Spaces:
Running
Running
Create aib4.py
Browse files
aib4.py
ADDED
@@ -0,0 +1,450 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import requests
|
2 |
+
import json
|
3 |
+
import base64
|
4 |
+
|
5 |
+
class BhashiniClient:
|
6 |
+
"""
|
7 |
+
A client for interacting with Bhashini's ASR, NMT, and TTS services.
|
8 |
+
|
9 |
+
Methods:
|
10 |
+
list_available_languages(task_type): Lists available languages for a given task.
|
11 |
+
get_supported_voices(source_language): Gets supported genders for TTS in a language.
|
12 |
+
asr(audio_content, source_language, audio_format='wav', sampling_rate=16000): Performs ASR.
|
13 |
+
translate(text, source_language, target_language): Translates text from source to target language.
|
14 |
+
tts(text, source_language, gender='female', sampling_rate=8000): Performs TTS.
|
15 |
+
"""
|
16 |
+
|
17 |
+
PIPELINE_CONFIG_ENDPOINT = "https://meity-auth.ulcacontrib.org/ulca/apis/v0/model/getModelsPipeline"
|
18 |
+
INFERENCE_ENDPOINT = "https://dhruva-api.bhashini.gov.in/services/inference/pipeline"
|
19 |
+
PIPELINE_ID = "64392f96daac500b55c543cd"
|
20 |
+
|
21 |
+
def __init__(self, user_id, api_key, pipeline_id = PIPELINE_ID):
|
22 |
+
"""
|
23 |
+
Initializes the BhashiniClient with user credentials and pipeline ID.
|
24 |
+
|
25 |
+
Args:
|
26 |
+
user_id (str): Your user ID.
|
27 |
+
api_key (str): Your ULCA API key.
|
28 |
+
pipeline_id (str): The pipeline ID.
|
29 |
+
|
30 |
+
Raises:
|
31 |
+
Exception: If the pipeline configuration retrieval fails.
|
32 |
+
"""
|
33 |
+
self.user_id = user_id
|
34 |
+
self.api_key = api_key
|
35 |
+
self.pipeline_id = pipeline_id
|
36 |
+
self.headers = {
|
37 |
+
"Content-Type": "application/json",
|
38 |
+
"userID": self.user_id,
|
39 |
+
"ulcaApiKey": self.api_key
|
40 |
+
}
|
41 |
+
self.config = self._get_pipeline_config()
|
42 |
+
self.pipeline_data = self._parse_pipeline_config(self.config)
|
43 |
+
self.inference_api_key = self.pipeline_data['inferenceApiKey']
|
44 |
+
|
45 |
+
def _get_pipeline_config(self):
|
46 |
+
"""
|
47 |
+
Retrieves the pipeline configuration.
|
48 |
+
|
49 |
+
Returns:
|
50 |
+
dict: The pipeline configuration.
|
51 |
+
|
52 |
+
Raises:
|
53 |
+
Exception: If the request fails.
|
54 |
+
"""
|
55 |
+
payload = {
|
56 |
+
"pipelineTasks": [
|
57 |
+
{"taskType": "asr"},
|
58 |
+
{"taskType": "translation"},
|
59 |
+
{"taskType": "tts"}
|
60 |
+
],
|
61 |
+
"pipelineRequestConfig": {
|
62 |
+
"pipelineId": self.pipeline_id
|
63 |
+
}
|
64 |
+
}
|
65 |
+
response = requests.post(
|
66 |
+
self.PIPELINE_CONFIG_ENDPOINT,
|
67 |
+
headers=self.headers,
|
68 |
+
data=json.dumps(payload)
|
69 |
+
)
|
70 |
+
response.raise_for_status()
|
71 |
+
return response.json()
|
72 |
+
|
73 |
+
def _parse_pipeline_config(self, config):
|
74 |
+
"""
|
75 |
+
Parses the pipeline configuration and extracts necessary information.
|
76 |
+
|
77 |
+
Args:
|
78 |
+
config (dict): The pipeline configuration.
|
79 |
+
|
80 |
+
Returns:
|
81 |
+
dict: Parsed pipeline data.
|
82 |
+
"""
|
83 |
+
inference_api_key = config['pipelineInferenceAPIEndPoint']['inferenceApiKey']['value']
|
84 |
+
callback_url = config['pipelineInferenceAPIEndPoint']['callbackUrl']
|
85 |
+
pipeline_data = {
|
86 |
+
'asr': {},
|
87 |
+
'tts': {},
|
88 |
+
'translation': {},
|
89 |
+
'inferenceApiKey': inference_api_key,
|
90 |
+
'callbackUrl': callback_url
|
91 |
+
}
|
92 |
+
|
93 |
+
for pipeline in config['pipelineResponseConfig']:
|
94 |
+
task_type = pipeline['taskType']
|
95 |
+
if task_type in ['asr', 'translation', 'tts']:
|
96 |
+
for language_config in pipeline['config']:
|
97 |
+
source_language = language_config['language']['sourceLanguage']
|
98 |
+
|
99 |
+
if task_type != 'translation':
|
100 |
+
if source_language not in pipeline_data[task_type]:
|
101 |
+
pipeline_data[task_type][source_language] = []
|
102 |
+
|
103 |
+
language_info = {
|
104 |
+
'serviceId': language_config['serviceId'],
|
105 |
+
'sourceScriptCode': language_config['language'].get('sourceScriptCode')
|
106 |
+
}
|
107 |
+
|
108 |
+
if task_type == 'tts':
|
109 |
+
language_info['supportedVoices'] = language_config.get('supportedVoices', [])
|
110 |
+
|
111 |
+
pipeline_data[task_type][source_language].append(language_info)
|
112 |
+
else:
|
113 |
+
target_language = language_config['language']['targetLanguage']
|
114 |
+
if source_language not in pipeline_data[task_type]:
|
115 |
+
pipeline_data[task_type][source_language] = {}
|
116 |
+
|
117 |
+
if target_language not in pipeline_data[task_type][source_language]:
|
118 |
+
pipeline_data[task_type][source_language][target_language] = []
|
119 |
+
|
120 |
+
language_info = {
|
121 |
+
'serviceId': language_config['serviceId'],
|
122 |
+
'sourceScriptCode': language_config['language'].get('sourceScriptCode'),
|
123 |
+
'targetScriptCode': language_config['language'].get('targetScriptCode')
|
124 |
+
}
|
125 |
+
|
126 |
+
pipeline_data[task_type][source_language][target_language].append(language_info)
|
127 |
+
|
128 |
+
return pipeline_data
|
129 |
+
|
130 |
+
def list_available_languages(self, task_type):
|
131 |
+
"""
|
132 |
+
Lists the available languages for the specified task.
|
133 |
+
|
134 |
+
Args:
|
135 |
+
task_type (str): The task type ('asr', 'translation', or 'tts').
|
136 |
+
|
137 |
+
Returns:
|
138 |
+
list or dict: A list of available languages, or a dictionary for translation.
|
139 |
+
|
140 |
+
Raises:
|
141 |
+
ValueError: If an invalid task type is provided.
|
142 |
+
|
143 |
+
Usage Example:
|
144 |
+
client = BhashiniClient(user_id, api_key, pipeline_id)
|
145 |
+
asr_languages = client.list_available_languages('asr')
|
146 |
+
print("Available ASR Languages:", asr_languages)
|
147 |
+
|
148 |
+
translation_languages = client.list_available_languages('translation')
|
149 |
+
print("Available Translation Languages:", translation_languages)
|
150 |
+
"""
|
151 |
+
if task_type not in ['asr', 'translation', 'tts']:
|
152 |
+
raise ValueError("Invalid task type. Choose from 'asr', 'translation', or 'tts'.")
|
153 |
+
|
154 |
+
if task_type == 'translation':
|
155 |
+
languages = {}
|
156 |
+
for src_lang in self.pipeline_data['translation']:
|
157 |
+
languages[src_lang] = list(self.pipeline_data['translation'][src_lang].keys())
|
158 |
+
return languages
|
159 |
+
else:
|
160 |
+
return list(self.pipeline_data[task_type].keys())
|
161 |
+
|
162 |
+
def get_supported_voices(self, source_language):
|
163 |
+
"""
|
164 |
+
Returns the supported genders for TTS in the specified language.
|
165 |
+
|
166 |
+
Args:
|
167 |
+
source_language (str): The language code (e.g., 'hi' for Hindi).
|
168 |
+
|
169 |
+
Returns:
|
170 |
+
list: A list of supported genders (e.g., ['male', 'female']).
|
171 |
+
|
172 |
+
Raises:
|
173 |
+
ValueError: If TTS is not supported for the language.
|
174 |
+
|
175 |
+
Usage Example:
|
176 |
+
client = BhashiniClient(user_id, api_key, pipeline_id)
|
177 |
+
voices = client.get_supported_voices('hi')
|
178 |
+
print("Supported voices for Hindi TTS:", voices)
|
179 |
+
"""
|
180 |
+
if source_language not in self.pipeline_data['tts']:
|
181 |
+
available_languages = ', '.join(self.list_available_languages('tts'))
|
182 |
+
raise ValueError(
|
183 |
+
f"TTS not supported for language '{source_language}'. "
|
184 |
+
f"Available languages: {available_languages}"
|
185 |
+
)
|
186 |
+
|
187 |
+
service_info = self.pipeline_data['tts'][source_language][0]
|
188 |
+
supported_voices = service_info.get('supportedVoices', [])
|
189 |
+
return supported_voices
|
190 |
+
|
191 |
+
|
192 |
+
def asr(self, audio_content, source_language, audio_format='wav', sampling_rate=16000):
|
193 |
+
"""
|
194 |
+
Performs Automatic Speech Recognition on the provided audio content.
|
195 |
+
|
196 |
+
Args:
|
197 |
+
audio_content (bytes): The audio content in bytes.
|
198 |
+
source_language (str): The language code of the audio (e.g., 'hi' for Hindi).
|
199 |
+
audio_format (str): supported formats of audio content: ('wav', 'mp3', 'flac', 'ogg'.)
|
200 |
+
sampling_rate (int): The sampling rate of the audio in Hz.
|
201 |
+
|
202 |
+
Returns:
|
203 |
+
dict: The ASR response from the API.
|
204 |
+
|
205 |
+
Raises:
|
206 |
+
ValueError: If the language is not supported.
|
207 |
+
Exception: If the API request fails.
|
208 |
+
|
209 |
+
Usage Example:
|
210 |
+
client = BhashiniClient(user_id, api_key, pipeline_id)
|
211 |
+
with open('audio.wav', 'rb') as f:
|
212 |
+
audio_content = f.read()
|
213 |
+
asr_result = client.asr(audio_content, source_language='hi', audio_format='wav')
|
214 |
+
print("ASR Result:", asr_result)
|
215 |
+
"""
|
216 |
+
if source_language not in self.pipeline_data['asr']:
|
217 |
+
available_languages = ', '.join(self.list_available_languages('asr'))
|
218 |
+
raise ValueError(
|
219 |
+
f"ASR not supported for language '{source_language}'. "
|
220 |
+
f"Available languages: {available_languages}"
|
221 |
+
)
|
222 |
+
|
223 |
+
service_info = self.pipeline_data['asr'][source_language][0]
|
224 |
+
service_id = service_info['serviceId']
|
225 |
+
|
226 |
+
payload = {
|
227 |
+
"pipelineTasks": [
|
228 |
+
{
|
229 |
+
"taskType": "asr",
|
230 |
+
"config": {
|
231 |
+
"language": {
|
232 |
+
"sourceLanguage": source_language
|
233 |
+
},
|
234 |
+
"serviceId": service_id,
|
235 |
+
"audioFormat": audio_format,
|
236 |
+
"samplingRate": sampling_rate
|
237 |
+
}
|
238 |
+
}
|
239 |
+
],
|
240 |
+
"inputData": {
|
241 |
+
"audio": [
|
242 |
+
{
|
243 |
+
"audioContent": base64.b64encode(audio_content).decode('utf-8')
|
244 |
+
}
|
245 |
+
]
|
246 |
+
}
|
247 |
+
}
|
248 |
+
|
249 |
+
headers = {
|
250 |
+
'Accept': '*/*',
|
251 |
+
'Authorization': self.inference_api_key,
|
252 |
+
'Content-Type': 'application/json'
|
253 |
+
}
|
254 |
+
|
255 |
+
response = requests.post(
|
256 |
+
self.INFERENCE_ENDPOINT,
|
257 |
+
headers=headers,
|
258 |
+
data=json.dumps(payload)
|
259 |
+
)
|
260 |
+
|
261 |
+
self._handle_response_errors(response)
|
262 |
+
return response.json()
|
263 |
+
|
264 |
+
def translate(self, text, source_language, target_language):
|
265 |
+
"""
|
266 |
+
Translates the provided text from the source language to the target language.
|
267 |
+
|
268 |
+
Args:
|
269 |
+
text (str): The text to translate.
|
270 |
+
source_language (str): The source language code.
|
271 |
+
target_language (str): The target language code.
|
272 |
+
|
273 |
+
Returns:
|
274 |
+
dict: The translation response from the API.
|
275 |
+
|
276 |
+
Raises:
|
277 |
+
ValueError: If the language pair is not supported.
|
278 |
+
Exception: If the API request fails.
|
279 |
+
|
280 |
+
Usage Example:
|
281 |
+
client = BhashiniClient(user_id, api_key, pipeline_id)
|
282 |
+
translation_result = client.translate(
|
283 |
+
'मेरा नाम विहिर है।',
|
284 |
+
source_language='hi',
|
285 |
+
target_language='gu'
|
286 |
+
)
|
287 |
+
print("Translation Result:", translation_result)
|
288 |
+
"""
|
289 |
+
if source_language not in self.pipeline_data['translation']:
|
290 |
+
available_languages = ', '.join(self.list_available_languages('translation').keys())
|
291 |
+
raise ValueError(
|
292 |
+
f"Translation not supported from language '{source_language}'. "
|
293 |
+
f"Available source languages: {available_languages}"
|
294 |
+
)
|
295 |
+
|
296 |
+
if target_language not in self.pipeline_data['translation'][source_language]:
|
297 |
+
available_targets = ', '.join(self.pipeline_data['translation'][source_language].keys())
|
298 |
+
raise ValueError(
|
299 |
+
f"Translation from '{source_language}' to '{target_language}' not supported. "
|
300 |
+
f"Available target languages for '{source_language}': {available_targets}"
|
301 |
+
)
|
302 |
+
|
303 |
+
service_info = self.pipeline_data['translation'][source_language][target_language][0]
|
304 |
+
service_id = service_info['serviceId']
|
305 |
+
|
306 |
+
payload = {
|
307 |
+
"pipelineTasks": [
|
308 |
+
{
|
309 |
+
"taskType": "translation",
|
310 |
+
"config": {
|
311 |
+
"language": {
|
312 |
+
"sourceLanguage": source_language,
|
313 |
+
"targetLanguage": target_language
|
314 |
+
},
|
315 |
+
"serviceId": service_id
|
316 |
+
}
|
317 |
+
}
|
318 |
+
],
|
319 |
+
"inputData": {
|
320 |
+
"input": [
|
321 |
+
{
|
322 |
+
"source": text
|
323 |
+
}
|
324 |
+
]
|
325 |
+
}
|
326 |
+
}
|
327 |
+
|
328 |
+
headers = {
|
329 |
+
'Accept': '*/*',
|
330 |
+
'Authorization': self.inference_api_key,
|
331 |
+
'Content-Type': 'application/json'
|
332 |
+
}
|
333 |
+
|
334 |
+
response = requests.post(
|
335 |
+
self.INFERENCE_ENDPOINT,
|
336 |
+
headers=headers,
|
337 |
+
data=json.dumps(payload)
|
338 |
+
)
|
339 |
+
|
340 |
+
self._handle_response_errors(response)
|
341 |
+
return response.json()
|
342 |
+
|
343 |
+
def tts(self, text, source_language, gender='female', sampling_rate=8000):
|
344 |
+
"""
|
345 |
+
Converts the provided text to speech in the specified language.
|
346 |
+
|
347 |
+
Args:
|
348 |
+
text (str): The text to convert to speech.
|
349 |
+
source_language (str): The language code of the text.
|
350 |
+
gender (str): The desired voice gender ('male' or 'female').
|
351 |
+
sampling_rate (int): The sampling rate in Hz.
|
352 |
+
|
353 |
+
Returns:
|
354 |
+
dict: The TTS response from the API.
|
355 |
+
|
356 |
+
Raises:
|
357 |
+
ValueError: If the language or gender is not supported.
|
358 |
+
Exception: If the API request fails.
|
359 |
+
|
360 |
+
Usage Example:
|
361 |
+
client = BhashiniClient(user_id, api_key, pipeline_id)
|
362 |
+
tts_result = client.tts(
|
363 |
+
'હેલો વર્લ્ડ',
|
364 |
+
source_language='gu',
|
365 |
+
gender='female'
|
366 |
+
)
|
367 |
+
# Save the audio output
|
368 |
+
audio_base64 = tts_result['pipelineResponse'][0]['audio'][0]['audioContent']
|
369 |
+
audio_data = base64.b64decode(audio_base64)
|
370 |
+
with open('output_audio.wav', 'wb') as f:
|
371 |
+
f.write(audio_data)
|
372 |
+
"""
|
373 |
+
if source_language not in self.pipeline_data['tts']:
|
374 |
+
available_languages = ', '.join(self.list_available_languages('tts'))
|
375 |
+
raise ValueError(
|
376 |
+
f"TTS not supported for language '{source_language}'. "
|
377 |
+
f"Available languages: {available_languages}"
|
378 |
+
)
|
379 |
+
|
380 |
+
service_info = self.pipeline_data['tts'][source_language][0]
|
381 |
+
service_id = service_info['serviceId']
|
382 |
+
supported_voices = service_info.get('supportedVoices', [])
|
383 |
+
|
384 |
+
if gender not in ['male', 'female']:
|
385 |
+
raise ValueError("Gender must be 'male' or 'female'.")
|
386 |
+
|
387 |
+
if supported_voices and gender not in supported_voices:
|
388 |
+
available_genders = ', '.join(supported_voices)
|
389 |
+
raise ValueError(
|
390 |
+
f"Gender '{gender}' not supported for language '{source_language}'. "
|
391 |
+
f"Available genders: {available_genders}"
|
392 |
+
)
|
393 |
+
|
394 |
+
payload = {
|
395 |
+
"pipelineTasks": [
|
396 |
+
{
|
397 |
+
"taskType": "tts",
|
398 |
+
"config": {
|
399 |
+
"language": {
|
400 |
+
"sourceLanguage": source_language
|
401 |
+
},
|
402 |
+
"serviceId": service_id,
|
403 |
+
"gender": gender,
|
404 |
+
"samplingRate": sampling_rate
|
405 |
+
}
|
406 |
+
}
|
407 |
+
],
|
408 |
+
"inputData": {
|
409 |
+
"input": [
|
410 |
+
{
|
411 |
+
"source": text
|
412 |
+
}
|
413 |
+
]
|
414 |
+
}
|
415 |
+
}
|
416 |
+
|
417 |
+
headers = {
|
418 |
+
'Accept': '*/*',
|
419 |
+
'Authorization': self.inference_api_key,
|
420 |
+
'Content-Type': 'application/json'
|
421 |
+
}
|
422 |
+
|
423 |
+
response = requests.post(
|
424 |
+
self.INFERENCE_ENDPOINT,
|
425 |
+
headers=headers,
|
426 |
+
data=json.dumps(payload)
|
427 |
+
)
|
428 |
+
|
429 |
+
self._handle_response_errors(response)
|
430 |
+
return response.json()
|
431 |
+
|
432 |
+
def _handle_response_errors(self, response):
|
433 |
+
"""
|
434 |
+
Handles errors in the response.
|
435 |
+
|
436 |
+
Args:
|
437 |
+
response (requests.Response): The response object.
|
438 |
+
|
439 |
+
Raises:
|
440 |
+
Exception: If an HTTP error occurs.
|
441 |
+
"""
|
442 |
+
try:
|
443 |
+
response.raise_for_status()
|
444 |
+
except requests.HTTPError as http_err:
|
445 |
+
try:
|
446 |
+
error_info = response.json()
|
447 |
+
error_message = error_info.get('message', 'An error occurred.')
|
448 |
+
except json.JSONDecodeError:
|
449 |
+
error_message = response.text
|
450 |
+
raise Exception(f"HTTP error occurred: {error_message}") from http_err
|