Aleix Conchillo Flaqué
5bb64098e7
transports(inputs): don't queue incoming system frames
2024-07-24 14:35:00 -07:00
Aleix Conchillo Flaqué
b583f5162b
transports(daily): fix completion callbacks handling
2024-07-23 15:25:59 -07:00
Aleix Conchillo Flaqué
ad88925154
vad(silero): expose cache and repo parameters
2024-07-23 12:12:28 -07:00
Aleix Conchillo Flaqué
08e0722d97
fix initial metrics format
2024-07-23 11:23:03 -07:00
Aleix Conchillo Flaqué
05d4fba551
processors(rtvi): send initial empty metrics
2024-07-23 11:22:41 -07:00
Aleix Conchillo Flaqué
f41c2b3c9f
transports(daily): don't send empty metrics
2024-07-23 11:22:41 -07:00
Aleix Conchillo Flaqué
69f64899fe
pipeline: add send_initial_empty_metrics flag
2024-07-23 11:22:41 -07:00
Aleix Conchillo Flaqué
ad5b9202ab
transports(daily): wait until start|stop_transcription are finished
...
Fixes #305
2024-07-22 22:59:30 -07:00
Aleix Conchillo Flaqué
0852b50b8f
services: transcription timestamp should use ISO8601 format
2024-07-22 22:40:28 -07:00
Aleix Conchillo Flaqué
087fe9a537
services(cartesia): fix TTFB
2024-07-22 15:30:16 -07:00
Aleix Conchillo Flaqué
c1170260b5
processors(rtvi): use generic LLM and TTS names
2024-07-22 15:27:33 -07:00
Aleix Conchillo Flaqué
65cdf50774
processors(rtvi): fix task cleanup
2024-07-22 15:01:45 -07:00
Aleix Conchillo Flaqué
9233bb490c
processors(rtvi): add support for "tts-text" messages
2024-07-22 11:40:17 -07:00
Aleix Conchillo Flaqué
43932220f7
processors(rtvi): use only user-transcription
2024-07-22 09:40:16 -07:00
Aleix Conchillo Flaqué
cea4d1894e
processors(rtvi): change voice before LLM updates
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
80baa0358d
processors(rtvi): lable is now rtvi
2024-07-22 09:32:18 -07:00
Chad Bailey
5d73db53a0
initial pseudo function calling
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
302ea90dce
processors(rtvi): messages now require an id
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
37b04ed283
processors(rtvi): use send a type=response as command responses
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
be6995cfdf
processors(rtvi): renamed realtime-ai to rtvi
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
dfbc11300c
processors(realtime-ai): use label instead of tag
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
82d539d174
processors(realtime-ai): add support for interrupting the bot
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
6fbf98d8e2
processors(realtime-ai): llm-context now uses a data field
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
f094c42728
processors(realtime-ai): add transcription messages
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
13827e1282
processors(realtime-ai): send a successful response for every command
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
32170b47d9
processors(realtime-ai): add user-[start|stopped]-speaking messages
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
09c05354c2
processors(realtime-ai): fix voice initialization
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
b0b1475563
processors(realtime-ai): add support making TTS to speak
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
b85dd7283a
processors(realtime-ai): add support for appending to the LLM context
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
846ae765e5
services(TTSService): fix sentence cleanup
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
4c629e538e
processors(realtime-ai): add assistant before output transport
...
Cartesia can do word-to-word output instead of full sentences. This means that
for properly adding things into the context we need to add it before the
transport, otherwise some words might be lost.
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
f6e22bb3b9
processors(realtime-ai): add silero vad to the transport
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
46a048d7f6
processors(realtime-ai): allow default setup to be None
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
bd9f4eea06
processors(realtime-ai): provide default values
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
0a672e61e2
processors(realtime-ai): update it to use groq by default
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
29a8530221
processors(realtime-ai): add support for updating config (model, voice...)
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
3e738642a7
processors(realtime-ai): add support for getting/updating LLM context
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
9f012c8002
processors: add new RealtimeAIProcessor
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
0a69a9e5ef
transport(daily): also accept TransportMessageFrame
2024-07-22 09:32:18 -07:00
Aleix Conchillo Flaqué
194790183a
processor: add support for setting a processor parent
2024-07-22 09:32:18 -07:00
weedge
e5363f65f0
fix: push_frame use TextFrame
...
Signed-off-by: weedge <weege007@gmail.com >
2024-07-22 17:29:06 +08:00
Lewis Wolfgang
f9fdadb4c0
Allow openai http connections to remain open in the pool indefinitely.
...
Rather than expiring in 5 seconds.
2024-07-18 11:18:21 -04:00
Aleix Conchillo Flaqué
37027f68cb
remove LLMResponseStartFrame and LLMResponseEndFrame
...
This was added in the past to properly handle interruptions for the
LLMAssistantContextAggregator. But this is not necessary anymore since we can
handle interruptions by just processing the StartInterruptionFrame, so there's
no need for these extra frames.
2024-07-17 20:53:35 -07:00
Kwindla Hultman Kramer
355fe01cb7
fixed forgotten renames
2024-07-17 20:28:27 -07:00
Kwindla Hultman Kramer
fa53c67606
comments re fixes
2024-07-17 18:30:45 -07:00
Kwindla Hultman Kramer
2204b8e205
cartesia streaming and context management via word-level timestamps
2024-07-17 15:17:00 -07:00
Kwindla Hultman Kramer
270007b17c
wip - using cartesia word timestamps for context management
2024-07-17 14:13:52 -07:00
Kwindla Hultman Kramer
568eb2ef4c
cartesia websockets and streaming
2024-07-17 14:13:52 -07:00
Kwindla Hultman Kramer
73ca9184a8
wip cartesia continuation (not working yet)
2024-07-17 14:13:52 -07:00
TomTom101
da2082b025
chore: Combined combinable lookaheads
2024-07-06 11:11:40 +02:00
