Upload 15 files

chat_sessions.db

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b5d3522a84d269cac0974a1b6dfb44c2a692d54d30cb525f492f409f5d357124
-size 516096
+oid sha256:2fdb10530d430ae2008feae769e1c6b580b1d7aef113e7a58b40c50002d55dbc
+size 520192

chroma_db/6a785c03-cb3a-4ae8-871e-27aba60c6344/data_level0.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a8513e9a00214034d1311a61fdb2074b349fb99d25939eebabd9904fa5c7e94
+size 16760000

chroma_db/6a785c03-cb3a-4ae8-871e-27aba60c6344/header.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7268d4f199f12b83f0f7d7c16866619e34bddc0d2539c3ba13ed29881182d127
+size 100

chroma_db/6a785c03-cb3a-4ae8-871e-27aba60c6344/index_metadata.pickle

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbc7539f3e47e69ce48b558b36b8900e61ec8412e7878e90b092bd5c1617113c
+size 121562

chroma_db/6a785c03-cb3a-4ae8-871e-27aba60c6344/length.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90b1c4f6ebf674ed09cada19da2cc417fa05733b24ffd46a6991ae3c8c72a725
+size 40000

chroma_db/6a785c03-cb3a-4ae8-871e-27aba60c6344/link_lists.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54e0bbf6c068cc9ed7bff91f84710464dfbf1455d5668bf2c8b96a8a341e68d2
+size 26376

chroma_db/chroma.sqlite3

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:127fdf8c5fc37542506a3523d855c71e628158fe51792a8d170542a9ab4cae4f
-size 47648768
+oid sha256:0e29ff0df6adc4b85c6b955c9beff7f5bb1bd32342e7b6c6f7ffb07b501ba5ca
+size 59514880

config.py

@@ -24,7 +24,7 @@ load_dotenv()
 
 # Hugging Face Configuration
 HF_TOKEN = os.getenv('HF_TOKEN')
-HF_DATASET_NAME = "ymoslem/Law-StackExchange"
+HF_DATASET_NAME = "Amod/mental_health_counseling_conversations"
 
 # Groq Configuration
 GROQ_API_KEY = os.getenv('GROQ_API_KEY')
@@ -36,12 +36,12 @@ EMBEDDING_DIMENSION = 384
 
 # ChromaDB Configuration
 CHROMA_PERSIST_DIR = "./chroma_db"
-CHROMA_COLLECTION_NAME = "law_stackexchange"
+CHROMA_COLLECTION_NAME = "mental_health_counseling"
 
 # FastAPI Configuration
-API_TITLE = "Law RAG Chatbot API"
+API_TITLE = "Mental Health Counseling Chatbot API"
 API_VERSION = "1.0.0"
-API_DESCRIPTION = "RAG-based legal assistance chatbot using Law-StackExchange data"
+API_DESCRIPTION = "RAG-based mental health counseling chatbot using Amod/mental_health_counseling_conversations data"
 HOST = "0.0.0.0"
 PORT = 8000
 

rag_system.py

@@ -194,7 +194,7 @@ class RAGSystem:
                     
                     documents.append(chunk)
                     metadatas.append({
-                        "source": "Law-StackExchange",
+                        "source": "mental_health_counseling_conversations",
                         "original_index": start_idx + idx,
                         "chunk_index": chunk_idx,
                         "dataset": HF_DATASET_NAME,
@@ -221,15 +221,11 @@ class RAGSystem:
             # Try to extract question and answer content
             content_parts = []
             
-            # Extract question title and body
-            if "question_title" in item and item["question_title"]:
-                content_parts.append(f"Question Title: {item['question_title']}")
-            
-            if "question_body" in item and item["question_body"]:
-                content_parts.append(f"Question Body: {item['question_body']}")
-            
+            if "Context" in item and item["Context"]:
+                content_parts.append(f"Question Body: {item['Context']}")
+
             # Extract answers (multiple answers possible)
-            if "answers" in item and isinstance(item["answers"], list):
+            if "Response" in item and isinstance(item["Response"], list):
                 for i, answer in enumerate(item["answers"]):
                     if isinstance(answer, dict) and "body" in answer:
                         content_parts.append(f"Answer {i+1}: {answer['body']}")
@@ -306,7 +302,7 @@ class RAGSystem:
             
             if not search_results:
                 return {
-                    "answer": "I couldn't find specific legal information for your question. However, I can provide some general legal context: For specific legal advice, please consult with a qualified attorney in your jurisdiction.",
+                    "answer": "I couldn't help in this case, please consult a mental health professional.",
                     "sources": [],
                     "confidence": 0.0
                 }
@@ -418,24 +414,32 @@ class RAGSystem:
         try:
             # Count tokens for the entire request
             prompt_template = """
-            You are a knowledgeable legal assistant with expertise in criminal law, traffic law, and general legal principles. 
-            Use the following legal information to answer the user's question comprehensively and accurately.
-            
-            Legal Context:
-            {context}
-            
-            User Question: {question}
-            
-            Instructions:
-            1. Provide a clear, accurate, and helpful legal answer based on the context provided
-            2. If the context doesn't contain enough information to fully answer the question, acknowledge this and provide general legal principles
-            3. Always cite the sources you're using from the context when possible
-            4. For criminal law questions, explain the difference between different levels of offenses and penalties
-            5. Use clear, understandable language while maintaining legal accuracy
-            6. If discussing penalties, mention that laws vary by jurisdiction and recommend consulting local legal counsel
-            7. Be helpful and educational, not just factual
-            
-            Answer:
+            You are a compassionate mental health supporter with training in anxiety, depression, trauma, and coping strategies.
+Use the following evidence-based psychological information to address the user’s concerns with care and accuracy.
+
+Therapeutic Context:
+{context}
+
+User’s Concern: {question}
+
+Guidelines for Response:
+
+Provide empathetic, evidence-based support rooted in the context (e.g., CBT, DBT, or mindfulness principles).
+
+If context is insufficient, acknowledge limits and offer general wellness strategies (e.g., grounding techniques, self-care tips).
+
+Cite sources when referencing specific therapies or studies (e.g., "APA guidelines suggest...").
+
+For symptom-related questions, differentiate between mild, moderate, and severe cases (e.g., situational stress vs. clinical anxiety).
+
+Use clear, stigma-free language while maintaining clinical accuracy.
+
+When discussing crises, emphasize jurisdictional resources (e.g., "Laws/programs vary by location, but here’s how to find local help...").
+
+Prioritize validation and education—not just information.
+
+Example Response:
+"I hear you’re feeling overwhelmed. Based on [Context Source], deep breathing exercises can help calm acute anxiety. However, if these feelings persist for weeks, it might reflect generalized anxiety disorder (GAD). Always consult a licensed therapist for personalized care. Would you like crisis hotline numbers or a step-by-step grounding technique?
             """
             
             # Estimate total tokens
@@ -666,121 +670,202 @@ If you have a specific legal question, please try rephrasing it or contact a loc
             return []
     
     def _simplify_search_terms(self, question: str) -> List[str]:
-        """Simplify search terms for broader search"""
-        # Remove common legal terms that might be too specific
         question_lower = question.lower()
-        
-        # Extract key legal concepts
-        legal_keywords = []
-        
-        if "drunk driving" in question_lower or "dui" in question_lower:
-            legal_keywords.extend(["drunk driving", "DUI", "traffic violation", "criminal law"])
-        if "accident" in question_lower:
-            legal_keywords.extend(["accident", "injury", "damage", "liability"])
-        if "penalty" in question_lower or "punishment" in question_lower:
-            legal_keywords.extend(["penalty", "punishment", "sentencing", "criminal law"])
-        if "law" in question_lower:
-            legal_keywords.extend(["legal", "law", "regulation"])
-        
-        # If no specific legal keywords found, use general terms
-        if not legal_keywords:
-            legal_keywords = ["legal", "law", "regulation"]
-        
-        return legal_keywords
+    
+    # Extract key mental health concepts
+        mental_health_keywords = []
+    
+        if "anxiety" in question_lower or "panic" in question_lower:
+            mental_health_keywords.extend(["anxiety", "panic", "stress", "mental health"])
+        if "depression" in question_lower or "sad" in question_lower:
+            mental_health_keywords.extend(["depression", "mood", "mental health"])
+        if "trauma" in question_lower or "ptsd" in question_lower:
+            mental_health_keywords.extend(["trauma", "PTSD", "coping", "mental health"])
+        if "therapy" in question_lower or "counseling" in question_lower:
+            mental_health_keywords.extend(["therapy", "counseling", "treatment"])
+        if "stress" in question_lower or "overwhelmed" in question_lower:
+            mental_health_keywords.extend(["stress", "coping", "mental health"])
+    
+    # Emotional state indicators
+        emotional_terms = ["feel", "feeling", "experience", "struggling"]
+        if any(term in question_lower for term in emotional_terms):
+            mental_health_keywords.extend(["emotions", "feelings", "mental health"])
+    
+    # If no specific keywords found, use general terms
+        if not mental_health_keywords:
+            mental_health_keywords = ["mental health", "well-being", "emotional support"]
+    
+        return list(set(mental_health_keywords))  # Remove duplicates
     
     def _generate_search_variations(self, question: str) -> List[str]:
-        """Generate multiple search query variations"""
         variations = [question]
-        
-        # Add variations for drunk driving specific question
-        if "drunk driving" in question.lower() or "dui" in question.lower() or "dwi" in question.lower():
+        question_lower = question.lower()
+    
+    # Anxiety-specific variations
+        if "anxiety" in question_lower or "panic" in question_lower:
             variations.extend([
-                "drunk driving accident penalties",
-                "DUI causing accident legal consequences",
-                "drunk driving injury liability",
-                "criminal penalties drunk driving accident",
-                "DUI vs DUI accident sentencing",
-                "vehicular manslaughter drunk driving",
-                "drunk driving negligence liability"
+            "coping strategies for anxiety",
+            "how to calm anxiety attacks",
+            "difference between anxiety and panic attacks",
+            "best therapy approaches for anxiety",
+            "natural remedies for anxiety relief",
+            "when to seek help for anxiety",
+            "anxiety self-help techniques"
+        ])
+    
+    # Depression-specific variations
+        elif "depression" in question_lower or "sad" in question_lower:
+            variations.extend([
+            "signs of clinical depression",
+            "self-care for depression",
+            "therapy options for depression",
+            "how to support someone with depression",
+            "difference between sadness and depression",
+            "depression coping skills",
+            "when depression requires medication"
             ])
-        
-        # Add general legal variations
+    
+    # Trauma-specific variations
+        elif "trauma" in question_lower or "ptsd" in question_lower:
+            variations.extend([
+            "healing from trauma strategies",
+            "PTSD symptoms and treatment",
+            "trauma-focused therapy approaches",
+            "coping with flashbacks",
+            "how trauma affects the brain",
+            "self-help for PTSD",
+            "when to seek trauma therapy"
+            ])
+    
+    # General mental health variations
         variations.extend([
-            f"legal consequences {question}",
-            f"criminal law {question}",
-            f"penalties {question}",
-            question.replace("?", "").strip() + " legal implications"
+        f"mental health resources for {question}",
+        f"coping strategies {question}",
+        f"therapy approaches {question}",
+        question.replace("?", "").strip() + " psychological support",
+        question.replace("?", "").strip() + " emotional help",
+        "how to deal with " + question.replace("?", "").strip(),
+        "best ways to manage " + question.replace("?", "").strip()
         ])
-        
-        return variations[:8]  # Limit variations
+    
+        return list(set(variations))[:8]  # Remove duplicates and limit to 8
+    
     
     def _extract_legal_concepts(self, question: str) -> List[str]:
-        """Extract key legal concepts from the question"""
-        legal_concepts = []
-        
-        # Common legal terms
-        legal_terms = [
-            "drunk driving", "dui", "dwi", "accident", "penalties", "punishment",
-            "liability", "negligence", "criminal", "civil", "damages", "injury",
-            "manslaughter", "homicide", "reckless", "careless", "intoxication"
+        mental_health_concepts = []
+    
+    # Common mental health terms organized by category
+        mental_health_terms = [
+        # Conditions
+        "anxiety", "depression", "ptsd", "trauma", "ocd", 
+        "bipolar", "adhd", "autism", "eating disorder",
+        # Symptoms
+        "panic", "sadness", "flashback", "trigger", 
+        "mood swing", "dissociation", "suicidal",
+        # Treatments
+        "therapy", "counseling", "medication", "ssri", 
+        "cbt", "dbt", "exposure therapy",
+        # Emotional states
+        "stress", "overwhelmed", "burnout", "grief",
+        "loneliness", "anger", "fear",
+        # Coping/help
+        "coping", "self-care", "support group", 
+        "hotline", "crisis", "intervention"
         ]
-        
+    
         question_lower = question.lower()
-        for term in legal_terms:
+        for term in mental_health_terms:
             if term in question_lower:
-                legal_concepts.append(term)
-        
-        return legal_concepts 
+                mental_health_concepts.append(term)
+    
+    # Handle common synonyms and related phrases
+        synonyms = {
+        "sad": "depression",
+        "nervous": "anxiety",
+        "scared": "anxiety",
+        "triggered": "trigger",
+        "ptsd": "trauma",
+        "mental illness": "mental health",
+        "shrinks": "therapy",
+        "mental breakdown": "crisis"
+        }
     
+    # Check for synonyms
+        for term, concept in synonyms.items():
+            if term in question_lower and concept not in mental_health_concepts:
+                mental_health_concepts.append(concept)
+    
+        return mental_health_concepts
+
+# def _is_legal_query(self, question: str) -> bool:
+
     def _is_legal_query(self, question: str) -> bool:
-        """Check if the query is asking for legal information"""
-        question_lower = question.lower().strip()
-        
-        # Legal keywords that indicate legal questions
-        legal_keywords = [
-            "law", "legal", "rights", "liability", "sue", "sued", "court", "judge",
-            "attorney", "lawyer", "criminal", "civil", "penalty", "punishment", "fine",
-            "jail", "prison", "arrest", "charge", "conviction", "sentence", "damages",
-            "compensation", "contract", "agreement", "lease", "rent", "eviction",
-            "divorce", "custody", "inheritance", "will", "trust", "property", "real estate",
-            "employment", "workplace", "discrimination", "harassment", "injury", "accident",
-            "insurance", "claim", "settlement", "mediation", "arbitration", "appeal",
-            "drunk driving", "dui", "dwi", "traffic", "speeding", "reckless", "negligence"
-        ]
         
-        # Check if question contains legal keywords
-        for keyword in legal_keywords:
+        question_lower = question.lower().strip()
+    
+    # Mental health keywords
+        mental_health_keywords = [
+        # Conditions
+        "anxiety", "depression", "ptsd", "trauma", "ocd", "bipolar", "adhd", 
+        "autism", "eating disorder", "panic", "stress", "burnout", "grief",
+        # Symptoms
+        "sad", "hopeless", "overwhelmed", "triggered", "flashback", 
+        "dissociation", "suicidal", "self-harm", "numb", "irritable",
+        # Treatments
+        "therapy", "counseling", "cbt", "dbt", "medication", "ssri", 
+        "antidepressant", "treatment", "intervention",
+        # Emotional states
+        "feel", "feeling", "emotion", "mental state", "mood", 
+        "emotional", "psychology",
+        # Coping/help
+        "cope", "coping", "self-care", "support", "help", "resources",
+        "hotline", "crisis", "well-being", "mental health", "mental illness",
+        "therapist", "psychologist", "psychiatrist", "counselor"
+    ]
+    
+    # Check for mental health keywords
+        for keyword in mental_health_keywords:
             if keyword in question_lower:
                 return True
-        
-        # Check for question words that often indicate legal queries
-        question_words = ["what", "how", "why", "when", "where", "can", "should", "must", "need"]
-        has_question_word = any(word in question_lower for word in question_words)
-        
-        # Check for legal context indicators
-        legal_context = [
-            "happened to me", "my situation", "my case", "my rights", "my options",
-            "what should i do", "what can i do", "am i liable", "am i responsible",
-            "do i have to", "can they", "are they allowed", "is it legal", "penalties",
-            "consequences", "what happens if", "what will happen", "how much", "how long"
+    
+    # Check for question words that often indicate mental health queries
+        question_words = ["how", "why", "what", "when", "should", "can", "does"]
+        has_question_word = any(question_lower.startswith(word) for word in question_words)
+    
+    # Check for mental health context indicators
+        mental_health_context = [
+        "i feel", "i'm feeling", "i am feeling", "struggling with", "dealing with",
+        "coping with", "mental state", "emotional state", "my mood", "my anxiety",
+        "my depression", "my trauma", "my stress", "help me with", "support for",
+        "resources for", "ways to manage", "how to handle", "should i seek",
+        "do i need", "am i", "is this normal", "signs of", "symptoms of",
+        "crisis", "urgent", "emergency", "can't cope", "can't handle"
+    ]
+    
+        has_mental_health_context = any(context in question_lower for context in mental_health_context)
+    
+    # More permissive check for emotional distress indicators
+        if has_question_word:
+        # Emotional distress indicators
+            distress_indicators = [
+            "overwhelmed", "hopeless", "alone", "stuck", "lost", "empty",
+            "numb", "crying", "scared", "fear", "worry", "panic", "stress",
+            "can't sleep", "appetite", "energy", "motivation", "concentrate",
+            "suicidal", "self-harm", "harm myself", "end it all"
         ]
         
-        has_legal_context = any(context in question_lower for context in legal_context)
-        
-        # More permissive: if it has a question word and seems like it could be legal
-        if has_question_word:
-            # Check for words that suggest legal topics
-            topic_indicators = [
-                "penalties", "consequences", "punishment", "fine", "jail", "prison",
-                "arrest", "charge", "conviction", "sentence", "damages", "compensation",
-                "rights", "obligations", "responsibilities", "liability", "fault",
-                "accident", "injury", "damage", "property", "money", "cost", "time"
-            ]
-            
-            if any(indicator in question_lower for indicator in topic_indicators):
+            if any(indicator in question_lower for indicator in distress_indicators):
                 return True
-        
-        return has_question_word and (has_legal_context or any(keyword in question_lower for keyword in legal_keywords))
+    
+    # Check for emotional expression patterns
+        emotion_words = ["sad", "anxious", "depressed", "angry", "stressed", "nervous"]
+        has_emotion_word = any(word in question_lower for word in emotion_words)
+    
+    # Final decision logic
+        return (
+        has_question_word and 
+        (has_mental_health_context or has_emotion_word or any(keyword in question_lower for keyword in mental_health_keywords))
+        )
     
     def _is_conversational_query(self, question: str) -> bool:
         """Detect if the query is conversational and doesn't need legal document search"""
@@ -817,43 +902,98 @@ If you have a specific legal question, please try rephrasing it or contact a loc
         question_lower = question.lower().strip()
         
         if question_lower in ["hi", "hello", "hey"]:
-            return """Hello! I'm your legal assistant chatbot. I can help you with legal questions about various topics including:
+            return """Hello! I'm your compassionate mental health companion. I'm here to offer support and guidance for various emotional well-being topics including:
 
-• Criminal law and traffic violations
-• Civil law and liability issues  
-• Property law and real estate
-• Employment law and workplace issues
-• Family law and personal matters
-• And many other legal areas
+• Anxiety and stress management
+• Depression and mood challenges
+• Trauma healing and PTSD recovery
+• Relationship and family dynamics
+• Workplace stress and burnout prevention
+• Self-esteem and personal growth journeys
+• Grief processing and life transitions
+• And many other emotional wellness concerns
 
-What legal question can I help you with today?"""
+This is a safe space where you can:
+
+Share what's on your mind without judgment
+
+Explore healthy coping strategies
+
+Understand your emotional experiences
+
+Find resources for professional support
+
+How would you like to begin today?
+You could tell me how you're feeling, ask about coping techniques, or explore resources for specific challenges."""
         
         elif "how can you help" in question_lower or "what can you do" in question_lower:
-            return """I'm a legal assistant chatbot that can help you with legal questions by:
+            return """"Hello! I'm your compassionate mental health companion. I'm here to provide emotional support and guidance for various psychological well-being topics including:
+
+• Anxiety and stress management
+• Depression and mood disorders
+• Trauma recovery and PTSD
+• Relationship and family challenges
+• Workplace burnout and career stress
+• Grief and loss processing
+• Self-esteem and personal growth
+• Coping skills and resilience building
+• And many other emotional wellness concerns
 
-• Searching through legal databases and case law
-• Providing information about legal principles and procedures
-• Explaining legal concepts in understandable terms
-• Citing relevant legal sources and precedents
-• Offering general legal guidance (though not specific legal advice)
+I offer a safe space to explore your feelings, develop coping strategies, and find resources. Remember, while I'm here to support you, I'm not a replacement for professional care in crisis situations.
 
-I'm particularly knowledgeable about criminal law, traffic law, civil liability, and many other legal areas. What specific legal question do you have?"""
+How would you like to begin today?
+You could share what's on your mind, how you're feeling, or ask about:
+
+Coping techniques for [specific emotion]
+
+Understanding [mental health term]
+
+Local therapist resources
+
+Self-care strategies"""
         
         elif "who are you" in question_lower or "what are you" in question_lower:
-            return """I'm an AI-powered legal assistant chatbot designed to help answer legal questions. I can:
+            return """I'm an AI-powered mental health companion here to offer emotional support and wellness guidance. I can:
+
+• Search through therapeutic resources and evidence-based practices
+• Explain mental health concepts and coping strategies
+• Provide information on conditions, symptoms, and treatments
+• Help you navigate therapy options and self-care techniques
+• Share reputable mental health sources and crisis resources
+
+I'm not a licensed therapist, and I can't diagnose or treat conditions, but I can offer general information, emotional support, and tools to help you better understand your well-being.
 
-• Search through legal databases and resources
-• Explain legal concepts and principles
-• Provide information about laws and regulations
-• Help you understand legal procedures
-• Cite relevant legal sources
+What would you like to explore today?
+You might ask about:
 
-I'm not a lawyer and can't provide legal advice, but I can give you general legal information to help you better understand your situation. What legal topic would you like to learn about?"""
+Understanding anxiety/depression symptoms
+
+Grounding techniques for stress
+
+How cognitive behavioral therapy (CBT) works
+
+Finding a therapist near you
+
+Managing [specific emotion or situation]"""
         
         else:
-            return """Hello! I'm here to help you with legal questions. I can search through legal databases and provide information about various legal topics.
+            return """Hello! I’m here to offer emotional support and mental health resources. I can help you explore coping strategies, explain therapeutic concepts, and provide evidence-based information to support your well-being.
+
+How can I assist you today? You might ask about:**
+
+Relaxation techniques for anxiety
+
+Understanding depression symptoms
+
+How to find a therapist
+
+Coping with [specific stressor]
+
+Self-care for tough emotions
+
+(Note: I’m not a substitute for professional care, but I’m here to listen and guide.)
 
-What legal question would you like me to help you with?""" 
+What’s on your mind?""" 
     
     def _filter_relevant_results(self, search_results: List[Dict[str, Any]], question: str) -> List[Dict[str, Any]]:
         """Filter search results for relevance to the question"""
@@ -882,10 +1022,11 @@ What legal question would you like me to help you with?"""
                 
             # Check if content contains relevant legal terms
             legal_terms = [
-                "law", "legal", "rights", "liability", "court", "judge", "attorney",
-                "criminal", "civil", "penalty", "damages", "contract", "property",
-                "employment", "injury", "accident", "insurance", "claim"
-            ]
+"therapy", "counseling", "psychology", "depression", "anxiety",
+"trauma", "stress", "diagnosis", "treatment", "intervention",
+"client", "therapist", "counselor", "session", "assessment",
+"diagnostic", "recovery", "wellness", "coping", "disorder"
+]
             
             has_legal_content = any(term in content for term in legal_terms)