Spaces:
Running
Running
push
Browse files- README.md +114 -12
- main.py +1 -1
- multilingual/multi.py +41 -98
- requirements.txt +1 -1
README.md
CHANGED
|
@@ -1,12 +1,114 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Farmlingua Multilingual Translation API
|
| 2 |
+
|
| 3 |
+
A FastAPI-based multilingual translation service supporting Igbo, Hausa, Yoruba, Swahili, and English for agricultural queries.
|
| 4 |
+
|
| 5 |
+
## Features
|
| 6 |
+
|
| 7 |
+
- **Simple Language Identifier**: Specify language with `lang` parameter
|
| 8 |
+
- **Single Endpoint**: `/ask` - Send text with language identifier
|
| 9 |
+
- **Supported Languages**: English (en), Igbo (ig), Hausa (ha), Yoruba (yo), Swahili (sw)
|
| 10 |
+
- **Agricultural Focus**: Specialized for farming and agricultural queries
|
| 11 |
+
- **External Integration**: Connects to Milestone 1 Q&A service
|
| 12 |
+
|
| 13 |
+
## Quick Start
|
| 14 |
+
|
| 15 |
+
### Local Development
|
| 16 |
+
|
| 17 |
+
1. **Install Dependencies**
|
| 18 |
+
```bash
|
| 19 |
+
pip install -r requirements.txt
|
| 20 |
+
```
|
| 21 |
+
|
| 22 |
+
2. **Run the Server**
|
| 23 |
+
```bash
|
| 24 |
+
python main.py
|
| 25 |
+
```
|
| 26 |
+
|
| 27 |
+
3. **Test the API**
|
| 28 |
+
```bash
|
| 29 |
+
python test_simple.py
|
| 30 |
+
```
|
| 31 |
+
|
| 32 |
+
4. **Access Documentation**
|
| 33 |
+
- Interactive API docs: http://localhost:8000/docs
|
| 34 |
+
- Health check: http://localhost:8000/health
|
| 35 |
+
|
| 36 |
+
### Docker Deployment
|
| 37 |
+
|
| 38 |
+
1. **Build Image**
|
| 39 |
+
```bash
|
| 40 |
+
docker build -t farmlingua-api .
|
| 41 |
+
```
|
| 42 |
+
|
| 43 |
+
2. **Run Container**
|
| 44 |
+
```bash
|
| 45 |
+
docker run -p 8000:8000 farmlingua-api
|
| 46 |
+
```
|
| 47 |
+
|
| 48 |
+
## API Usage
|
| 49 |
+
|
| 50 |
+
### Single Endpoint: `/ask`
|
| 51 |
+
|
| 52 |
+
**POST** `/ask`
|
| 53 |
+
|
| 54 |
+
**Request Body:**
|
| 55 |
+
```json
|
| 56 |
+
{
|
| 57 |
+
"text": "Kedu ka m ga-esi na-akụ ọka?",
|
| 58 |
+
"lang": "ig"
|
| 59 |
+
}
|
| 60 |
+
```
|
| 61 |
+
|
| 62 |
+
**Response:**
|
| 63 |
+
```json
|
| 64 |
+
{
|
| 65 |
+
"language": "Igbo",
|
| 66 |
+
"language_code": "ig",
|
| 67 |
+
"original_text": "Kedu ka m ga-esi na-akụ ọka?",
|
| 68 |
+
"english_translation": "How can I plant grain?",
|
| 69 |
+
"response": "Jiri oghere mkpụrụ osisi ma ọ bụ osisi aka..."
|
| 70 |
+
}
|
| 71 |
+
```
|
| 72 |
+
|
| 73 |
+
### Supported Language Codes
|
| 74 |
+
|
| 75 |
+
| Code | Language | NLLB Code |
|
| 76 |
+
|------|----------|-----------|
|
| 77 |
+
| `en` | English | `eng_Latn` |
|
| 78 |
+
| `ig` | Igbo | `ibo_Latn` |
|
| 79 |
+
| `ha` | Hausa | `hau_Latn` |
|
| 80 |
+
| `yo` | Yoruba | `yor_Latn` |
|
| 81 |
+
| `sw` | Swahili | `swh_Latn` |
|
| 82 |
+
|
| 83 |
+
## Project Structure
|
| 84 |
+
|
| 85 |
+
```
|
| 86 |
+
FARMLINGUAMULTILINGUA/
|
| 87 |
+
├── models/
|
| 88 |
+
│ └── model.py # Hugging Face translation model
|
| 89 |
+
├── multilingual/
|
| 90 |
+
│ └── multi.py # FastAPI application
|
| 91 |
+
├── main.py # Application entry point
|
| 92 |
+
├── test_simple.py # Test suite
|
| 93 |
+
├── requirements.txt # Python dependencies
|
| 94 |
+
├── Dockerfile # Docker configuration
|
| 95 |
+
├── .gitignore # Git ignore rules
|
| 96 |
+
└── README.md # This file
|
| 97 |
+
```
|
| 98 |
+
|
| 99 |
+
## Environment Variables
|
| 100 |
+
|
| 101 |
+
- `HOST`: Server host (default: `0.0.0.0`)
|
| 102 |
+
- `PORT`: Server port (default: `8000`)
|
| 103 |
+
|
| 104 |
+
## Technology Stack
|
| 105 |
+
|
| 106 |
+
- **FastAPI**: Web framework
|
| 107 |
+
- **Transformers**: Hugging Face model pipeline
|
| 108 |
+
- **NLLB Model**: `drrobot9/nllb-ig-yo-ha-finetuned`
|
| 109 |
+
- **Uvicorn**: ASGI server
|
| 110 |
+
- **Docker**: Containerization
|
| 111 |
+
|
| 112 |
+
## License
|
| 113 |
+
|
| 114 |
+
MIT License
|
main.py
CHANGED
|
@@ -4,7 +4,7 @@ import uvicorn
|
|
| 4 |
|
| 5 |
if __name__ == "__main__":
|
| 6 |
host = os.getenv("HOST", "0.0.0.0")
|
| 7 |
-
port = int(os.getenv("PORT", "
|
| 8 |
uvicorn.run("multilingual.multi:app", host=host, port=port, reload=False)
|
| 9 |
|
| 10 |
|
|
|
|
| 4 |
|
| 5 |
if __name__ == "__main__":
|
| 6 |
host = os.getenv("HOST", "0.0.0.0")
|
| 7 |
+
port = int(os.getenv("PORT", "8000"))
|
| 8 |
uvicorn.run("multilingual.multi:app", host=host, port=port, reload=False)
|
| 9 |
|
| 10 |
|
multilingual/multi.py
CHANGED
|
@@ -9,6 +9,7 @@ import requests
|
|
| 9 |
import os
|
| 10 |
from models.model import translator
|
| 11 |
|
|
|
|
| 12 |
nest_asyncio.apply()
|
| 13 |
|
| 14 |
lang_codes = {
|
|
@@ -19,6 +20,8 @@ lang_codes = {
|
|
| 19 |
"sw": "swh_Latn"
|
| 20 |
}
|
| 21 |
|
|
|
|
|
|
|
| 22 |
app = FastAPI(
|
| 23 |
title="Farmlingua Translation API",
|
| 24 |
description="Multilingual ability for farmlingua supporting Igbo, Hausa, Yoruba, and Swahili",
|
|
@@ -27,7 +30,7 @@ app = FastAPI(
|
|
| 27 |
|
| 28 |
class Query(BaseModel):
|
| 29 |
text: str
|
| 30 |
-
lang: str
|
| 31 |
|
| 32 |
def translate(text: str, src: str, tgt: str) -> str:
|
| 33 |
try:
|
|
@@ -57,13 +60,18 @@ def root():
|
|
| 57 |
"""Root endpoint with API information"""
|
| 58 |
return {
|
| 59 |
"message": "Welcome to Farmlingua Translation API",
|
| 60 |
-
"version": "
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
"supported_languages": {
|
| 62 |
-
"
|
| 63 |
-
"
|
| 64 |
-
"
|
| 65 |
-
"
|
| 66 |
-
"
|
| 67 |
},
|
| 68 |
"documentation": "/docs"
|
| 69 |
}
|
|
@@ -73,112 +81,47 @@ def health_check():
|
|
| 73 |
"""Health check endpoint"""
|
| 74 |
return {"status": "healthy", "model_loaded": translator is not None}
|
| 75 |
|
| 76 |
-
# endpoints for each language
|
| 77 |
-
@app.post("/igbo")
|
| 78 |
-
def igbo_endpoint(query: Query):
|
| 79 |
-
"""Endpoint specifically for Igbo language queries"""
|
| 80 |
-
user_text = query.text
|
| 81 |
-
|
| 82 |
-
# Translate Igbo to English
|
| 83 |
-
english_text = translate(user_text, "ig", "en")
|
| 84 |
-
|
| 85 |
-
# Get response from Milestone 1
|
| 86 |
-
english_response = query_milestone1(english_text)
|
| 87 |
-
|
| 88 |
-
# Translate response back to Igbo
|
| 89 |
-
igbo_response = translate(english_response, "en", "ig")
|
| 90 |
-
|
| 91 |
-
return {
|
| 92 |
-
"language": "Igbo",
|
| 93 |
-
"original_text": user_text,
|
| 94 |
-
"english_translation": english_text,
|
| 95 |
-
"response": igbo_response
|
| 96 |
-
}
|
| 97 |
-
|
| 98 |
-
@app.post("/hausa")
|
| 99 |
-
def hausa_endpoint(query: Query):
|
| 100 |
-
"""Endpoint specifically for Hausa language queries"""
|
| 101 |
-
user_text = query.text
|
| 102 |
-
|
| 103 |
-
# Translate Hausa to English
|
| 104 |
-
english_text = translate(user_text, "ha", "en")
|
| 105 |
-
|
| 106 |
-
# Get response from Milestone 1
|
| 107 |
-
english_response = query_milestone1(english_text)
|
| 108 |
-
|
| 109 |
-
# Translate response back to Hausa
|
| 110 |
-
hausa_response = translate(english_response, "en", "ha")
|
| 111 |
-
|
| 112 |
-
return {
|
| 113 |
-
"language": "Hausa",
|
| 114 |
-
"original_text": user_text,
|
| 115 |
-
"english_translation": english_text,
|
| 116 |
-
"response": hausa_response
|
| 117 |
-
}
|
| 118 |
-
|
| 119 |
-
@app.post("/yoruba")
|
| 120 |
-
def yoruba_endpoint(query: Query):
|
| 121 |
-
"""Endpoint specifically for Yoruba language queries"""
|
| 122 |
-
user_text = query.text
|
| 123 |
-
|
| 124 |
-
# Translate Yoruba to English
|
| 125 |
-
english_text = translate(user_text, "yo", "en")
|
| 126 |
-
|
| 127 |
-
# Get response from Milestone 1
|
| 128 |
-
english_response = query_milestone1(english_text)
|
| 129 |
-
|
| 130 |
-
# Translate response back to Yoruba
|
| 131 |
-
yoruba_response = translate(english_response, "en", "yo")
|
| 132 |
-
|
| 133 |
-
return {
|
| 134 |
-
"language": "Yoruba",
|
| 135 |
-
"original_text": user_text,
|
| 136 |
-
"english_translation": english_text,
|
| 137 |
-
"response": yoruba_response
|
| 138 |
-
}
|
| 139 |
-
|
| 140 |
-
@app.post("/swahili")
|
| 141 |
-
def swahili_endpoint(query: Query):
|
| 142 |
-
"""Endpoint specifically for Swahili language queries"""
|
| 143 |
-
user_text = query.text
|
| 144 |
-
|
| 145 |
-
# Translate Swahili to English
|
| 146 |
-
english_text = translate(user_text, "sw", "en")
|
| 147 |
-
|
| 148 |
-
# Get response from Milestone 1
|
| 149 |
-
english_response = query_milestone1(english_text)
|
| 150 |
-
|
| 151 |
-
# Translate response back to Swahili
|
| 152 |
-
swahili_response = translate(english_response, "en", "sw")
|
| 153 |
-
|
| 154 |
-
return {
|
| 155 |
-
"language": "Swahili",
|
| 156 |
-
"original_text": user_text,
|
| 157 |
-
"english_translation": english_text,
|
| 158 |
-
"response": swahili_response
|
| 159 |
-
}
|
| 160 |
-
|
| 161 |
@app.post("/ask")
|
| 162 |
def multilingual_ask(query: Query):
|
|
|
|
| 163 |
user_text = query.text
|
| 164 |
user_lang = query.lang.lower()
|
| 165 |
-
|
|
|
|
| 166 |
if user_lang not in lang_codes:
|
| 167 |
return {"error": "Unsupported language. Use one of: en, ig, ha, yo, sw."}
|
| 168 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 169 |
if user_lang != "en":
|
| 170 |
english_text = translate(user_text, user_lang, "en")
|
| 171 |
else:
|
| 172 |
english_text = user_text
|
| 173 |
-
|
|
|
|
| 174 |
english_response = query_milestone1(english_text)
|
| 175 |
-
|
|
|
|
| 176 |
if user_lang != "en":
|
| 177 |
final_response = translate(english_response, "en", user_lang)
|
| 178 |
else:
|
| 179 |
final_response = english_response
|
| 180 |
-
|
| 181 |
-
return {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 182 |
|
| 183 |
# Start ngrok tunnel and run FastAPI server
|
| 184 |
if __name__ == "__main__":
|
|
|
|
| 9 |
import os
|
| 10 |
from models.model import translator
|
| 11 |
|
| 12 |
+
|
| 13 |
nest_asyncio.apply()
|
| 14 |
|
| 15 |
lang_codes = {
|
|
|
|
| 20 |
"sw": "swh_Latn"
|
| 21 |
}
|
| 22 |
|
| 23 |
+
# Simple language identifier - no auto-detection needed
|
| 24 |
+
|
| 25 |
app = FastAPI(
|
| 26 |
title="Farmlingua Translation API",
|
| 27 |
description="Multilingual ability for farmlingua supporting Igbo, Hausa, Yoruba, and Swahili",
|
|
|
|
| 30 |
|
| 31 |
class Query(BaseModel):
|
| 32 |
text: str
|
| 33 |
+
lang: str
|
| 34 |
|
| 35 |
def translate(text: str, src: str, tgt: str) -> str:
|
| 36 |
try:
|
|
|
|
| 60 |
"""Root endpoint with API information"""
|
| 61 |
return {
|
| 62 |
"message": "Welcome to Farmlingua Translation API",
|
| 63 |
+
"version": "3.0",
|
| 64 |
+
"features": {
|
| 65 |
+
"simple_language_identifier": "Specify language with 'lang' parameter",
|
| 66 |
+
"single_endpoint": "/ask - Send text with language identifier",
|
| 67 |
+
"supported_codes": "en, ig, ha, yo, sw"
|
| 68 |
+
},
|
| 69 |
"supported_languages": {
|
| 70 |
+
"en": "English",
|
| 71 |
+
"ig": "Igbo",
|
| 72 |
+
"ha": "Hausa",
|
| 73 |
+
"yo": "Yoruba",
|
| 74 |
+
"sw": "Swahili"
|
| 75 |
},
|
| 76 |
"documentation": "/docs"
|
| 77 |
}
|
|
|
|
| 81 |
"""Health check endpoint"""
|
| 82 |
return {"status": "healthy", "model_loaded": translator is not None}
|
| 83 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 84 |
@app.post("/ask")
|
| 85 |
def multilingual_ask(query: Query):
|
| 86 |
+
"""Multilingual endpoint with simple language identifier"""
|
| 87 |
user_text = query.text
|
| 88 |
user_lang = query.lang.lower()
|
| 89 |
+
|
| 90 |
+
# Validate language code
|
| 91 |
if user_lang not in lang_codes:
|
| 92 |
return {"error": "Unsupported language. Use one of: en, ig, ha, yo, sw."}
|
| 93 |
+
|
| 94 |
+
# Get language name for response
|
| 95 |
+
lang_names = {
|
| 96 |
+
"en": "English",
|
| 97 |
+
"ig": "Igbo",
|
| 98 |
+
"ha": "Hausa",
|
| 99 |
+
"yo": "Yoruba",
|
| 100 |
+
"sw": "Swahili"
|
| 101 |
+
}
|
| 102 |
+
|
| 103 |
+
# Translate to English if not English
|
| 104 |
if user_lang != "en":
|
| 105 |
english_text = translate(user_text, user_lang, "en")
|
| 106 |
else:
|
| 107 |
english_text = user_text
|
| 108 |
+
|
| 109 |
+
# Get response from Milestone 1
|
| 110 |
english_response = query_milestone1(english_text)
|
| 111 |
+
|
| 112 |
+
# Translate response back to original language if not English
|
| 113 |
if user_lang != "en":
|
| 114 |
final_response = translate(english_response, "en", user_lang)
|
| 115 |
else:
|
| 116 |
final_response = english_response
|
| 117 |
+
|
| 118 |
+
return {
|
| 119 |
+
"language": lang_names[user_lang],
|
| 120 |
+
"language_code": user_lang,
|
| 121 |
+
"original_text": user_text,
|
| 122 |
+
"english_translation": english_text,
|
| 123 |
+
"response": final_response
|
| 124 |
+
}
|
| 125 |
|
| 126 |
# Start ngrok tunnel and run FastAPI server
|
| 127 |
if __name__ == "__main__":
|
requirements.txt
CHANGED
|
@@ -5,4 +5,4 @@ requests
|
|
| 5 |
transformers
|
| 6 |
torch
|
| 7 |
nest-asyncio
|
| 8 |
-
pyngrok
|
|
|
|
| 5 |
transformers
|
| 6 |
torch
|
| 7 |
nest-asyncio
|
| 8 |
+
pyngrok
|