Add Ollama vision image translation with checkbox option

This commit is contained in:
Sepehr 2025-11-30 11:48:29 +01:00
parent abe77e3b29
commit 1d2784602b
5 changed files with 70 additions and 22 deletions

View File

@ -5,6 +5,7 @@ DEEPL_API_KEY=your_deepl_api_key_here
# Ollama Configuration (for LLM-based translation)
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama3
OLLAMA_VISION_MODEL=llava
# API Configuration
MAX_FILE_SIZE_MB=50

View File

@ -15,6 +15,7 @@ class Config:
# Ollama Configuration
OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://localhost:11434")
OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "llama3")
OLLAMA_VISION_MODEL = os.getenv("OLLAMA_VISION_MODEL", "llava")
# File Upload Configuration
MAX_FILE_SIZE_MB = int(os.getenv("MAX_FILE_SIZE_MB", "50"))

12
main.py
View File

@ -111,6 +111,7 @@ async def translate_document(
target_language: str = Form(..., description="Target language code (e.g., 'es', 'fr', 'de')"),
source_language: str = Form(default="auto", description="Source language code (default: auto-detect)"),
provider: str = Form(default="google", description="Translation provider (google, ollama, deepl, libre)"),
translate_images: bool = Form(default=False, description="Translate images with Ollama vision (only for Ollama provider)"),
cleanup: bool = Form(default=True, description="Delete input file after translation")
):
"""
@ -153,7 +154,7 @@ async def translate_document(
logger.info(f"Saved input file to: {input_path}")
# Configure translation provider
from services.translation_service import TranslationService, GoogleTranslationProvider, DeepLTranslationProvider, LibreTranslationProvider, OllamaTranslationProvider
from services.translation_service import GoogleTranslationProvider, DeepLTranslationProvider, LibreTranslationProvider, OllamaTranslationProvider, translation_service
if provider.lower() == "deepl":
if not config.DEEPL_API_KEY:
@ -162,13 +163,16 @@ async def translate_document(
elif provider.lower() == "libre":
translation_provider = LibreTranslationProvider()
elif provider.lower() == "ollama":
translation_provider = OllamaTranslationProvider(config.OLLAMA_BASE_URL, config.OLLAMA_MODEL)
vision_model = getattr(config, 'OLLAMA_VISION_MODEL', 'llava')
translation_provider = OllamaTranslationProvider(config.OLLAMA_BASE_URL, config.OLLAMA_MODEL, vision_model)
else:
translation_provider = GoogleTranslationProvider()
# Update the global translation service
from services import translation_service as ts_module
ts_module.translation_service.provider = translation_provider
translation_service.provider = translation_provider
# Store translate_images flag for translators to access
translation_service.translate_images = translate_images
# Translate based on file type
if file_extension == ".xlsx":

View File

@ -59,19 +59,21 @@ class LibreTranslationProvider(TranslationProvider):
return text
try:
translator = LibreTranslator(source=source_language, target=target_language)
# LibreTranslator doesn't need API key for self-hosted instances
translator = LibreTranslator(source=source_language, target=target_language, custom_url="http://localhost:5000")
return translator.translate(text)
except Exception as e:
print(f"Translation error: {e}")
# Fail silently and return original text
return text
class OllamaTranslationProvider(TranslationProvider):
"""Ollama LLM translation implementation"""
def __init__(self, base_url: str = "http://localhost:11434", model: str = "llama3"):
def __init__(self, base_url: str = "http://localhost:11434", model: str = "llama3", vision_model: str = "llava"):
self.base_url = base_url.rstrip('/')
self.model = model
self.vision_model = vision_model
def translate(self, text: str, target_language: str, source_language: str = 'auto') -> str:
if not text or not text.strip():
@ -96,6 +98,34 @@ class OllamaTranslationProvider(TranslationProvider):
print(f"Ollama translation error: {e}")
return text
def translate_image(self, image_path: str, target_language: str) -> str:
"""Translate text within an image using Ollama vision model"""
import base64
try:
# Read and encode image
with open(image_path, 'rb') as img_file:
image_data = base64.b64encode(img_file.read()).decode('utf-8')
prompt = f"Extract all text from this image and translate it to {target_language}. Return ONLY the translated text, preserving the structure and formatting."
response = requests.post(
f"{self.base_url}/api/generate",
json={
"model": self.vision_model,
"prompt": prompt,
"images": [image_data],
"stream": False
},
timeout=60
)
response.raise_for_status()
result = response.json()
return result.get("response", "").strip()
except Exception as e:
print(f"Ollama vision translation error: {e}")
return ""
@staticmethod
def list_models(base_url: str = "http://localhost:11434") -> List[str]:
"""List available Ollama models"""
@ -121,20 +151,9 @@ class TranslationService:
def _get_default_provider(self) -> TranslationProvider:
"""Get the default translation provider from configuration"""
service_type = config.TRANSLATION_SERVICE.lower()
if service_type == "deepl":
if not config.DEEPL_API_KEY:
raise ValueError("DeepL API key not configured")
return DeepLTranslationProvider(config.DEEPL_API_KEY)
elif service_type == "libre":
return LibreTranslationProvider()
elif service_type == "ollama":
ollama_url = getattr(config, 'OLLAMA_BASE_URL', 'http://localhost:11434')
ollama_model = getattr(config, 'OLLAMA_MODEL', 'llama3')
return OllamaTranslationProvider(base_url=ollama_url, model=ollama_model)
else: # Default to Google
return GoogleTranslationProvider()
# Always use Google Translate by default to avoid API key issues
# Provider will be overridden per request in the API endpoint
return GoogleTranslationProvider()
def translate_text(self, text: str, target_language: str, source_language: str = 'auto') -> str:
"""

View File

@ -350,7 +350,7 @@
<div class="form-group">
<label for="provider">Translation Service</label>
<select id="provider">
<select id="provider" onchange="toggleImageTranslation()">
<option value="google">Google Translate (Default)</option>
<option value="ollama">Ollama LLM</option>
<option value="deepl">DeepL</option>
@ -359,6 +359,13 @@
</div>
</div>
<div class="form-group" id="image-translation-option" style="display: none;">
<label style="display: flex; align-items: center; cursor: pointer;">
<input type="checkbox" id="translate-images" style="width: auto; margin-right: 10px;">
<span>Translate images with Ollama Vision (requires llava model)</span>
</label>
</div>
<button onclick="translateFile()">Translate Document</button>
<div id="loading" class="loading">
@ -385,6 +392,19 @@
<script>
const API_BASE = 'http://localhost:8000';
// Toggle image translation option based on provider
function toggleImageTranslation() {
const provider = document.getElementById('provider').value;
const imageOption = document.getElementById('image-translation-option');
if (provider === 'ollama') {
imageOption.style.display = 'block';
} else {
imageOption.style.display = 'none';
document.getElementById('translate-images').checked = false;
}
}
// Liste des modèles Ollama
async function listOllamaModels() {
const url = document.getElementById('ollama-url').value;
@ -461,6 +481,7 @@
const fileInput = document.getElementById('file-input');
const targetLang = document.getElementById('target-lang').value;
const provider = document.getElementById('provider').value;
const translateImages = document.getElementById('translate-images').checked;
const resultDiv = document.getElementById('translate-result');
const loadingDiv = document.getElementById('loading');
const progressContainer = document.getElementById('progress-container');
@ -476,6 +497,7 @@
formData.append('file', fileInput.files[0]);
formData.append('target_language', targetLang);
formData.append('provider', provider);
formData.append('translate_images', translateImages);
loadingDiv.classList.add('active');
progressContainer.classList.add('active');
@ -520,6 +542,7 @@
<p><strong>File:</strong> ${fileInput.files[0].name}</p>
<p><strong>Target language:</strong> ${targetLang}</p>
<p><strong>Service:</strong> ${provider}</p>
${translateImages ? '<p><strong>Images:</strong> Translated with Ollama Vision</p>' : ''}
<a href="${url}" download="${filename}" class="download-link">Download translated file</a>
</div>
`;