File size: 7,263 Bytes
73bb3e6 |
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 |
import os
from crewai.tools import BaseTool
from crewai.tools import tool
from transformers import pipeline
from backend.crew_ai.data_retriever_util import get_user_profile
from backend.crew_ai.config import get_config
import psycopg2
from psycopg2.extras import RealDictCursor
from typing import ClassVar
from langchain_huggingface import HuggingFaceEmbeddings
from transformers import pipeline
from gradio_client import Client
class MentalHealthTools:
"""Tools for mental health chatbot"""
@tool("Bhutanese Helplines")
def get_bhutanese_helplines() -> str:
"""
Retrieves Bhutanese mental health helplines from the PostgreSQL `resources` table.
"""
try:
db_uri = os.getenv("SUPABASE_DB_URI")
if not db_uri:
raise ValueError("SUPABASE_DB_URI not set in environment")
conn = psycopg2.connect(db_uri)
cursor = conn.cursor(cursor_factory=RealDictCursor)
query = """
SELECT name, description, phone, website, address, operation_hours
FROM resources
"""
cursor.execute(query)
helplines = cursor.fetchall()
if not helplines:
return "No helplines found in the database."
response = "📞 Bhutanese Mental Health Helplines:\n"
for h in helplines:
response += f"\n📌 {h['name']}"
if h['description']:
response += f"\n Description: {h['description']}"
if h['phone']:
response += f"\n 📱 Phone: {h['phone']}"
if h['website']:
response += f"\n 🌐 Website: {h['website']}"
if h['address']:
response += f"\n 🏠 Address: {h['address']}"
if h['operation_hours']:
response += f"\n ⏰ Hours: {h['operation_hours']}"
response += "\n"
cursor.close()
conn.close()
return response.strip()
except Exception as e:
return f"⚠️ Failed to fetch helplines from DB: {str(e)}"
class CrisisClassifierTool(BaseTool):
name: str = "Crisis Classifier"
description: str = (
"A tool that classifies text into predefined categories. "
"Input should be the text to classify."
)
def _run(self, text: str) -> str:
"""
Classifies the given text using the Hugging Face model.
Returns the classification label and score.
"""
try:
# Initialize the pipeline here (will happen on every tool call)
classifier = pipeline("sentiment-analysis", model="sentinet/suicidality")
result = classifier(text)
if result:
label = result[0]['label']
score = result[0]['score']
return f"Classification: {label} (Score: {score:.4f})"
return "Could not classify the text."
except Exception as e:
return f"Error during text classification: {e}"
class MentalConditionClassifierTool(BaseTool):
name: str = "Mental condition Classifier"
description: str = (
"A tool that classifies text into predefined categories. "
"Input should be the text to classify."
)
# Class-level cache for the client
_client = None
def _get_client(self):
if self._client is None:
self.__class__._client = Client("ety89/mental_health_text_classifiaction") # ✅ fixed typo
return self._client
def _run(self, text: str) -> str:
"""
Classifies the given text using the Hugging Face model.
Returns the classification label and score.
"""
try:
# Initialize the pipeline here (will happen on every tool call)
client = Client("ety89/mental_health_text_classifiaction")
result = client.predict(
input_text=text,
api_name="/predict"
)
if result:
label = result.split(':')[-2].split('(')[-2].strip()
score = result.split(':')[-1].strip(')').strip()
return label, score
return "Could not classify the text."
except Exception as e:
return f"Error during text classification: {e}"
class DataRetrievalTool(BaseTool):
name: str = "Data Retrieval"
description: str = (
"A tool that fetched the user profile data from the database. "
"Input should be User Profile ID."
)
def _run(self, user_profile_id: str) -> str:
"""
Fetches the user profile data from the database using the user profile ID.
Returns the user profile information or an error message.
"""
try:
config = get_config()
if user_profile_id.strip() == "anon_user":
return config['default_user_profile']
# Retrieve user profile using the utility function
user_profile = get_user_profile(user_profile_id)
if user_profile:
return f"User Profile: {user_profile}"
return "User profile not found."
except Exception as e:
return f"Error retrieving user profile: {e}"
class QueryVectorStoreTool(BaseTool):
name: str = "Query Vector Store"
description: str = (
"Queries the Supabase-hosted PostgreSQL vector database with a user query and classified condition, "
"and retrieves the top 3 most relevant documents."
)
# Shared across all instances
embedding_model: ClassVar = HuggingFaceEmbeddings(
model_name="sentence-transformers/all-MiniLM-L6-v2"
)
def _run(self, user_query: str, classified_condition: str) -> dict:
query_text = f"{user_query} Condition: {classified_condition}"
embedding = self.embedding_model.embed_query(query_text)
db_uri = os.getenv("SUPABASE_DB_URI")
if not db_uri:
raise ValueError("SUPABASE_DB_URI not set in environment")
conn = psycopg2.connect(db_uri)
cursor = conn.cursor()
cursor.execute("""
SELECT ac.chunk_text, a.title, a.topic, a.source, ac.embedding <-> %s::vector AS score
FROM article_chunks ac
JOIN articles a ON ac.doc_id = a.id
ORDER BY score
LIMIT 3;
""", (embedding,))
rows = cursor.fetchall()
docs = [
{
"text": row[0],
"title": row[1],
"topic": row[2],
"source": row[3],
"score": row[4]
}
for row in rows
]
cursor.close()
conn.close()
return {"docs": docs}
def _arun(self, *args, **kwargs):
raise NotImplementedError("Async version not implemented")
|