llm_tts_stt_conversation/double_trouble.py at main · IP050/llm_tts_stt_conversation · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
from RealtimeTTS import TextToAudioStream, CoquiEngine, SystemEngine
from langchain_ollama import ChatOllama
import time

def install(voice=None):
    engine = CoquiEngine(voice=voice,pretrained=True)
    return engine

def install_system(voice):
    engine = SystemEngine(voice)
    return engine

def system_prompt(prompt, user_message):
    return [
        ("system", prompt),
        ("user", user_message)
    ]

if __name__ == '__main__':
    # Initialize the TTS engine and LLMs
    print("Initializing")
    engine = install()
    enginetwo = install(voice="bane.wav")
    print("created engine")
    stream = TextToAudioStream(engine)
    second = TextToAudioStream(enginetwo)
    llm1 = ChatOllama(model="bob:latest", base_url="http://localhost:11434", num_predict=-2, disable_streaming=True)
    llm2 = ChatOllama(model="lisa:latest", base_url="http://localhost:11434", num_predict=-2, disable_streaming=True)

    # Initialize chat memory and system prompt
    chat_memory = []
    initial_prompt = input("Prompt bot one:")
    initial_prompttwo = input("Prompt bot two")

    print("We here")
    user_input = input("User:")

    while True:
        try:
            # LLM1 interaction
            messages = system_prompt(initial_prompt, user_input)
            llm1_response = llm1.invoke(messages)
            response1 = llm1_response.content
            chat_memory.extend((("user", user_input), ("Assistant1", response1)))
            # Play LLM1's response
            stream.feed(response1)

            stream.play_async(output_wavfile="newtests.wav")
            print(f"LLM1: {response1}")
            while stream.is_playing():
                time.sleep(0.1)

            #engine.set_voice("bane.wav")
            # Pass LLM1's response to LLM2
            messages = system_prompt(initial_prompt, response1)
            llm2_response = llm2.invoke(messages)
            response2 = llm2_response.content
            chat_memory.append(("Assistant2", response2))

            # Play LLM2's response
            second.feed(response2)
            second.play_async(output_wavfile="newtests2.wav")
            print(f"LLM2: {response2}")
            while second.is_playing():
                time.sleep(0.1)

            # Update user_input with LLM2's response for the next loop iteration
            user_input = response2

            # Optional: Add a stopping condition if the conversation should not be infinite
            if "stop" in response2.lower():
                break

        except KeyboardInterrupt:
            engine.shutdown()
            break