LMX app

.gitignore

@@ -0,0 +1,2 @@
+.venv
+.DS_Store

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Jobin Thavu Varghese
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

app.py

@@ -0,0 +1,95 @@
+import gradio as gr
+from services.ocr import upload_file, ocr, ocr_languages
+from services.interleave import interleave
+from services.translation import translation_languages, translation
+from services.transliteration import transliterate_languages, transliterate
+
+img_path="bmc/qr-code.png"
+
+html_header = f"""
+<div style="display:flex; align-items:center; justify-content:space-between;">
+    <h1 style="margin:0;">LMX</h1>
+</div>
+"""
+
+html_footer = f"""
+<footer>
+</footer>
+"""
+
+with gr.Blocks(title="LMX", analytics_enabled=True) as interface:
+
+    with gr.Row():
+        gr.HTML(html_header)
+        gr.Button(variant="primary", scale=1, min_width=5, value="☕️ Buy me a coffee", size="lg", link_target="_blank", link="https://buymeacoffee.com/jobvargh")
+        gr.Button(variant="primary", scale=1, min_width=5, value="Github", size="lg", link_target="_blank", link="https://github.com/jobinv12/LMX")
+
+    
+    with gr.Tab("OCR"):
+        with gr.Row():
+            with gr.Column():
+                ocr_path = gr.Textbox(interactive=False, visible=False)
+                ocr_languages = gr.Dropdown(label="Source language", choices=ocr_languages, filterable=True, info="Select source language")
+                ocr_upload_file = gr.UploadButton(label="Upload a Image", file_types=[".png",".jpeg",".webp",".jpg"], file_count="single")
+                ocr_file = gr.Image(height=250, width=700, type="filepath", show_label=False, buttons=[], container=False)
+                ocr_upload_file.upload(upload_file, inputs=ocr_upload_file, outputs=[ocr_file, ocr_path])
+
+            with gr.Column():
+                ocr_output = gr.TextArea(label="Output", lines=15 ,interactive=False)
+
+        with gr.Row():
+            clear_btn = gr.ClearButton(value="Clear", variant="secondary")
+            ocr_submit_btn = gr.Button(value="Run", variant="huggingface")
+
+        ocr_submit_btn.click(fn=ocr, inputs=[ocr_path, ocr_languages], outputs=ocr_output)
+
+    with gr.Tab("Transliteration"):
+        with gr.Row():
+            with gr.Column():
+                transliteration_src_lang = gr.Dropdown(label="Source Language", choices=transliterate_languages, filterable=True, interactive=True)
+                transliteration_input_area = gr.TextArea(label="Input")
+
+            with gr.Column():
+                transliteration_trgt_lang = gr.Dropdown(label="Target Language", choices=transliterate_languages, filterable=True, interactive=True)
+                transliteration_output_area = gr.TextArea(label="Output", interactive=False)
+    
+        with gr.Row():
+            clear_btn = gr.ClearButton(value="Clear", variant="secondary")
+            transliteration_submit_btn = gr.Button(value="Run", variant="huggingface")
+        
+        transliteration_submit_btn.click(fn=transliterate, inputs=[transliteration_src_lang, transliteration_trgt_lang, transliteration_input_area], outputs=transliteration_output_area)
+    
+    with gr.Tab("Translation"):
+        with gr.Row():
+            with gr.Column():
+                translation_src_lang = gr.Dropdown(label="Source Language", choices=translation_languages, filterable=True, interactive=True)
+                translation_input_area = gr.TextArea(label="Input")
+            with gr.Column():
+                translation_trgt_lang = gr.Dropdown(label="Target Language", choices=translation_languages, filterable=True, interactive=True)
+                translation_output_area = gr.TextArea(label="Output", interactive=False)
+    
+        with gr.Row():
+            clear_btn = gr.ClearButton(value="Clear", variant="secondary")
+            translation_submit_btn = gr.Button(value="Run", variant="huggingface")  
+
+        translation_submit_btn.click(fn=translation, inputs=[translation_src_lang, translation_trgt_lang, translation_input_area], outputs=translation_output_area) 
+
+
+    with gr.Tab("Interleave"):
+        with gr.Row():
+            interleave_input_area = gr.TextArea(label="First Input", type="text")
+            interleave_input2_area = gr.TextArea(label="Second Input", type="text")
+
+        with gr.Row():
+            clear_btn = gr.ClearButton(value="Clear", variant="secondary")
+            interleave_submit_btn = gr.Button(value="Run" ,variant="huggingface")    
+
+        with gr.Row():
+            interleave_output_area = gr.TextArea(label="Output", interactive=False, type="text")
+        interleave_submit_btn.click(fn=interleave, inputs=[interleave_input_area, interleave_input2_area], outputs=[interleave_output_area])
+    
+    with gr.Row():
+        gr.HTML(html_template=html_footer)
+
+if __name__ == "__main__":
+    interface.launch(pwa=True, share=True, allowed_paths=["assets", "bmc"])

pyproject.toml

@@ -0,0 +1,7 @@
+[project]
+name = "alephlm"
+version = "0.1.0"
+description = "Add your description here"
+readme = "README.md"
+requires-python = ">=3.12"
+dependencies = []

requirements.txt

@@ -0,0 +1,91 @@
+aiofiles==24.1.0
+alabaster==1.0.0
+annotated-doc==0.0.4
+annotated-types==0.7.0
+anyio==4.12.1
+babel==2.18.0
+brotli==1.2.0
+certifi==2026.2.25
+charset-normalizer==3.4.6
+click==8.3.1
+colorama==0.4.6
+cython==3.2.4
+docutils==0.22.4
+fastapi==0.135.1
+ffmpy==1.0.0
+filelock==3.25.2
+fsspec==2026.2.0
+gradio==6.9.0
+gradio-client==2.3.0
+groovy==0.1.2
+h11==0.16.0
+hf-xet==1.4.2
+httpcore==1.0.9
+httpx==0.28.1
+huggingface-hub==0.36.2
+idna==3.11
+imagesize==2.0.0
+indic-nlp-library-itt==0.1.1
+indictranstoolkit==1.1.1
+jinja2==3.1.6
+joblib==1.5.3
+lxml==6.0.2
+markdown-it-py==4.0.0
+markupsafe==3.0.3
+mdurl==0.1.2
+mlx==0.31.1
+mlx-metal==0.31.1
+morfessor==2.0.6
+mpmath==1.3.0
+networkx==3.6.1
+numpy==2.4.3
+orjson==3.11.7
+packaging==26.0
+pandas==3.0.1
+pillow==12.1.1
+portalocker==3.2.0
+pydantic==2.12.5
+pydantic-core==2.41.5
+pydub==0.25.1
+pygments==2.19.2
+python-dateutil==2.9.0.post0
+python-multipart==0.0.22
+pytz==2026.1.post1
+pyyaml==6.0.3
+regex==2026.2.28
+requests==2.32.5
+rich==14.3.3
+roman-numerals==4.1.0
+sacrebleu==2.6.0
+sacremoses==0.1.1
+safehttpx==0.1.7
+safetensors==0.7.0
+semantic-version==2.10.0
+setuptools==82.0.1
+shellingham==1.5.4
+six==1.17.0
+snowballstemmer==3.0.1
+sphinx==9.1.0
+sphinx-argparse==0.5.2
+sphinx-rtd-theme==3.1.0
+sphinxcontrib-applehelp==2.0.0
+sphinxcontrib-devhelp==2.0.0
+sphinxcontrib-htmlhelp==2.1.0
+sphinxcontrib-jquery==4.1
+sphinxcontrib-jsmath==1.0.1
+sphinxcontrib-qthelp==2.0.0
+sphinxcontrib-serializinghtml==2.0.0
+starlette==0.52.1
+sympy==1.14.0
+tabulate==0.10.0
+tokenizers==0.21.4
+tomlkit==0.13.3
+torch==2.10.0
+torchvision==0.25.0
+tqdm==4.67.3
+transformers==4.53.2
+typer==0.24.1
+typing-extensions==4.15.0
+typing-inspection==0.4.2
+urllib3==2.6.3
+uvicorn==0.42.0

services/interleave.py

@@ -0,0 +1,11 @@
+def interleave(first_input_text:str, second_input_text:str) -> str:
+    first_input = [
+        line.strip() for line in first_input_text.splitlines() if line.strip()
+    ]
+    second_input = [
+        line.strip() for line in second_input_text.splitlines() if line.strip()
+    ]
+
+    result = "\n\n".join(f"{x}\n{y}" for x, y in zip(first_input, second_input))
+
+    return result

services/ocr.py

@@ -0,0 +1,74 @@
+import os
+from datetime import datetime
+import shutil
+import mlx.core as mx
+from mlx_vlm import load, generate
+from mlx_vlm.prompt_utils import apply_chat_template
+from mlx_vlm.utils import load_config
+
+ocr_languages = [
+    "English", "Afrikaans", "Amharic", "Arabic", "Assamese", "Azerbaijani", 
+    "Azerbaijani - Cyrillic", "Belarusian", "Bengali", "Tibetan", "Bosnian", 
+    "Breton", "Bulgarian", "Catalan; Valencian", "Cebuano", "Czech", 
+    "Chinese", "Cherokee", "Corsican", "Welsh", 
+    "Danish", "Danish - Fraktur", "German", "German", 
+    "Dzongkha", "Greek", "Esperanto", 
+    "Estonian", "Basque", "Faroese", "Persian", "Filipino", 
+    "Finnish", "French", "French", "Western Frisian", 
+    "Scottish Gaelic", "Irish", "Galician", "Gujarati", 
+    "Haitian; Haitian Creole", "Hebrew", "Hindi", "Croatian", "Hungarian", 
+    "Armenian", "Inuktitut", "Indonesian", "Icelandic", "Italian",
+    "Javanese", "Japanese", "Kannada", "Georgian", "Georgian - Old", "Kazakh", 
+    "Central Khmer", "Kirghiz; Kyrgyz", "Kurmanji", "Korean", 
+    "Korean", "Kurdish", "Lao", "Latin", "Latvian", 
+    "Lithuanian", "Luxembourgish", "Malayalam", "Marathi", "Macedonian", 
+    "Maltese", "Mongolian", "Maori", "Malay", "Burmese", "Nepali", 
+    "Dutch; Flemish", "Norwegian", "Occitan", "Oriya", 
+    "Panjabi; Punjabi", "Polish", "Portuguese", "Pushto; Pashto", "Quechua", 
+    "Romanian; Moldavian; Moldovan", "Russian", "Sanskrit", "Sinhala; Sinhalese", 
+    "Slovak", "Slovak - Fraktur", "Slovenian", "Sindhi", "Spanish; Castilian", 
+    "Spanish; Castilian - Old", "Albanian", "Serbian", "Serbian - Latin", 
+    "Sundanese", "Swahili", "Swedish", "Syriac", "Tamil", "Tatar", "Telugu", 
+    "Tajik", "Thai", "Tigrinya", "Tonga", "Turkish", 
+    "Uighur; Uyghur", "Ukrainian", "Urdu", "Uzbek", 
+    "Vietnamese", "Yiddish", "Yoruba"
+]
+
+model_id = "mlx-community/olmOCR-2-7B-1025-8bit"
+
+try:
+    # Load the model
+    model, processor = load(model_id, trust_remote_code=True)
+    config = load_config(model_id)
+    print("olmocr2 model loaded successfully")
+except Exception as e:
+    print(f"Failed to load deepseekocr2 model {e}")
+
+def upload_file(file):
+
+    if file is None:
+        return "No file uploaded"
+    
+    timestamp = datetime.now().strftime("%Y%m%d")
+
+    original_name = os.path.basename(file.name)
+
+    name, ext = os.path.splitext(original_name)
+
+    new_filename = f"{name}_{timestamp}{ext}"
+
+    dest_path = os.path.join("assets", new_filename)
+
+    shutil.copy(file, dest_path)
+
+    return dest_path, dest_path
+    
+def ocr(image_path:str, language:str) -> str:
+
+    prompt = f"Convert the {language} document to markdown."
+    
+    formatted_output = apply_chat_template(processor, config, prompt, num_images=1)
+
+    output = generate(model, processor, formatted_output, [image_path], verbose=False, max_tokens=2000)
+
+    return output.text

services/translation.py

@@ -0,0 +1,352 @@
+import torch
+from transformers import AutoModelForImageTextToText, AutoProcessor
+
+translation_languages: list = ['English', 'Abkhaz', 'Acehnese', 'Acholi', 
+                         'Afar', 'Afrikaans', 'Albanian', 'Alur', 
+                         'Amharic', 'Arabic', 'Armenian', 'Assamese', 
+                         'Avar', 'Awadhi', 'Aymara', 'Azerbaijani', 
+                         'Balinese', 'Baluchi', 'Bambara', 'Baoulé', 
+                         'Bashkir', 'Basque', 'Batak Karo', 'Batak Simalungun', 
+                         'Batak Toba', 'Belarusian', 'Bemba', 'Bengali', 'Betawi', 
+                         'Bhojpuri', 'Bikol', 'Bosnian', 'Breton', 'Bulgarian', 
+                         'Buryat', 'Cantonese', 'Catalan', 'Cebuano', 'Chamorro', 
+                         'Chechen', 'Chichewa', 'Chinese (Simplified)', 
+                         'Chinese (Traditional)', 'Chuukese', 'Chuvash', 
+                         'Corsican', 'Crimean Tatar (Cyrillic)', 
+                         'Crimean Tatar (Latin)', 'Croatian', 'Czech', 
+                         'Danish', 'Dari', 'Dhivehi', 'Dinka', 'Dogri', 
+                         'Dombe', 'Dutch', 'Dyula', 'Dzongkha', 'Esperanto', 
+                         'Estonian', 'Ewe', 'Faroese', 'Fijian', 'Filipino', 
+                         'Finnish', 'Fon', 'French', 'French (Canada)', 'Frisian', 
+                         'Friulian', 'Fulani', 'Ga', 'Galician', 'Georgian', 'German', 
+                         'Greek', 'Guarani', 'Gujarati', 'Haitian Creole', 'Hakha Chin',
+                         'Hausa', 'Hawaiian', 'Hebrew', 'Hiligaynon', 'Hindi', 
+                         'Hmong', 'Hungarian', 'Hunsrik', 'Iban', 'Icelandic', 
+                         'Igbo', 'Ilocano', 'Indonesian', 'Inuktut (Latin)', 
+                         'Inuktut (Syllabics)', 'Irish', 'Italian', 
+                         'Jamaican Patois', 'Japanese', 'Javanese', 'Jingpo', 
+                         'Kalaallisut', 'Kannada', 'Kanuri', 'Kapampangan', 
+                         'Kazakh', 'Khasi', 'Khmer', 'Kiga', 'Kikongo', 
+                         'Kinyarwanda', 'Kituba', 'Kokborok', 'Komi', 'Konkani', 
+                         'Korean', 'Krio', 'Kurdish (Kurmanji)', 'Kurdish (Sorani)',
+                         'Kyrgyz', 'Lao', 'Latgalian', 'Latin', 'Latvian', 
+                         'Ligurian', 'Limburgish', 'Lingala', 'Lithuanian', 
+                         'Lombard', 'Luganda', 'Luo', 'Luxembourgish', 
+                         'Macedonian', 'Madurese', 'Maithili', 'Makassar', 
+                         'Malagasy', 'Malay', 'Malay (Jawi)', 'Malayalam', 
+                         'Maltese', 'Mam', 'Manx', 'Maori', 'Marathi', 
+                         'Marshallese', 'Marwadi', 'Mauritian Creole', 
+                         'Meadow Mari', 'Meiteilon (Manipuri)', 'Minang', 
+                         'Mizo', 'Mongolian', 'Myanmar (Burmese)', 
+                         'Nahuatl (Eastern Huasteca)', 'Ndau', 'Ndebele (South)', 
+                         'Nepalbhasa (Newari)', 'Nepali', 'NKo', 'Norwegian', 
+                         'Nuer', 'Occitan', 'Odia (Oriya)', 'Oromo', 'Ossetian',
+                         'Pangasinan', 'Papiamento', 'Pashto', 'Persian', 
+                         'Polish', 'Portuguese (Brazil)', 'Portuguese (Portugal)', 
+                         'Punjabi (Gurmukhi)', 'Punjabi (Shahmukhi)', 'Quechua', 
+                         'Qʼeqchiʼ', 'Romani', 'Romanian', 'Rundi', 'Russian', 
+                         'Sami (North)', 'Samoan', 'Sango', 'Sanskrit', 'Santali (Latin)', 
+                         'Santali (Ol Chiki)', 'Scots Gaelic', 'Sepedi', 'Serbian', 
+                         'Sesotho', 'Seychellois Creole', 'Shan', 'Shona', 'Sicilian', 
+                         'Silesian', 'Sindhi', 'Sinhala', 'Slovak', 'Slovenian', 'Somali', 
+                         'Spanish', 'Sundanese', 'Susu', 'Swahili', 'Swati', 'Swedish', 
+                         'Tahitian', 'Tajik', 'Tamazight', 'Tamazight (Tifinagh)', 'Tamil', 
+                         'Tatar', 'Telugu', 'Tetum', 'Thai', 'Tibetan', 'Tigrinya', 'Tiv', 
+                         'Tok Pisin', 'Tongan', 'Tshiluba', 'Tsonga', 'Tswana', 'Tulu', 
+                         'Tumbuka', 'Turkish', 'Turkmen', 'Tuvan', 'Twi', 'Udmurt', 
+                         'Ukrainian', 'Urdu', 'Uyghur', 'Uzbek', 'Venda', 'Venetian', 
+                         'Vietnamese', 'Waray', 'Welsh', 'Wolof', 'Xhosa', 'Yakut', 
+                         'Yiddish', 'Yoruba', 'Yucatec Maya', 'Zapotec', 'Zulu']
+
+languages_iso_codes: dict = {
+    "English": "en",
+    "Abkhaz": "ab",
+    "Acehnese": "ace",
+    "Acholi": "ach",
+    "Afar": "aa",
+    "Afrikaans": "af",
+    "Albanian": "sq",
+    "Alur": "alz",
+    "Amharic": "am",
+    "Arabic": "ar",
+    "Armenian": "hy",
+    "Assamese": "as",
+    "Avar": "av",
+    "Awadhi": "awa",
+    "Aymara": "ay",
+    "Azerbaijani": "az",
+    "Balinese": "ban",
+    "Baluchi": "bal",
+    "Bambara": "bm",
+    "Baoulé": "bci",
+    "Bashkir": "ba",
+    "Basque": "eu",
+    "Batak Karo": "btx",
+    "Batak Simalungun": "bts",
+    "Batak Toba": "bbc",
+    "Belarusian": "be",
+    "Bemba": "bem",
+    "Bengali": "bn",
+    "Betawi": "bew",
+    "Bhojpuri": "bho",
+    "Bikol": "bik",
+    "Bosnian": "bs",
+    "Breton": "br",
+    "Bulgarian": "bg",
+    "Buryat": "bua",
+    "Cantonese": "yue",
+    "Catalan": "ca",
+    "Cebuano": "ceb",
+    "Chamorro": "ch",
+    "Chechen": "ce",
+    "Chichewa": "ny",
+    "Chinese (Simplified)": "zh-CN",
+    "Chinese (Traditional)": "zh-TW",
+    "Chuukese": "chk",
+    "Chuvash": "cv",
+    "Corsican": "co",
+    "Crimean Tatar (Cyrillic)": "crh-Cyrl",
+    "Crimean Tatar (Latin)": "crh-Latn",
+    "Croatian": "hr",
+    "Czech": "cs",
+    "Danish": "da",
+    "Dari": "prs",
+    "Dhivehi": "dv",
+    "Dinka": "din",
+    "Dogri": "doi",
+    "Dombe": "dov",
+    "Dutch": "nl",
+    "Dyula": "dyu",
+    "Dzongkha": "dz",
+    "Esperanto": "eo",
+    "Estonian": "et",
+    "Ewe": "ee",
+    "Faroese": "fo",
+    "Fijian": "fj",
+    "Filipino": "fil",
+    "Finnish": "fi",
+    "Fon": "fon",
+    "French": "fr",
+    "French (Canada)": "fr-CA",
+    "Frisian": "fy",
+    "Friulian": "fur",
+    "Fulani": "ff",
+    "Ga": "gaa",
+    "Galician": "gl",
+    "Georgian": "ka",
+    "German": "de",
+    "Greek": "el",
+    "Guarani": "gn",
+    "Gujarati": "gu",
+    "Haitian Creole": "ht",
+    "Hakha Chin": "cnh",
+    "Hausa": "ha",
+    "Hawaiian": "haw",
+    "Hebrew": "he",
+    "Hiligaynon": "hil",
+    "Hindi": "hi",
+    "Hmong": "hmn",
+    "Hungarian": "hu",
+    "Hunsrik": "hrx",
+    "Iban": "iba",
+    "Icelandic": "is",
+    "Igbo": "ig",
+    "Ilocano": "ilo",
+    "Indonesian": "id",
+    "Inuktut (Latin)": "iu-Latn",
+    "Inuktut (Syllabics)": "iu-Cans",
+    "Irish": "ga",
+    "Italian": "it",
+    "Jamaican Patois": "jam",
+    "Japanese": "ja",
+    "Javanese": "jv",
+    "Jingpo": "kac",
+    "Kalaallisut": "kl",
+    "Kannada": "kn",
+    "Kanuri": "kr",
+    "Kapampangan": "pam",
+    "Kazakh": "kk",
+    "Khasi": "kha",
+    "Khmer": "km",
+    "Kiga": "cgg",
+    "Kikongo": "kg",
+    "Kinyarwanda": "rw",
+    "Kituba": "ktu",
+    "Kokborok": "trp",
+    "Komi": "kv",
+    "Konkani": "gom",
+    "Korean": "ko",
+    "Krio": "kri",
+    "Kurdish (Kurmanji)": "ku",
+    "Kurdish (Sorani)": "ckb",
+    "Kyrgyz": "ky",
+    "Lao": "lo",
+    "Latgalian": "ltg",
+    "Latin": "la",
+    "Latvian": "lv",
+    "Ligurian": "lij",
+    "Limburgish": "li",
+    "Lingala": "ln",
+    "Lithuanian": "lt",
+    "Lombard": "lmo",
+    "Luganda": "lg",
+    "Luo": "luo",
+    "Luxembourgish": "lb",
+    "Macedonian": "mk",
+    "Madurese": "mad",
+    "Maithili": "mai",
+    "Makassar": "mak",
+    "Malagasy": "mg",
+    "Malay": "ms",
+    "Malay (Jawi)": "ms-Arab",
+    "Malayalam": "ml",
+    "Maltese": "mt",
+    "Mam": "mam",
+    "Manx": "gv",
+    "Maori": "mi",
+    "Marathi": "mr",
+    "Marshallese": "mh",
+    "Marwadi": "mwr",
+    "Mauritian Creole": "mfe",
+    "Meadow Mari": "mhr",
+    "Meiteilon (Manipuri)": "mni",
+    "Minang": "min",
+    "Mizo": "lus",
+    "Mongolian": "mn",
+    "Myanmar (Burmese)": "my",
+    "Nahuatl (Eastern Huasteca)": "nhe",
+    "Ndau": "ndc",
+    "Ndebele (South)": "nr",
+    "Nepalbhasa (Newari)": "new",
+    "Nepali": "ne",
+    "NKo": "nqo",
+    "Norwegian": "no",
+    "Nuer": "nus",
+    "Occitan": "oc",
+    "Odia (Oriya)": "or",
+    "Oromo": "om",
+    "Ossetian": "os",
+    "Pangasinan": "pag",
+    "Papiamento": "pap",
+    "Pashto": "ps",
+    "Persian": "fa",
+    "Polish": "pl",
+    "Portuguese (Brazil)": "pt-BR",
+    "Portuguese (Portugal)": "pt-PT",
+    "Punjabi (Gurmukhi)": "pa",
+    "Punjabi (Shahmukhi)": "pa-Arab",
+    "Quechua": "qu",
+    "Qʼeqchiʼ": "kek",
+    "Romani": "rom",
+    "Romanian": "ro",
+    "Rundi": "rn",
+    "Russian": "ru",
+    "Sami (North)": "se",
+    "Samoan": "sm",
+    "Sango": "sg",
+    "Sanskrit": "sa",
+    "Santali (Latin)": "sat-Latn",
+    "Santali (Ol Chiki)": "sat",
+    "Scots Gaelic": "gd",
+    "Sepedi": "nso",
+    "Serbian": "sr",
+    "Sesotho": "st",
+    "Seychellois Creole": "crs",
+    "Shan": "shn",
+    "Shona": "sn",
+    "Sicilian": "scn",
+    "Silesian": "szl",
+    "Sindhi": "sd",
+    "Sinhala": "si",
+    "Slovak": "sk",
+    "Slovenian": "sl",
+    "Somali": "so",
+    "Spanish": "es",
+    "Sundanese": "su",
+    "Susu": "sus",
+    "Swahili": "sw",
+    "Swati": "ss",
+    "Swedish": "sv",
+    "Tahitian": "ty",
+    "Tajik": "tg",
+    "Tamazight": "tzm",
+    "Tamazight (Tifinagh)": "zgh",
+    "Tamil": "ta",
+    "Tatar": "tt",
+    "Telugu": "te",
+    "Tetum": "tet",
+    "Thai": "th",
+    "Tibetan": "bo",
+    "Tigrinya": "ti",
+    "Tiv": "tiv",
+    "Tok Pisin": "tpi",
+    "Tongan": "to",
+    "Tshiluba": "lua",
+    "Tsonga": "ts",
+    "Tswana": "tn",
+    "Tulu": "tcy",
+    "Tumbuka": "tum",
+    "Turkish": "tr",
+    "Turkmen": "tk",
+    "Tuvan": "tyv",
+    "Twi": "tw",
+    "Udmurt": "udm",
+    "Ukrainian": "uk",
+    "Urdu": "ur",
+    "Uyghur": "ug",
+    "Uzbek": "uz",
+    "Venda": "ve",
+    "Venetian": "vec",
+    "Vietnamese": "vi",
+    "Waray": "war",
+    "Welsh": "cy",
+    "Wolof": "wo",
+    "Xhosa": "xh",
+    "Yakut": "sah",
+    "Yiddish": "yi",
+    "Yoruba": "yo",
+    "Yucatec Maya": "yua",
+    "Zapotec": "zap",
+    "Zulu": "zu"
+}
+
+model_id = "google/translategemma-4b-it"
+
+try:
+   processor = AutoProcessor.from_pretrained(model_id)
+   model = AutoModelForImageTextToText.from_pretrained(model_id, device_map="auto")
+   print("Translation model loaded successfully")
+except Exception as e:
+    print(f"Failed to load translation model {e}")
+
+def translation(src_lang:str, trgt_lang:str, text:str) -> str:
+
+    src_lang_iso = languages_iso_codes[src_lang]
+    trgt_lang_iso = languages_iso_codes[trgt_lang]
+
+    messages = [
+        {
+            "role": "user",
+            "content": [
+                {
+                    "type": "text",
+                    "source_lang_code": src_lang_iso,
+                    "target_lang_code": trgt_lang_iso,
+                    "text": text,
+                }
+            ],
+        }
+    ]
+
+    inputs = processor.apply_chat_template(
+        messages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors="pt"
+    ).to(model.device, dtype=torch.bfloat16)
+    input_len = len(inputs['input_ids'][0])
+
+    with torch.inference_mode():
+        generation = model.generate(**inputs, do_sample=False, max_new_tokens=2000)
+
+    generation = generation[0][input_len:]
+    decoded = processor.decode(generation, skip_special_tokens=True)
+
+    return decoded

services/transliteration.py

@@ -0,0 +1,105 @@
+import torch
+from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
+from IndicTransToolkit.processor import IndicProcessor
+
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+
+
+transliterate_languages: list = [
+    "English",
+    "Assamese",
+    "Bengali",
+    "Bodo",
+    "Dogri",
+    "Konkani",
+    "Gujarati",
+    "Hindi",
+    "Kannada",
+    "Kashmiri",
+    "Kashmiri",
+    "Maithili",
+    "Malayalam",
+    "Marathi",
+    "Manipuri",
+    "Manipuri",
+    "Nepali",
+    "Odia",
+    "Punjabi",
+    "Sanskrit",
+    "Santali",
+    "Sindhi",
+    "Sindhi",
+    "Tamil",
+    "Telugu",
+    "Urdu"
+]
+
+transliteration_iso_codes: dict = {
+    "English": "eng_Latn",
+    "Assamese": "asm_Beng",
+    "Bengali": "ben_Beng",
+    "Bodo": "brx_Deva",
+    "Dogri": "doi_Deva",
+    "Konkani": "gom_Deva",
+    "Gujarati": "guj_Gujr",
+    "Hindi": "hin_Deva",
+    "Kannada": "kan_Knda",
+    "Kashmiri (Arabic)": "kas_Arab",
+    "Kashmiri (Devanagari)": "kas_Deva",
+    "Maithili": "mai_Deva",
+    "Malayalam": "mal_Mlym",
+    "Marathi": "mar_Deva",
+    "Manipuri (Bengali)": "mni_Beng",
+    "Manipuri (Meitei)": "mni_Mtei",
+    "Nepali": "npi_Deva",
+    "Odia": "ory_Orya",
+    "Punjabi": "pan_Guru",
+    "Sanskrit": "san_Deva",
+    "Santali": "sat_Olck",
+    "Sindhi (Arabic)": "snd_Arab",
+    "Sindhi (Devanagari) ()": "snd_Deva",
+    "Tamil": "tam_Taml",
+    "Telugu": "tel_Telu",
+    "Urdu": "urd_Arab"
+}
+
+def indic_transliteration(model_name:str, src_lang:str, trgt_lang:str, text:str) -> str:
+
+    model_id = model_name
+    try:
+        tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
+        model = AutoModelForSeq2SeqLM.from_pretrained(model_id,trust_remote_code=True,torch_dtype=torch.bfloat16).to(DEVICE)
+        print(f"{model_id} model loaded sucessfully.")
+    except Exception as e:
+        print(f"Failed to load {model_id} model.")
+
+    ip = IndicProcessor(inference=True)
+
+    batch = ip.preprocess_batch(
+        [text],
+        src_lang=src_lang,
+        tgt_lang=trgt_lang
+    )
+
+    inputs = tokenizer(batch, truncation=True, padding="longest", return_tensors="pt", return_attention_mask=True).to(DEVICE)
+
+    with torch.no_grad():
+        generated_tokens = model.generate(**inputs, use_cache=True, min_length=0, max_length=2048, num_beams=5, num_return_sequences=1)
+    
+    generated_tokens = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True, clean_up_tokenization_spaces=True)
+
+    translations = ip.postprocess_batch(generated_tokens, lang=trgt_lang)
+
+    return "\n".join(translations)
+
+def transliterate(src_lang:str, trgt_lang:str, text:str) -> str:
+
+    src_lang_iso:str = transliteration_iso_codes.get(src_lang)
+    trgt_lang_iso:str = transliteration_iso_codes.get(trgt_lang)
+
+    if "eng_Latn" in src_lang_iso and "eng_Latn" not in trgt_lang_iso:
+        return indic_transliteration("ai4bharat/indictrans2-en-indic-1B", src_lang_iso, trgt_lang_iso, text)
+    elif "eng_Latn" not in src_lang_iso and "eng_Latn" in trgt_lang_iso:
+        return indic_transliteration("ai4bharat/indictrans2-indic-en-1B", src_lang_iso, trgt_lang_iso, text)
+    else:
+        return indic_transliteration("ai4bharat/indictrans2-indic-indic-1B", src_lang_iso, trgt_lang_iso, text)

uv.lock

@@ -0,0 +1,8 @@
+version = 1
+revision = 3
+requires-python = ">=3.12"
+
+[[package]]
+name = "alephlm"
+version = "0.1.0"
+source = { virtual = "." }