(status_code=400, detail="Uploaded image is empty")
img = Image.open(io.BytesIO(image_data))
# Translate prompt to English if src_lang is not English
if src_lang != "eng_Latn":
translated_prompt = await call_external_translation(
sentences=[prompt],
src_lang=src_lang,
tgt_lang="eng_Latn"
)
prompt_to_process = translated_prompt[0]
logger.info(f"Translated prompt to English: {prompt_to_process}")
else:
prompt_to_process = prompt
logger.info("Prompt already in English, no translation needed")
decoded = await llm_manager.chat_v2(img, prompt_to_process)
logger.info(f"Generated English response: {decoded}")
# Translate response to target language if tgt_lang is not English
if tgt_lang != "eng_Latn":
translated_response = await call_external_translation(
sentences=[decoded],
src_lang="eng_Latn",
tgt_lang=tgt_lang
)
final_response = translated_response[0]
logger.info(f"Translated response to {tgt_lang}: {final_response}")
else:
final_response = decoded
logger.info("Response kept in English, no translation needed")
else:
# Translate prompt to English if src_lang is not English
if src_lang != "eng_Latn":
translated_prompt = await call_external_translation(
sentences=[prompt],
src_lang=src_lang,
tgt_lang="eng_Latn"
)
prompt_to_process = translated_prompt[0]
logger.info(f"Translated prompt to English: {prompt_to_process}")
else:
prompt_to_process = prompt
logger.info("Prompt already in English, no translation needed")
decoded = await llm_manager.generate(prompt_to_process, settings.max_tokens)
logger.info(f"Generated English response: {decoded}")
# Translate response to target language if tgt_lang is not English
if tgt_lang != "eng_Latn":
translated_response = await call_external_translation(
sentences=[decoded],
src_lang="eng_Latn",
tgt_lang=tgt_lang
)
final_response = translated_response[0]
logger.info(f"Translated response to {tgt_lang}: {final_response}")
else:
final_response = decoded
logger.info("Response kept in English, no translation needed")
return ChatResponse(response=final_response)
except Exception as e:
logger.error(f"Error processing request: {str(e)}")
raise HTTPException(status_code=500, detail=f"An error occurred: {str(e)}")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="Run the FastAPI server.")
parser.add_argument("--port", type=int, default=settings.port, help="Port to run the server on.")
parser.add_argument("--host", type=str, default=settings.host, help="Host to run the server on.")
args = parser.parse_args()
uvicorn.run(app, host=args.host, port=args.port) |