1+ import asyncio
2+ import getpass
3+
4+ import pyaudio
5+ import oracledb
6+ import oci
7+ from oci .config import from_file
8+ from oci .auth .signers .security_token_signer import SecurityTokenSigner
9+ from oci .ai_speech_realtime import (
10+ RealtimeClient ,
11+ RealtimeClientListener ,
12+ RealtimeParameters ,
13+ )
14+
15+ pw = getpass .getpass ("Enter database user password:" )
16+
17+ # Use this when making a connection with a wallet
18+ connection = oracledb .connect (
19+ user = "moviestream" ,
20+ password = pw ,
21+ dsn = "selectaidb_high" ,
22+ config_dir = "/Users/pparkins/Downloads/Wallet_SelectAIDB" ,
23+ wallet_location = "/Users/pparkins/Downloads/Wallet_SelectAIDB"
24+ )
25+ print ("Successfully connected to Oracle Database" )
26+ print (f"Connection details: { connection } )
27+
28+ # Create a FIFO queue
29+ queue = asyncio .Queue ()
30+
31+ # Set audio parameters
32+ SAMPLE_RATE = 16000
33+ FORMAT = pyaudio .paInt16
34+ CHANNELS = 1
35+ BUFFER_DURATION_MS = 96
36+
37+ # Calculate the number of frames per buffer
38+ FRAMES_PER_BUFFER = int (SAMPLE_RATE * BUFFER_DURATION_MS / 1000 )
39+
40+ # Variables to keep track of results and state
41+ cummulativeResult = ""
42+ isSelect = False
43+ last_result_time = None
44+
45+ def authenticator ():
46+ config = from_file ("~/.oci/config" , "paulspeechai" )
47+ with open (config ["security_token_file" ], "r" ) as f :
48+ token = f .readline ()
49+ private_key = oci .signer .load_private_key_from_file (config ["key_file" ])
50+ auth = SecurityTokenSigner (token = token , private_key = private_key )
51+ return auth
52+
53+ def audio_callback (in_data , frame_count , time_info , status ):
54+ # This function will be called by PyAudio when there's new audio data
55+ queue .put_nowait (in_data )
56+ return (None , pyaudio .paContinue )
57+
58+ p = pyaudio .PyAudio ()
59+
60+ # Open the stream
61+ stream = p .open (
62+ format = FORMAT ,
63+ channels = CHANNELS ,
64+ rate = SAMPLE_RATE ,
65+ input = True ,
66+ frames_per_buffer = FRAMES_PER_BUFFER ,
67+ stream_callback = audio_callback ,
68+ )
69+
70+ stream .start_stream ()
71+ config = from_file ()
72+
73+ async def send_audio (client ):
74+ while True :
75+ data = await queue .get ()
76+ # Send it over the websocket
77+ await client .send_data (data )
78+
79+ class SpeechListener (RealtimeClientListener ):
80+ def on_result (self , result ):
81+ global cummulativeResult , isSelect , last_result_time
82+ if result ["transcriptions" ][0 ]["isFinal" ]:
83+ transcription = result ['transcriptions' ][0 ]['transcription' ]
84+ cummulativeResult += transcription
85+ print (f"Received final results: { transcription } )
86+ print (f"Current cummulative result: { cummulativeResult } )
87+ if cummulativeResult .lower ().startswith ("select ai" ):
88+ isSelect = True
89+ last_result_time = asyncio .get_event_loop ().time ()
90+ else :
91+ print (f"Received partial results: { result ['transcriptions' ][0 ]['transcription' ]} )
92+
93+ def on_ack_message (self , ackmessage ):
94+ return super ().on_ack_message (ackmessage )
95+
96+ def on_connect (self ):
97+ return super ().on_connect ()
98+
99+ def on_connect_message (self , connectmessage ):
100+ return super ().on_connect_message (connectmessage )
101+
102+ def on_network_event (self , ackmessage ):
103+ return super ().on_network_event (ackmessage )
104+
105+ def on_error (self ):
106+ return super ().on_error ()
107+
108+ async def check_idle ():
109+ global last_result_time , isSelect
110+ while True :
111+ if isSelect and last_result_time and (asyncio .get_event_loop ().time () - last_result_time > 2 ):
112+ executeSelectAI ()
113+ isSelect = False
114+ await asyncio .sleep (1 )
115+
116+ def executeSelectAI ():
117+ global cummulativeResult
118+ print (f"executeSelectAI called cummulative result: { cummulativeResult } )
119+ # for example prompt => 'select ai I am looking for the top 5 selling movies for the latest month please',
120+ query = """SELECT DBMS_CLOUD_AI.GENERATE(
121+ prompt => :prompt,
122+ profile_name => 'openai_gpt35',
123+ action => 'narrate')
124+ FROM dual"""
125+ with connection .cursor () as cursor :
126+ cursor .execute (query , prompt = cummulativeResult )
127+ result = cursor .fetchone ()
128+ if result and isinstance (result [0 ], oracledb .LOB ):
129+ text_result = result [0 ].read ()
130+ print (text_result )
131+ else :
132+ print (result )
133+ # Reset cumulativeResult after execution
134+ cummulativeResult = ""
135+
136+
137+ # logic such as the following could be added to make the app further dynamic as far as action type...
138+ # actionValue = 'narrate'
139+ # if cummulativeResult.lower().startswith("select ai narrate"):
140+ # actionValue = "narrate"
141+ # elif cummulativeResult.lower().startswith("select ai chat"):
142+ # actionValue = "chat"
143+ # elif cummulativeResult.lower().startswith("select ai showsql"):
144+ # actionValue = "showsql"
145+ # elif cummulativeResult.lower().startswith("select ai show sql"):
146+ # actionValue = "showsql"
147+ # elif cummulativeResult.lower().startswith("select ai runsql"):
148+ # actionValue = "runsql"
149+ # elif cummulativeResult.lower().startswith("select ai run sql"):
150+ # actionValue = "runsql"
151+ # # Note that "runsql" is not currently supported as action value
152+ # query = """SELECT DBMS_CLOUD_AI.GENERATE(
153+ # prompt => :prompt,
154+ # profile_name => 'openai_gpt35',
155+ # action => :actionValue)
156+ # FROM dual"""
157+
158+ if __name__ == "__main__" :
159+ # Run the event loop
160+ def message_callback (message ):
161+ print (f"Received message: { message } )
162+
163+ realtime_speech_parameters : RealtimeParameters = RealtimeParameters ()
164+ realtime_speech_parameters .language_code = "en-US"
165+ realtime_speech_parameters .model_domain = (
166+ realtime_speech_parameters .MODEL_DOMAIN_GENERIC
167+ )
168+ realtime_speech_parameters .partial_silence_threshold_in_ms = 0
169+ realtime_speech_parameters .final_silence_threshold_in_ms = 2000
170+ realtime_speech_parameters .should_ignore_invalid_customizations = False
171+ realtime_speech_parameters .stabilize_partial_results = (
172+ realtime_speech_parameters .STABILIZE_PARTIAL_RESULTS_NONE
173+ )
174+
175+ realtime_speech_url = "wss://realtime.aiservice.us-phoenix-1.oci.oraclecloud.com"
176+ client = RealtimeClient (
177+ config = config ,
178+ realtime_speech_parameters = realtime_speech_parameters ,
179+ listener = SpeechListener (),
180+ service_endpoint = realtime_speech_url ,
181+ signer = authenticator (),
182+ compartment_id = "ocid1.compartment.oc1..MYCOMPARMENTID" ,
183+ )
184+
185+ loop = asyncio .get_event_loop ()
186+ loop .create_task (send_audio (client ))
187+ loop .create_task (check_idle ())
188+ loop .run_until_complete (client .connect ())
189+
190+ if stream .is_active ():
191+ stream .close ()
192+
193+ print ("Closed now" )
0 commit comments