File size: 9,955 Bytes
3fa63a4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
from dotenv import load_dotenv
import os, asyncio, re
from crewai import Agent, Task, Crew
from langchain_openai import ChatOpenAI

try:
    from gtts import gTTS
except Exception:
    gTTS = None

try:
    from elevenlabs import ElevenLabs
except Exception:
    ElevenLabs = None

try:
    import edge_tts
except Exception:
    edge_tts = None

try:
    from TTS.api import TTS
except Exception:
    TTS = None

# load_dotenv()
# openai_api_key = os.getenv('open_ai_key_for_crew_ai')
# os.environ["OPENAI_API_KEY"] = openai_api_key

TTS_PROVIDER   = os.getenv("TTS_PROVIDER", "gtts").lower()  
ELEVEN_KEY     = os.getenv("ELEVEN_API_KEY") or os.getenv("ELEVENLABS_API_KEY")
ELEVEN_VOICE   = os.getenv("ELEVEN_VOICE", "Rachel")   
ELEVEN_MODEL   = os.getenv("ELEVEN_MODEL", "eleven_multilingual_v2")

EDGE_VOICE     = os.getenv("EDGE_VOICE", "en-US-JennyNeural")  
COQUI_MODEL    = os.getenv("COQUI_MODEL", "tts_models/multilingual/multi-dataset/xtts_v2")
OUTPUT_FILE    = "guided_meditation.mp3"

def _clean_text(text: str) -> str:
    text = re.sub(r"[*_`#>\[\](){}]", " ", str(text))
    text = re.sub(r"\s{2,}", " ", text).strip()
    return text

def _chunk(text: str, max_chars: int = 4000):
    """Split long scripts into manageable chunks for TTS providers with limits."""
    text = text.strip()
    if len(text) <= max_chars:
        yield text
        return
    parts = re.split(r"(?<=[.!?])\s+", text)
    cur = ""
    for p in parts:
        if len(cur) + len(p) + 1 <= max_chars:
            cur = (cur + " " + p).strip()
        else:
            if cur:
                yield cur
            cur = p
    if cur:
        yield cur

def tts_with_gtts(text: str, out_path: str = OUTPUT_FILE):
    if gTTS is None:
        raise RuntimeError("gTTS not installed. pip install gTTS")
    audio = None
    from tempfile import NamedTemporaryFile
    tts = gTTS(text=_clean_text(text), lang="en", slow=False)
    tts.save(out_path)
    return out_path

def tts_with_elevenlabs(text: str, out_path: str = OUTPUT_FILE, voice: str = ELEVEN_VOICE, model: str = ELEVEN_MODEL):
    if ElevenLabs is None:
        raise RuntimeError("elevenlabs SDK not installed. pip install elevenlabs")
    if not ELEVEN_KEY:
        raise RuntimeError("Set ELEVEN_API_KEY in environment to use ElevenLabs TTS.")
    client = ElevenLabs(api_key=ELEVEN_KEY)

    # ElevenLabs supports streaming; we’ll write chunks sequentially
    with open(out_path, "wb") as f:
        for chunk in _chunk(_clean_text(text), max_chars=4800):
            audio_stream = client.generate(text=chunk, voice=voice, model=model)
            for b in audio_stream:
                f.write(b)
    return out_path

async def _edge_tts_async(text: str, out_path: str, voice: str):
    communicate = edge_tts.Communicate(text=_clean_text(text), voice=voice)
    await communicate.save(out_path)

def tts_with_edge(text: str, out_path: str = OUTPUT_FILE, voice: str = EDGE_VOICE):
    if edge_tts is None:
        raise RuntimeError("edge-tts not installed. pip install edge-tts")
    asyncio.run(_edge_tts_async(text, out_path, voice))
    return out_path

def tts_with_coqui(text: str, out_path: str = OUTPUT_FILE, model_name: str = COQUI_MODEL):
    if TTS is None:
        raise RuntimeError("Coqui TTS not installed. pip install TTS")

    tts = TTS(model_name=model_name, progress_bar=False, gpu=False)
    # If you hit speed/memory issues, chunk:
    chunks = list(_chunk(_clean_text(text), max_chars=800))
    if len(chunks) == 1:
        tts.tts_to_file(text=chunks[0], file_path=out_path, language="en")
    else:
        try:
            from pydub import AudioSegment  # pip install pydub
            segs = []
            for i, ch in enumerate(chunks):
                tmp = f"__tmp_{i}.wav"
                tts.tts_to_file(text=ch, file_path=tmp, language="en")
                segs.append(AudioSegment.from_file(tmp))
                os.remove(tmp)
            final = sum(segs[1:], segs[0])
            final.export(out_path, format="mp3")
        except Exception:
            tts.tts_to_file(text=chunks[-1], file_path=out_path, language="en")
    return out_path

def synthesize_tts(text: str, out_path: str = OUTPUT_FILE) -> str:
    provider = TTS_PROVIDER
    try:
        if provider == "elevenlabs":
            return tts_with_elevenlabs(text, out_path)
        elif provider == "edge":
            return tts_with_edge(text, out_path)
        elif provider == "coqui":
            return tts_with_coqui(text, out_path)
        elif provider == "gtts":
            return tts_with_gtts(text, out_path)
        else:
            try:
                return tts_with_elevenlabs(text, out_path)
            except Exception:
                try:
                    return tts_with_edge(text, out_path)
                except Exception:
                    return tts_with_gtts(text, out_path)
    except Exception as e:
        if provider != "gtts":
            try:
                return tts_with_gTTS(text, out_path)
            except Exception:
                pass
        raise

from typing import Any

def _extract_task_output(task: Any) -> str:
    """
    CrewAI versions differ; this reads whatever shape is present.
    Returns a plain string.
    """
    out = getattr(task, "output", None)
    if out is None:
        out = getattr(task, "result", None)
        if isinstance(out, str):
            return out or ""
        return str(out) if out is not None else ""

    if isinstance(out, str):
        return out
    for attr in ("raw", "result", "final_output", "output"):
        val = getattr(out, attr, None)
        if isinstance(val, str) and val.strip():
            return val
        if val is not None:
            try:
                return str(val)
            except Exception:
                pass
    try:
        return str(out)
    except Exception:
        return ""

def task_agent_pipeline(chat_transcript, openai_api_key):
    
    os.environ["OPENAI_API_KEY"] = openai_api_key
    print("Reached crew_ai with transcript")
    print(chat_transcript)

    llm = ChatOpenAI(model="gpt-4o", temperature=0.7)  

    transcript_analysis_agent = Agent(
        name="Transcript Analyzer",
        role="Analyzes the client's chat with the AI therapist to extract emotions, key concerns, and sentiment trends.",
        goal="Extract user's emotional state and well-being indicators from the chat transcript.",
        backstory="An AI therapist assistant skilled in NLP-based sentiment and topic analysis.",
        llm=llm,
        verbose=True
    )

    plan_generator_agent = Agent(
        name="Plan Generator",
        role="Creates a personalized 1-week plan with activities, exercises, and affirmations.",
        goal="Generate a structured 7-day well-being improvement plan",
        backstory="An AI wellness coach that specializes in personalized mental health plans.",
        llm=llm,
        verbose=True
    )

    meditation_audio_agent = Agent(
        name="Meditation Generator",
        role="Creates a guided meditation script and generates an audio file for relaxation.",
        goal="Generate a calming meditation based on the user's emotional state and well-being plan.",
        backstory="An AI meditation coach that creates mindfulness and relaxation exercises.",
        llm=llm,
        verbose=True
    )
    transcript_task = Task(
        description=(
            "Analyze the chat transcript:\n"
            "{user_input}\n\n"
            "Extract key emotions, concerns, triggers, coping patterns, and sentiment trends. "
            "Output a crisp bullet summary."
        ),
        agent=transcript_analysis_agent,
        expected_output="Bullet summary of emotions/concerns/triggers/trends (no diagnosis, no PII).",
    )

    plan_task = Task(
        description=(
            "Based on the transcript summary, generate a customized 7-day well-being plan.\n"
            "- For each day, include short bullets under **Morning**, **Midday**, **Evening**.\n"
            "- Include CBT/behavioral techniques (worry time, thought labeling, activity scheduling).\n"
            "- Include sleep hygiene and anti-rumination steps.\n"
            "- Include one small social connection action daily.\n"
            "- End with a brief safety note: 'This is not medical advice. If you're in crisis, seek local help.'\n"
            "Format with markdown headers exactly as:\n"
            "## Day 1 ... ## Day 7"
        ),
        agent=plan_generator_agent,
        expected_output="Markdown with sections: ## Day 1 ... ## Day 7 plus a final Safety Note.",
        context=[transcript_task]
    )

    def generate_meditation_audio(result):
        script_text = _clean_text(str(result))
        path = synthesize_tts(script_text, out_path=OUTPUT_FILE)
        return f"Guided meditation audio generated: {path}  (provider={TTS_PROVIDER})"

    meditation_task = Task(
        description=(
            "Create a guided meditation script (≈5–7 minutes) based on the summary and plan. "
            "Tone: calm, supportive, inclusive. Avoid special characters like *."
        ),
        agent=meditation_audio_agent,
        expected_output="A short meditation script; confirm MP3 generation.",
        context=[transcript_task, plan_task],
        callback=generate_meditation_audio
    )

    wellness_crew = Crew(
        agents=[transcript_analysis_agent, plan_generator_agent, meditation_audio_agent],
        tasks=[transcript_task, plan_task, meditation_task]
    )

    # Run
    result = wellness_crew.kickoff(inputs={"user_input": chat_transcript})

    summary_text = _extract_task_output(transcript_task)
    plan_text    = _extract_task_output(plan_task)
    meditation   = _extract_task_output(meditation_task)

    return {
        "summary": summary_text,
        "plan": plan_text,
        "meditation": meditation,
        "final": getattr(result, "raw", result),  
    }