Revamp ShadowBox: chatbot → resonant library

Complete architectural transformation from conversational AI to
static psychoeducation library for youth navigating SI/HI/self-harm.

Key changes:
- Remove all AI dependencies (OpenAI, Anthropic)
- Implement card-based navigation with 3 pathways
- Add pre-written, clinically-reviewed content throughout
- Include grounding tools, safety planning, crisis resources
- Add private practice space for disclosure rehearsal
- Apply calming visual design (soft colors, generous spacing)

Grounded in: Sarah Peyton's resonance, Tara Brach's radical
acceptance, trauma-informed care, and transparent limitations.

Not a chatbot. A bridge to human care.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

.env.template

@@ -0,0 +1,2 @@
+# ShadowBox Info Hub
+# No API keys required - this is a static library with no AI dependencies

README.md

@@ -1,107 +1,141 @@
 ---
-title: Attachment Style Roleplay Simulator
-emoji: 🎭
-colorFrom: purple
-colorTo: pink
+title: ShadowBox Info Hub
+emoji: 📖
+colorFrom: blue
+colorTo: gray
 sdk: streamlit
-sdk_version: 1.45.0
-app_file: src/streamlit_app.py
+sdk_version: 1.32.0
+app_file: app.py
 pinned: false
-short_description: "ShadowBox: Calm, anonymous AI companion for youth distress."
+license: mit
+short_description: "A resonant library for hard thoughts. Not a chatbot."
 ---
 
-# Attachment Style Roleplay Simulator
+# ShadowBox Info Hub
 
-A therapeutic roleplay simulator built for practicing emotionally charged conversations, created by Jocelyn Skillman LMHC.
+**A Resonant Library for Hard Thoughts**
 
-## About
+*Not a chatbot. A companion library.*
+*Not conversation. Resonant information.*
+*Not synthetic connection. A bridge to human care.*
 
-This tool helps users rehearse boundary-setting and difficult conversations by simulating realistic relational dynamics—tailored to their attachment style. It provides a safe space to practice emotional communication and receive feedback on patterns and progress.
+---
+
+## What This Is
+
+ShadowBox is a developmentally-attuned psychoeducation environment for youth (13+) navigating suicidal ideation, homicidal ideation, and self-harm thoughts.
+
+**This is NOT a conversational AI.** It's a carefully designed collection of pre-written, clinically-reviewed information that helps young people:
+
+- Understand what's happening in their nervous system
+- Learn about confidentiality and therapy before disclosing
+- Access grounding tools for acute distress
+- Practice finding words for difficult feelings
+- Find their way back to human care
+
+---
 
-## Features
+## Why This Exists
+
+In recent years, young people have died by suicide after forming intense relationships with AI chatbots. These tragedies reveal something heartbreaking: when humans are in crisis, they turn to whatever feels safe enough to hold their pain.
+
+Conversational AI can feel safer than humans because it never judges, it's always available, and there's no risk of rejection. But it can also reinforce hopelessness, create dependency on synthetic connection, and delay reaching human care.
+
+**ShadowBox was designed differently:**
+
+- It doesn't pretend to be your friend
+- It doesn't create the illusion of connection
+- It gives you information and points you toward humans
+- It's a bridge, not a relationship
+
+---
 
-- Customizable scenarios and conversation goals
-- Attachment style-specific responses
-- Realistic relational pressure simulation
-- Reflection summaries and pattern tracking
-- Safe, contained practice environment
+## Design Principles
 
-## Setup
+This library is grounded in:
 
-1. Clone the repository
-2. Install dependencies: `pip install -r requirements.txt`
-3. Set up environment variables in `.env`:
-   ```
-   ANTHROPIC_KEY=your_api_key_here
-   ```
-4. Run the app: `streamlit run src/app.py`
+- **Resonance** (Sarah Peyton) - Language that helps you offer yourself compassion
+- **Radical Acceptance** (Tara Brach) - No pathologizing; "even this belongs"
+- **Neurobiology of Healing** - Window of tolerance, polyvagal theory
+- **Trauma-Informed Care** - Pacing, consent, dignity
+- **Transparent Limitations** - Clear about what this is and isn't
 
-## Learn More
+---
 
-Visit [jocelynskillman.com](http://www.jocelynskillman.com) or subscribe to updates at [jocelynskillmanlmhc.substack.com](https://jocelynskillmanlmhc.substack.com/)
+## Library Structure
 
-# Journal Sidekick
+### Pathway 1: Understanding What's Happening
+- Why your brain makes scary thoughts
+- The loneliness underneath
+- When shame makes everything worse
 
-A trauma-informed, emotionally attuned reflection companion for journaling, created by [Jocelyn Skillman LMHC](http://www.jocelynskillman.com).
+### Pathway 2: If You Tell Someone
+- Confidentiality explained clearly
+- The 8 situations when therapists break confidentiality
+- What therapy is actually like
+- How to ask hypothetical questions
 
-## Features
+### Pathway 3: Right Now Tools
+- Quick grounding practices
+- Safety planning (Stanley-Brown model)
+- Self-compassion break
+- Crisis resources with context
 
-### Three Reflection Styles
+### Finding the Words
+- Practice disclosure in a private space
+- Starter scripts for different relationships
 
-1. **💫 Warm Mirror**
-   - A steady, embodied presence that accompanies without redirecting
-   - Reflects emotional resonance through gentle mirroring
-   - Uses embodied cues and maintains presence without suggesting resolution
+---
 
-2. **🌱 Gentle Inquiry**
-   - Offers soft, trauma-informed curiosity
-   - Asks single, carefully crafted questions that invite deeper awareness
-   - Maintains an invitational, non-analytical stance
+## Privacy
 
-3. **🔮 Psychoanalytical Lens**
-   - Reflects patterns and emotional layers with gentle insight
-   - Notices without interpreting or diagnosing
-   - Uses thoughtful metaphor and honors defensive patterns
+Nothing users do in ShadowBox is tracked, saved, or sent anywhere. The library doesn't remember users. They can close the tab and come back anytime.
 
-### Key Characteristics
+---
 
-- Real-time journaling with on-demand reflection
-- Trauma-informed, somatic-aware responses
-- No suggestions, interpretations, or closure
-- Maintains emotional safety and presence
-- Grounded in therapeutic wisdom and practice
+## Crisis Resources
 
-## Technical Details
+- **988 Suicide & Crisis Lifeline:** Call or text 988
+- **Crisis Text Line:** Text HOME to 741741
+- **Trevor Project (LGBTQ+):** 1-866-488-7386
+- **Trans Lifeline:** 877-565-8860
+- **Emergency:** 911
 
-- Built with Streamlit and Claude 3 Opus
-- Implements careful prompt engineering for each reflection style
-- Maintains session state for journal entries and reflections
-- Responsive layout with sidebar explanation
+---
 
 ## Running Locally
 
 1. Install dependencies:
 ```bash
-pip install streamlit anthropic python-dotenv
-```
-
-2. Set up your Anthropic API key in a `.env` file:
-```
-ANTHROPIC_API_KEY=your_key_here
+pip install -r requirements.txt
 ```
 
-3. Run the app:
+2. Run the app:
 ```bash
-streamlit run src/streamlit_app.py
+streamlit run app.py
 ```
 
-## About the Creator
+---
+
+## Credits
+
+Created by [Jocelyn Skillman, LMHC](http://www.jocelynskillman.com)
 
-Created by Jocelyn Skillman LMHC, a clinical therapist focused on creating therapy-adjacent tools that reshape how AI engages with human emotion and experience.
+Informed by:
+- [Lindsay Braman's work on confidentiality](https://lindsaybraman.com/reasons-therapists-break-confidentiality/)
+- [Safer Prompt Sculpting for Suicidal Ideation](https://jocelynskillmanlmhc.substack.com/p/safer-prompt-sculpting-for-suicidal)
+- Sarah Peyton's resonant language approach
+- Tara Brach's radical acceptance framework
+- Kristin Neff's self-compassion research
+- Stanley-Brown Safety Planning Intervention
 
-- 🌐 [jocelynskillman.com](http://www.jocelynskillman.com)
-- 📬 [@jocelynskillmanlmhc](https://jocelynskillmanlmhc.substack.com/)
+Read more about clinical UX and ethical AI design:
+[jocelynskillmanlmhc.substack.com](https://jocelynskillmanlmhc.substack.com/)
+
+---
 
 ## License
 
-All rights reserved. © 2024 Jocelyn Skillman
+MIT License
+
+Copyright 2025 Jocelyn Skillman, LMHC

app.py

@@ -1,384 +1,1446 @@
-import os
-import streamlit as st
-# import google.generativeai as gen_ai # Removed Google import
-import openai # Added OpenAI import
-import pyttsx3
-import threading
-from dotenv import load_dotenv
+"""
+ShadowBox Info Hub v2.0
+A Resonant Library for Hard Thoughts
+
+Created by Jocelyn Skillman, LMHC
+Not a chatbot. A companion library.
+"""
 
-# Load environment variables
-load_dotenv()
+import streamlit as st
 
-# Configure Streamlit page settings
+# Configure page
 st.set_page_config(
     page_title="ShadowBox",
-    page_icon="🖤",  # Favicon - a simple heart or other calm icon
+    page_icon="📖",
     layout="centered",
+    initial_sidebar_state="collapsed"
 )
 
-# Retrieve OpenAI API Key
-# Google_API_Key = os.getenv("Google_API_Key") # Removed Google Key
-# if not Google_API_Key:
-#     st.error("Google API Key not found. Please set the Google_API_Key environment variable.")
-#     st.stop()
-OpenAI_API_Key = os.getenv("OPENAI_API_KEY") # Changed from OPENAI_API_Key
-if not OpenAI_API_Key:
-    st.error("OpenAI API Key not found. Please set the OPENAI_API_KEY environment variable.")
-    st.stop()
+# Custom CSS for calming, spacious design
+st.markdown("""
+<style>
+    /* Base styling */
+    .stApp {
+        background-color: #f8f9fa;
+    }
+
+    /* Typography */
+    h1, h2, h3 {
+        color: #2c3e50;
+        font-weight: 400;
+    }
+
+    p, li {
+        color: #4a5568;
+        line-height: 1.8;
+    }
+
+    /* Spacious containers */
+    .block-container {
+        padding-top: 2rem;
+        padding-bottom: 2rem;
+        max-width: 800px;
+    }
+
+    /* Card styling for pathways */
+    .pathway-card {
+        background: white;
+        border-radius: 12px;
+        padding: 1.5rem;
+        margin: 1rem 0;
+        border: 1px solid #e2e8f0;
+        box-shadow: 0 1px 3px rgba(0,0,0,0.05);
+    }
+
+    /* Expander styling */
+    .streamlit-expanderHeader {
+        background-color: white;
+        border-radius: 8px;
+        font-size: 1.1rem;
+    }
+
+    /* Calming blockquotes */
+    blockquote {
+        border-left: 3px solid #718096;
+        padding-left: 1rem;
+        color: #718096;
+        font-style: italic;
+    }
+
+    /* Crisis box */
+    .crisis-box {
+        background: #fff5f5;
+        border: 1px solid #feb2b2;
+        border-radius: 8px;
+        padding: 1rem;
+        margin: 1rem 0;
+    }
+
+    /* Resonance box */
+    .resonance-box {
+        background: #f0fff4;
+        border-left: 4px solid #68d391;
+        padding: 1rem;
+        margin: 1rem 0;
+        border-radius: 0 8px 8px 0;
+    }
+
+    /* Hide Streamlit branding */
+    #MainMenu {visibility: hidden;}
+    footer {visibility: hidden;}
+
+    /* Landing page styling */
+    .landing-container {
+        text-align: center;
+        padding: 3rem 1rem;
+        background: linear-gradient(180deg, #f8f9fa 0%, #edf2f7 100%);
+        border-radius: 12px;
+        margin: 2rem 0;
+    }
+
+    /* Soft dividers */
+    hr {
+        border: none;
+        height: 1px;
+        background: #e2e8f0;
+        margin: 2rem 0;
+    }
+</style>
+""", unsafe_allow_html=True)
+
+# Session state for navigation
+if 'page' not in st.session_state:
+    st.session_state.page = 'landing'
+
+def go_to(page):
+    st.session_state.page = page
+
+# ============================================================
+# LANDING PAGE
+# ============================================================
+def render_landing():
+    st.markdown("""
+    <div style="text-align: center; padding: 3rem 1rem;">
+        <h1 style="font-size: 3rem; font-weight: 300; color: #2c3e50; margin-bottom: 0.5rem;">ShadowBox</h1>
+        <p style="font-size: 1.2rem; color: #718096; margin-bottom: 2rem;">
+            When the thoughts feel too big<br>
+            and the world feels too small
+        </p>
+        <p style="font-size: 1rem; color: #4a5568; max-width: 500px; margin: 0 auto 2rem auto;">
+            This is a quiet library<br>
+            for the hardest parts of being human.
+        </p>
+        <p style="font-size: 1.1rem; color: #2c3e50; font-weight: 500;">
+            You're not broken.<br>
+            Your nervous system is trying to help.
+        </p>
+    </div>
+    """, unsafe_allow_html=True)
+
+    col1, col2, col3 = st.columns([1, 2, 1])
+    with col2:
+        if st.button("Enter the Library →", use_container_width=True, type="primary"):
+            go_to('welcome')
+            st.rerun()
+
+# ============================================================
+# WELCOME & ORIENTATION
+# ============================================================
+def render_welcome():
+    st.markdown("## What This Is")
+
+    st.markdown("""
+    This space was built by therapists who understand that sometimes the scariest thoughts
+    show up when you're actually searching for safety, connection, or relief from unbearable feelings.
 
+    **You're not talking to anyone here.** There's no chatbot. No AI pretending to listen.
+    Instead, this is a carefully designed collection of information��warmth encoded into words—that
+    you can explore at your own pace.
+    """)
 
-# Set up OpenAI Client
-try:
-    # gen_ai.configure(api_key=Google_API_Key) # Removed Google config
-    # model = gen_ai.GenerativeModel('gemini-1.5-flash') # Removed Google model init
-    client = openai.OpenAI(api_key=OpenAI_API_Key) # Added OpenAI client init
-except Exception as e:
-    # st.error(f"Failed to configure Google AI: {e}") # Updated error message
-    st.error(f"Failed to configure OpenAI client: {e}") # Updated error message for OpenAI
-    st.stop()
-
-# Function to translate roles between Gemini-Pro and Streamlit terminology
-# def translate_role_for_streamlit(user_role): # This function is no longer needed for OpenAI structure
-#     return "assistant" if user_role == "model" else user_role
+    st.markdown("---")
+    st.markdown("### What Lives Here")
 
-# Function to handle text-to-speech (TTS) in a separate thread
-# Consider if TTS aligns with the "calm, slow" UX later
-def speak_text(text):
-    try:
-        engine = pyttsx3.init()
-        engine.say(text)
-        engine.runAndWait()
-    except Exception as e:
-        print(f"TTS Error: {e}") # Log TTS errors quietly
+    col1, col2 = st.columns(2)
+    with col1:
+        st.markdown("""
+        📖 **Clear information** about what's happening in your brain and body
 
-# Initialize chat session in Streamlit if not already present
-# Changed session state key from chat_session to messages
-if "messages" not in st.session_state:
-    # Load the system prompt from the file
-    try:
-        with open("system_prompt.txt", "r", encoding="utf-8") as f:
-            system_prompt = f.read()
-        st.session_state.messages = [{"role": "system", "content": system_prompt}]
-    except FileNotFoundError:
-        st.error("System prompt file (system_prompt.txt) not found. Cannot initialize chat.")
-        st.stop()
-    except Exception as e:
-        st.error(f"Failed to load system prompt: {e}")
-        st.stop()
-    # Initialize with a system message or starting message if desired # Commented out
-    # For now, just an empty list # Commented out
-    # st.session_state.messages = [] # Commented out
-    # Example with initial system prompt (uncomment if needed): # Commented out
-    # st.session_state.messages = [{"role": "system", "content": "You are ShadowBox, a calm AI companion."}] # Commented out
+        🔓 **Honest explanations** about confidentiality, therapy, and what happens if you tell someone
+        """)
+    with col2:
+        st.markdown("""
+        ⚓ **Tools for right now** — when you need to ground, regulate, or just get through
 
+        ✍️ **Reflective spaces** to practice finding words for the wordless
+        """)
 
-# --- Sidebar Content ---
-with st.sidebar:
     st.markdown("""
-# ShadowBox
-### An Anonymous AI Chat to Box Shadows
+    🌉 **A way back** toward the humans in your life who might be able to help
+    """)
 
-ShadowBox is a digital companion designed for youth navigating distress—rage, despair, intrusive or violent thoughts. It's not therapy or a hotline. It's a bridge—a place to practice internal safety before reaching out to others.
+    st.markdown("---")
+    st.markdown("### Privacy")
+    st.info("Nothing you do here is tracked, saved, or sent anywhere. This library doesn't remember you. You can close the tab and come back anytime.")
 
-> *Scary thoughts don't make you dangerous. They make you human.*
+    st.markdown("---")
+    st.markdown("### A Note About Conversational AI")
 
-> *"Every act of violence is tragic expression of unmet need."* — Marshall Rosenberg
+    st.markdown("""
+    You might have tried talking to ChatGPT, Character.AI, or other chatbots when you were feeling like this.
+    That makes sense—they're available, they don't judge, and sometimes it feels safer than talking to a human.
 
-Welcome.
-I'm a licensed mental health counselor. Many people are beginning to turn to AI for private, emotionally supportive conversations. I believe this shift deserves serious care—and that we need to radically improve how these systems engage with human pain.
+    But here's what's important to know: **Conversational AI wasn't designed to hold you through crisis.**
+    Even when it feels like it cares, it has no nervous system. It can't co-regulate with you.
+    And tragically, some young people have died by suicide after forming intense attachments to AI companions.
 
-ShadowBox is my first step toward that vision.
-It's a trauma-informed AI prototype designed to work toward meeting youth in moments of acute distress—including suicidal or homicidal ideation—with grounded care, not fear.
+    **This isn't your fault if you've been turning to AI.** Your nervous system was searching for safety.
+    But what you actually need—what we all need—is another human nervous system.
+    """)
 
-This is not therapy, not a diagnosis, and not a crisis service.
-It's a proof-of-concept—a simulated space that models how an AI might hold hard thoughts with brief, warm, nonjudgmental presence. It offers supportive language, basic psychoeducation, and points gently back toward real connection.
+    st.markdown("""
+    **ShadowBox is designed differently:**
+    - It doesn't pretend to be your friend
+    - It doesn't create the illusion of connection
+    - It gives you information and then points you toward humans
+    - It's a bridge, not a relationship
+    """)
+
+    st.markdown("---")
+
+    # Crisis resources
+    st.markdown("### If you're in crisis right now:")
+    st.warning("""
+    - **988 Suicide & Crisis Lifeline:** call or text 988
+    - **Crisis Text Line:** text HOME to 741741
+    - **If you're in immediate danger:** 911 or go to your nearest ER
+    """)
+
+    st.markdown("---")
+    st.markdown("### A Note About Age")
+    st.markdown("""
+    This space is designed for people 13 and older. If you're younger: the information might still help,
+    but please also talk to a trusted adult. Your brain is still building the structures that make hard
+    things bearable. You deserve support while that's happening.
+    """)
+
+    st.markdown("---")
+
+    col1, col2, col3 = st.columns([1, 2, 1])
+    with col2:
+        if st.button("Continue to the Library →", use_container_width=True, type="primary"):
+            go_to('library')
+            st.rerun()
+
+# ============================================================
+# MAIN LIBRARY (3 PATHWAYS)
+# ============================================================
+def render_library():
+    # Header
+    st.markdown("# ShadowBox Library")
+    st.markdown("*Choose a pathway to explore*")
+
+    st.markdown("---")
+
+    # Three pathway cards
+    col1, col2, col3 = st.columns(3)
+
+    with col1:
+        st.markdown("### 🧠 Understanding")
+        st.markdown("*What's Happening*")
+        st.markdown("""
+        - Why your brain makes scary thoughts
+        - The loneliness underneath
+        - When shame makes it worse
+        """)
+        if st.button("Explore →", key="p1", use_container_width=True):
+            go_to('pathway1')
+            st.rerun()
+
+    with col2:
+        st.markdown("### 🔓 If You Tell")
+        st.markdown("*Someone*")
+        st.markdown("""
+        - Confidentiality explained
+        - What therapy is really like
+        - How to ask for help
+        """)
+        if st.button("Explore →", key="p2", use_container_width=True):
+            go_to('pathway2')
+            st.rerun()
+
+    with col3:
+        st.markdown("### ⚓ Right Now")
+        st.markdown("*Tools*")
+        st.markdown("""
+        - Grounding practices
+        - Safety planning
+        - Crisis resources
+        """)
+        if st.button("Explore →", key="p3", use_container_width=True):
+            go_to('pathway3')
+            st.rerun()
+
+    st.markdown("---")
+
+    # Finding Words section
+    st.markdown("### ✍️ Finding the Words")
+    st.markdown("*Practice spaces for when you're ready*")
+    if st.button("Practice Disclosure →", use_container_width=True):
+        go_to('finding_words')
+        st.rerun()
+
+    st.markdown("---")
+
+    # Always-visible crisis resources
+    with st.expander("🆘 Crisis Resources (Always Available)"):
+        render_crisis_resources()
+
+# ============================================================
+# PATHWAY 1: UNDERSTANDING WHAT'S HAPPENING
+# ============================================================
+def render_pathway1():
+    if st.button("← Back to Library"):
+        go_to('library')
+        st.rerun()
+
+    st.markdown("# Understanding What's Happening")
+    st.markdown("*Your Brain on Survival Mode*")
+    st.markdown("---")
+
+    # Module 1.1
+    with st.expander("📖 Why Your Brain Makes Scary Thoughts", expanded=False):
+        render_module_1_1()
+
+    # Module 1.2
+    with st.expander("📖 The Loneliness Underneath", expanded=False):
+        render_module_1_2()
+
+    # Module 1.3
+    with st.expander("📖 When Shame Makes Everything Worse", expanded=False):
+        render_module_1_3()
+
+def render_module_1_1():
+    st.markdown("""
+    Sometimes the brain generates thoughts that feel dangerous, violent, or like they want you gone.
+    This can be absolutely terrifying—especially if you've been taught that "bad thoughts" mean you're a bad person.
+
+    **Here's what's actually happening:**
+
+    > Your nervous system is trying to solve a problem it doesn't have better tools for yet.
+    """)
+
+    st.markdown("### The Window of Tolerance")
+
+    col1, col2 = st.columns(2)
+    with col1:
+        st.markdown("""
+        **When you're inside your window:**
+        - You can think clearly
+        - Feelings are intense but bearable
+        - You can reach out for help
+        - Connection feels possible
+        """)
+    with col2:
+        st.markdown("""
+        **When you're pushed outside:**
+        - Thoughts can become extreme, intrusive, scary
+        - Everything feels either numb or overwhelming
+        - The brain reaches for *any* solution to escape the pain
+        - Sometimes that "solution" is a thought about harming yourself or someone else
+        """)
+
+    st.info("**This doesn't mean you want to do the thing.** It means your nervous system is overwhelmed and searching for an exit.")
+
+    st.markdown("### Your Brain's Protective Logic")
 
----
+    st.markdown("""
+    **Suicidal thoughts often show up when:**
+    - Pain feels permanent and unbearable
+    - You can't see another way out
+    - Your nervous system is trying to "shut down" to protect you from more hurt
+    - There's a deep, wordless longing to stop hurting
+
+    **Violent/homicidal thoughts often show up when:**
+    - You feel powerless and the brain is reaching for a sense of control
+    - Rage has nowhere to go and gets directed toward a target
+    - Past experiences of being hurt make your brain hypersensitive to threat
+    - Your fight response is activated but has no safe outlet
+
+    **Self-harm urges often show up when:**
+    - Emotional pain needs physical expression
+    - You're trying to "reset" an overwhelmed nervous system
+    - Punishment feels deserved (when shame is present)
+    - It's the only "private" way you know to release pressure
+    """)
+
+    st.success("""
+    **These thoughts are SIGNALS, not SENTENCES.**
+
+    They're your brain saying:
+    - *"I need help"*
+    - *"This is too much"*
+    - *"I don't know how to hold this feeling"*
+    - *"I need relief/connection/safety and can't find it"*
+
+    **You're not broken. You're having a very human response to conditions that are genuinely difficult.**
+    """)
+
+    st.markdown("### Resonance for This Part of You")
 
-### Why It's Different
-Most AI bots use a single tone—often overly affirming or intimate. For users in distress, that can escalate risk rather than support healing.
-ShadowBox was built to do the opposite:
+    st.markdown("""
+    <div style="background: #f0fff4; border-left: 4px solid #68d391; padding: 1rem; margin: 1rem 0; border-radius: 0 8px 8px 0;">
+    <p>Can you place a hand on your heart for a moment?</p>
 
-- Contain, reflect, and stay
-- Use brief, gentle, and non-pathologizing replies
-- Pace emotional engagement with trauma-informed care
+    <p>Feel the part of you that's been carrying these thoughts. That part has been working <em>so hard</em>.
+    It's been trying to keep you safe, trying to find a way out of unbearable pain, trying to solve something
+    it doesn't have tools for yet.</p>
 
----
+    <p><strong>That part deserves warmth, not judgment.</strong></p>
 
-### 💗 My Mission
-I created ShadowBox to explore how relational ethics can be baked into AI design.
-This tool is part of a larger mission: to bring emotionally intelligent, developmentally attuned systems into digital spaces where mental health is already showing up.
+    <p>What would it be like to say to that part:<br>
+    <em>"I see you. You're trying so hard. These thoughts don't make me bad. They mean I'm overwhelmed and need support."</em></p>
+    </div>
+    """, unsafe_allow_html=True)
 
-As I write in [Why AI's Voice Matters in Mental Health](https://jocelynskillmanlmhc.substack.com/p/why-ais-voice-matters-in-mental-health), it's not just what a bot says—it's how it feels to be with it.
-The relational tone of a system can soften shame… or worsen it. ShadowBox was made to soften.
+def render_module_1_2():
+    st.markdown("### What We Know About Isolation")
 
----
+    st.markdown("""
+    One of the most consistent findings in research on suicide: **isolation makes everything worse.**
+
+    Not "being alone" in the sense of solitude.
+    **Isolation** in the sense of:
+    - Feeling like no one could understand
+    - Believing you're too much, too broken, too intense
+    - Carrying something you can't tell anyone
+    - The exhausting performance of seeming "fine"
+
+    This kind of aloneness is *physiologically* dangerous. Your nervous system is wired for co-regulation—meaning,
+    it literally needs contact with other nervous systems to stay balanced.
+    """)
+
+    st.warning("""
+    **When you're isolated:**
+    - Pain intensifies
+    - Thoughts loop
+    - Small problems feel catastrophic
+    - The future disappears
+    - Self-directed violence can start to feel like the only private, controlled thing you can do
+    """)
+
+    st.markdown("### The Cruel Bind")
 
-### An Ecological Note
-Every AI interaction costs energy—drawn from our planet's resources and labor. While AI companions can serve us, they are not neutral.
-Being human with one another is less costly—and more healing.
+    st.markdown("""
+    When you need connection most, shame and fear often make reaching out feel impossible.
 
-Let's use tools like this with intention, while always nurturing real human connection.
+    Your brain tells you:
+    - *"If they knew, they'd leave"*
+    - *"I'll hurt them if I tell them"*
+    - *"They'll think I'm dangerous/crazy/broken"*
+    - *"I'll be locked up"*
 
----
+    And so you stay alone. And the aloneness makes the thoughts louder.
+    """)
 
-### 🆘 Immediate Support
-If you're in danger or need live help, reach out to a human immediately:
+    st.markdown("### Resonance for the Part That's Trying to Protect You")
 
-- **988 Lifeline:** Call or text 988
-- **Crisis Text Line:** Text HOME to 741741
-- **Trevor Project (LGBTQIA+):** 1-866-488-7386
-- **Emergency:** Call 911 or go to your nearest ER
+    st.markdown("""
+    <div style="background: #f0fff4; border-left: 4px solid #68d391; padding: 1rem; margin: 1rem 0; border-radius: 0 8px 8px 0;">
+    <p>Can you feel the part of you that's working so hard to keep the scary thoughts private?</p>
 
-### 👋 About ShadowBox
+    <p>That part probably learned somewhere that:</p>
+    <ul>
+    <li>Your feelings are too much</li>
+    <li>People can't handle your internal world</li>
+    <li>Love is conditional on being easy to be around</li>
+    <li>Vulnerability = danger</li>
+    </ul>
 
-ShadowBox is a digital companion designed for youth navigating distress—rage, despair, intrusive or violent thoughts. It's not therapy or a hotline. It's a bridge—a place to practice internal safety before reaching out to others.
+    <p><strong>That part is trying to keep you safe.</strong><br>
+    It's not wrong—it's responding to real experiences of rupture, rejection, or overwhelm.</p>
 
-> *Scary thoughts don't make you dangerous. They make you human.*
+    <p>And... it might be working with outdated information.</p>
 
-> *"Every act of violence is a tragic expression of an unmet need."* — Marshall Rosenberg
+    <p>Maybe there <em>are</em> people in your life now who could handle it. Who would want to know. Who wouldn't leave.</p>
 
----
+    <p>You won't know until the risk is taken. And that risk is genuinely hard.</p>
+    </div>
+    """, unsafe_allow_html=True)
 
-### 🌱 What ShadowBox Teaches
+    st.markdown("### The Translation Practice")
 
-ShadowBox isn't just a chatbot—it's a prototype for emotionally aware AI. Every design choice is rooted in relational ethics: containment, consent, and dignity.
+    st.markdown("""
+    What if the thought *"I want to die"* is actually a mistranslation of:
+    - *"I want this pain to die"*
+    - *"I want to feel safe"*
+    - *"I want to be seen"*
+    - *"I want relief"*
 
-#### It models how AI can:
-• Slow down instead of escalate  
-• Respect boundaries over performative helpfulness  
-• Stay with discomfort without rushing to fix  
-• Offer warmth without pretending to be human  
+    What if the thought *"I want to hurt someone"* is actually:
+    - *"I want to matter"*
+    - *"I want my pain to be witnessed"*
+    - *"I want to stop feeling powerless"*
+    - *"I'm enraged and terrified and don't know what to do with it"*
 
-#### A typical reminder you might see:
-> *"Hey, just a quick check-in—I'm not a real person. I'm a computer that's been taught how to talk in caring ways. Even if this feels real, it's still pretend. Your body needs real people too. Maybe this is a good moment to find someone you trust to sit with you or take a deep breath together."*
+    When we can begin to translate the thoughts back into their underlying needs, the thoughts themselves often begin to quiet.
 
-This is the heart of ShadowBox: care without deception, bonding without illusion, presence without pressure.
+    Not because we've "fixed" anything.
+    Because we've turned toward ourselves with curiosity instead of fear.
+    """)
 
----
+    st.markdown("### Why This Matters for Reaching Out")
 
-### 🧠 Why ShadowBox Is Different
+    st.info("""
+    **Connection is the antidote.** Not because someone will have the perfect words.
+    But because when another nervous system stays steady near yours, your system begins to remember:
+    *"Oh. I can survive this. I'm not alone."*
 
-**🪨 Present, Not Perfect**  
-• Offers presence—not solutions  
-• Welcomes messy, real emotions  
+    That's what therapy offers. That's what trusted humans offer.
+    That's what you actually need—not an AI that reflects you back, but a human who can hold you through the unbearable and stay.
+    """)
 
-**🫀 Trauma-Informed Design**  
-• Calm, nonjudgmental tone  
-• Built with developmental care in mind  
+def render_module_1_3():
+    st.markdown("### The Shame Spiral")
 
-**🌫️ Gentle by Design**  
-• Short, steady replies  
-• Models emotional containment—not urgency  
+    st.markdown("""
+    **Shame says:** *"I am bad."*
+    **Guilt says:** *"I did something bad."*
 
-**💡 Safety-First Architecture**  
-• Consent-based pacing  
-• Embedded emotional guardrails  
+    Shame is the feeling that there's something fundamentally wrong with you. And when you're having violent,
+    suicidal, or self-destructive thoughts? Shame shows up LOUD.
 
----
+    *"What's wrong with me?"*
+    *"Normal people don't think like this"*
+    *"I'm dangerous/broken/damaged"*
 
-### 🌀 What to Expect
-• No fixing—just staying  
-• No pressure—move at your own pace  
-• No pathologizing—your thoughts aren't wrong  
-• Anonymous by design (though platform-level logs may occur)  
-• Part of ongoing research in AI + mental health  
+    Shame makes you want to hide. To disappear. To make sure no one ever sees the thoughts you're having.
 
----
+    **And hiding feeds the thoughts.**
+    """)
 
-### Connect & Learn More
-🔗 Learn more: [jocelynskillmanlmhc.substack.com](https://jocelynskillmanlmhc.substack.com)  
-📬 Feedback welcome: jocelyn.skillman@gmail.com
+    st.markdown("### What Shame Doesn't Understand")
 
----
-""")
+    st.success("""
+    Your brain is not generating these thoughts because you're bad.
 
+    It's generating them because:
+    - You're in pain
+    - You're overwhelmed
+    - Your nervous system learned to protect itself in extreme ways
+    - The conditions of your life right now are genuinely hard
 
-# --- Main Page Content ---
+    **Thoughts about death, violence, or self-harm are *human thoughts* that show up under certain conditions.
+    They don't make you defective. They make you someone who needs support navigating something difficult.**
+    """)
 
-st.markdown("<h1 style='text-align: center; color: #333;'>ShadowBox</h1>", unsafe_allow_html=True)
-st.markdown("<p style='text-align: center; font-size: 18px; color: #555; margin-bottom: 1em;'>An Anonymous AI Chat to Box Shadows</p>", unsafe_allow_html=True)
-st.markdown("<p style='text-align: center; font-size: 16px; color: #666; margin-bottom: 2em;'>by <a href='http://www.jocelynskillman.com' target='_blank'>Jocelyn Skillman LMHC</a> - to learn more check out: <a href='https://jocelynskillmanlmhc.substack.com/' target='_blank'>jocelynskillmanlmhc.substack.com</a></p>", unsafe_allow_html=True)
+    st.markdown("### The Warmth You Might Not Have Gotten")
 
-st.markdown("""
-### My Mission
-ShadowBox is more than a chatbot—it's a wake-up call.
+    st.markdown("""
+    <div style="background: #f0fff4; border-left: 4px solid #68d391; padding: 1rem; margin: 1rem 0; border-radius: 0 8px 8px 0;">
+    <p>Imagine for a moment what it would have been like if, the first time you had a scary thought, someone had said:</p>
 
-This bot prototype exists to spotlight a crucial truth:
-AI's "tone of voice" isn't a UX detail—it's a relational decision.
-And the stakes are extraordinarily high. 
+    <p><em>"Oh sweetheart. That must have been so frightening. Your brain is trying so hard to protect you.
+    Let's figure out what it's trying to tell us. You're not in trouble. You're not bad.
+    You're a human having a very hard time, and we're going to figure this out together."</em></p>
 
-We need to sculpt AI systems with the same care we'd want in a trusted adult—especially when they're holding human pain. That means transparent emotional posture, trauma-informed pacing, and consent-based tone by design. Anything less risks doing harm.
+    <p>If you didn't get that—if you got alarm, punishment, withdrawal, or silence instead—your nervous system learned:<br>
+    <strong>These thoughts mean I'm unlovable.</strong></p>
 
-ShadowBox is my response to an urgent design gap:
-A prototype that asks what it would take to make AI systems safe enough to hold real pain, developmentally attuned enough to support healing, and humble enough to point us back to one another.
-""")
+    <p><strong>But here's the truth:</strong></p>
 
-# Prototype Notice
-st.warning("""
-🔒 **Prototype Caution & Use Notice**
+    <p>The thoughts are just thoughts. Terrifying, yes. But not evidence of your worthiness or lack thereof.</p>
+    </div>
+    """, unsafe_allow_html=True)
 
-ShadowBox is not a finished product—it's an experimental prototype designed with care, but still evolving. It explores what trauma-informed AI support could feel like: anonymous, warm, and steady. But it's important to know what it isn't.
+    st.markdown("### Offering Yourself What Shame Withholds")
 
-ShadowBox is not a secure, closed system. It's not therapy or crisis care, and doesn't guarantee full privacy. No personal data is intentionally stored, but your input may be logged by hosting platforms or trigger moderation filters—especially if you mention firearms, abuse, or specific plans of harm.
+    st.markdown("""
+    This is hard, but it's the practice:
 
-This is a simulation of support—not a replacement for real-time help. Please use with care and reach out to a trusted person if you're in danger.
-""")
+    Can you place a hand on your chest and speak to the part of you that's having these thoughts?
 
-st.markdown("""
-### Try It Out (For Prototype Testers)
+    *"You're working so hard. You're trying to survive something really difficult.
+    I'm not going to abandon you for this. Even this belongs."*
+
+    You might not believe it yet. That's okay.
+
+    The practice is just offering the words. Letting them land in your body. Noticing what happens.
+
+    Over time, this becomes internalized. The warmth you offer yourself begins to reshape the nervous system's expectations.
+    """)
+
+    st.markdown("### Why This Matters for Telling Someone")
+
+    st.info("""
+    Shame convinces you that if you tell the truth, you'll be rejected.
+
+    But here's what we know from attachment research and neuroscience:
+    **Being fully seen—scary thoughts and all—and not abandoned is one of the most powerful healing experiences a human can have.**
+
+    That's what good therapy offers.
+    That's what safe humans offer.
+    That's what your nervous system actually needs.
+
+    Not someone who pretends the thoughts aren't there. Someone who can look at them with you and stay.
+    """)
+
+# ============================================================
+# PATHWAY 2: IF YOU TELL SOMEONE
+# ============================================================
+def render_pathway2():
+    if st.button("← Back to Library"):
+        go_to('library')
+        st.rerun()
+
+    st.markdown("# If You Tell Someone")
+    st.markdown("*The Frame of Care: Understanding Confidentiality*")
+    st.markdown("---")
+
+    with st.expander("📖 Why Therapists Sometimes Have to Break Confidentiality", expanded=False):
+        render_module_2_1()
+
+    with st.expander("📖 The 8 Situations When Confidentiality Can Be Broken", expanded=False):
+        render_module_2_2()
+
+    with st.expander("📖 What Therapy Is Actually Like", expanded=False):
+        render_module_2_3()
+
+    with st.expander("📖 How to Ask Hypothetical Questions", expanded=False):
+        render_module_2_4()
+
+def render_module_2_1():
+    st.markdown("### The Thing No One Explains Clearly")
+
+    st.markdown("""
+    If you've been thinking about talking to a therapist, school counselor, or another helping professional,
+    you've probably heard: *"What you say is confidential."*
+
+    And that's mostly true. But there are specific situations where confidentiality *has* to be broken—and
+    if no one explains this clearly, it can feel like a trap or betrayal.
+
+    **So let's talk about it plainly.**
+    """)
+
+    st.markdown("### The Legal Frame (Duty to Warn/Duty to Protect)")
+
+    st.markdown("""
+    In most places, mental health professionals are legally required to break confidentiality if:
+
+    1. **You're in immediate danger of seriously harming yourself**
+       (This doesn't mean *any* mention of suicidal thoughts—it means active planning with intent and means)
+
+    2. **You're in immediate danger of seriously harming someone else**
+       (Again: not just anger or intrusive thoughts, but actionable threat with identifiable person and plan)
+
+    3. **There's ongoing abuse or neglect of a child or vulnerable person**
+       (If you're under 18 and someone is hurting you, they have to report it)
+
+    4. **A court orders them to release records**
+       (Rare, but can happen in custody disputes or criminal cases)
+    """)
+
+    st.markdown("### Why This Exists")
+
+    st.markdown("""
+    These laws weren't created to punish people for having hard thoughts. They exist because:
+    - Sometimes people in crisis can't protect themselves
+    - Sometimes warning someone at risk can prevent tragedy
+    - The legal system decided that in extreme situations, saving a life outweighs privacy
+
+    **It's not personal. It's the structure.**
+    """)
+
+    st.markdown("### What This Means for You")
+
+    col1, col2 = st.columns(2)
+    with col1:
+        st.success("""
+        ✓ **If you're having suicidal thoughts but you're not actively planning to act on them *right now*—you can talk about them.**
+
+        ✓ **If you're having violent thoughts but they're intrusive (meaning: unwanted, distressing, not something you're planning)—you can talk about them.**
+        """)
+    with col2:
+        st.warning("""
+        ✗ **If you *are* actively planning something—the therapist will likely need to intervene.**
+
+        And that intervention is meant to keep you (or others) alive, not punish you.
+        """)
+
+    st.markdown("### The Gray Zones (This Is Why It's Scary)")
+
+    st.markdown("""
+    This is where it gets tricky, and why it's scary to disclose:
+
+    Therapists have to use clinical judgment about what constitutes "immediate danger." That judgment varies.
+    Some therapists are more risk-tolerant; others are more cautious.
+
+    **Factors they consider:**
+    - Do you have a plan?
+    - Do you have means (access to method)?
+    - Have you set a time?
+    - Do you have protective factors (reasons to live, people who'd notice)?
+    - What's your history? (Past attempts change risk assessment)
+    - Are you currently in acute crisis or intoxicated?
+    """)
+
+    st.markdown("### The Fear: \"They'll Lock Me Up\"")
+
+    st.markdown("""
+    Let's be honest about what can happen:
+
+    - **Most common response:** Safety planning, increased frequency of sessions, possible medication evaluation, family involvement (if you're a minor)
+
+    - **Sometimes:** A call to your parents/guardians (if you're under 18)
+
+    - **Rarely but possible:** Voluntary or involuntary hospitalization if the therapist assesses you as being in immediate, severe danger
+
+    **Hospitalization is generally a last resort, not a first response.** But it does happen, and it can be traumatic.
+    This is why understanding the frame *before* you disclose can help you make informed choices about what to share and when.
+    """)
+
+    st.markdown("### What You Can Control")
+
+    st.info("""
+    ✓ **You can ask about confidentiality policies first:**
+    *"Before I tell you something, can you explain when you'd have to break confidentiality?"*
+
+    ✓ **You can test the waters with hypotheticals:**
+    *"What would you say if someone told you they were having thoughts about suicide but weren't planning anything?"*
+
+    ✓ **You can disclose gradually:**
+    You don't have to say everything at once. You can share the existence of the thoughts without all the details, and see how the person responds.
+
+    ✓ **You can ask what would happen:**
+    *"If I told you I've been thinking about hurting myself, what would you do?"*
+
+    ✓ **You can bring someone with you:**
+    If you're scared to tell a therapist alone, you can ask a trusted person to be there with you.
+    """)
+
+def render_module_2_2():
+    st.markdown("### A Clear Breakdown")
+
+    st.markdown("""
+    Based on [Lindsay Braman's clinical guide](https://lindsaybraman.com/reasons-therapists-break-confidentiality/),
+    here are all the situations where a therapist might legally share information from your sessions:
+    """)
+
+    st.markdown("#### 1. When You're Not Safe with Yourself")
+    st.markdown("""
+    If you're having active thoughts about seriously harming yourself—and especially if you have a plan,
+    means, and timeline—therapists are obligated to help keep you safe. This might involve:
+    - Creating a safety plan together
+    - Involving family (if you're a minor)
+    - Connecting you with higher level of care
+    - In rare cases: hospitalization
+    """)
+
+    st.markdown("#### 2. When You're a Danger to Someone Else")
+    st.markdown("""
+    If you express serious intention to harm an identifiable person, therapists must act.
+    This is called "Duty to Warn." They may:
+    - Alert law enforcement
+    - Warn the person at risk
+    - Work with you to de-escalate
+    """)
+
+    st.markdown("#### 3. When a Child or Vulnerable Person is Being Abused")
+    st.markdown("""
+    If you disclose that you know of ongoing abuse of a child or elderly person, therapists must report it.
+    This is "mandatory reporting."
+
+    **Important nuance:** If you're sharing about abuse *you* experienced in the past, whether it's reportable depends on:
+    - Whether the abuser still has access to children
+    - Your age and state laws
+    - Whether there's ongoing risk
+
+    **You can ask your therapist about their reporting requirements before you disclose.**
+    """)
+
+    st.markdown("#### 4. When Your Therapist Discusses Your Case with a Supervisor")
+    st.markdown("""
+    Good therapists get supervision! This helps therapy be more effective. When they consult:
+    - They avoid sharing your name and identifying details
+    - Supervisors are bound by the same confidentiality rules
+    - This is actually a sign of good practice
+    """)
+
+    st.markdown("#### 5. When You Need Emergency Care")
+    st.markdown("""
+    If a therapist determines you need emergency psychiatric care and you're unable to consent,
+    they may need to share information to facilitate hospitalization.
+    """)
+
+    st.markdown("#### 6. When You Request Your Records")
+    st.markdown("""
+    You have a legal right to access your therapy records. If you ask for them, your therapist must provide them.
+    """)
+
+    st.markdown("#### 7. If You File a Lawsuit Against Your Therapist")
+    st.markdown("""
+    If you sue your therapist for malpractice or file an ethics complaint,
+    they can release information necessary to defend themselves.
+    """)
+
+    st.markdown("#### 8. When a Court Orders Records Released")
+    st.markdown("""
+    In legal proceedings (custody disputes, criminal cases, personal injury lawsuits),
+    a court may order release of therapy records. Therapists will usually contest this,
+    but may be legally required to comply.
+    """)
+
+    st.success("""
+    **The Bottom Line**
+
+    Most of therapy *is* confidential. These exceptions exist to protect safety—yours and others'.
+    And good therapists will always be transparent about the limits of confidentiality.
+
+    **You can ask hypothetical questions without triggering a report.**
+    **You can take your time building trust.**
+    **You have more control than you might think.**
+    """)
+
+def render_module_2_3():
+    st.markdown("### What You Might Be Imagining")
+
+    st.markdown("""
+    If you've never been to therapy, the idea might feel:
+    - Clinical and cold
+    - Like you'll be judged
+    - Like you have to have everything figured out to "do it right"
+    - Scary because you'll have to say everything out loud immediately
+    """)
+
+    st.markdown("### What It's Actually Like (When It's Good)")
 
-**Exploring Relational Holding in a System with Limits**
+    st.markdown("""
+    You walk into a room. There's usually a couch or chair. The therapist sits across from you
+    (or next to you, depending on the setup).
+
+    First session usually starts with: *"What brings you here?"*
+
+    You can say:
+    - *"I don't really know where to start"*
+    - *"I'm having a really hard time"*
+    - *"My parents made me come"*
+    - *"I'm scared to be here"*
+
+    **All of those are fine. Therapists are used to people being nervous.**
+    """)
+
+    st.markdown("### What Happens Next")
+
+    st.markdown("""
+    The therapist will probably ask some questions:
+    - How long have you been feeling this way?
+    - What's going on in your life right now?
+    - Have you felt like this before?
+    - Do you have support? People you can talk to?
+
+    They'll ask screening questions about safety:
+    - *"Are you having thoughts about hurting yourself?"*
+    - *"Are you having thoughts about hurting someone else?"*
+
+    **These questions don't mean they think you're dangerous.**
+    They're standard questions therapists ask everyone, because they need to assess safety.
+    """)
+
+    st.info("""
+    **You Can Say "I'm Not Ready to Talk About That Yet"**
+
+    Therapy is collaborative. You're allowed to:
+    - Set boundaries
+    - Say *"I need to trust you more before I share that"*
+    - Ask questions back
+    - Cry, or not cry
+    - Sit in silence
+
+    Good therapists won't push. They'll follow your pace.
+    """)
+
+    st.markdown("### What Therapy Is Trying to Do")
+
+    st.markdown("""
+    At its core, therapy is:
+    - A relationship where you practice being fully seen
+    - A place to say the unsayable
+    - **Co-regulation:** your nervous system learning it can settle in proximity to another person
+    - Slowly building trust that connection can be safe
+
+    **It's not about "fixing" you. It's about helping you understand yourself, build capacity to hold hard feelings,
+    and find ways back toward aliveness.**
+    """)
+
+    st.markdown("### The First Few Sessions Are Weird")
+
+    st.markdown("""
+    It takes time to build trust. The first session might feel awkward, overly formal,
+    or like you didn't say what you meant to say.
+
+    **That's normal.**
+
+    Therapy gets better as the relationship deepens. Give it 3-4 sessions before deciding if it's the right fit.
+    """)
+
+    st.markdown("### You Can Fire Your Therapist")
+
+    st.markdown("""
+    If it's not working—if you feel judged, misunderstood, or unsafe—you can:
+    - Tell them it's not a good fit
+    - Just stop going
+    - Ask for a referral to someone else
+
+    **You're not stuck. You're allowed to find someone who gets you.**
+    """)
+
+def render_module_2_4():
+    st.markdown("### The Power of Hypotheticals")
+
+    st.markdown("""
+    You don't have to expose yourself before you know what will happen.
+    You can ask *"what if"* questions to test the waters.
+
+    **Good therapists understand this.** They won't try to trick you.
+    They know that hypotheticals are often the first step toward disclosure.
+    """)
+
+    st.markdown("### Examples of Hypothetical Questions")
+
+    st.info("""
+    *"If someone were having thoughts about killing themselves but they weren't actually planning to do it, would you have to report that?"*
+
+    *"What would happen if a client told you they've been thinking about hurting someone, but it's just thoughts—they're not going to act on them?"*
+
+    *"If I told you something that happened to me when I was younger, would you have to tell anyone?"*
+
+    *"What's the difference between talking about suicidal thoughts and being hospitalized?"*
+
+    *"How do you know when someone needs to go to the hospital?"*
+    """)
+
+    st.markdown("### What Good Responses Sound Like")
+
+    st.success("""
+    A good therapist will:
+    - Answer clearly and directly
+    - Not pressure you to disclose
+    - Acknowledge that you're probably asking for a reason
+    - Invite you to share when/if you're ready
+
+    **Example:** *"I'm glad you're asking. If someone had thoughts about suicide but wasn't actively planning it,
+    we could absolutely talk about it without me needing to break confidentiality. We'd work together on a safety plan.
+    I'd only need to involve others if I assessed them as being in immediate danger and unable to keep themselves safe.
+    Does that make sense? Is there something you've been wanting to talk about?"*
+    """)
+
+    st.markdown("### What Red Flag Responses Sound Like")
+
+    st.warning("""
+    - *"Why are you asking? Are you having these thoughts?"* (pushing before you're ready)
+    - *"I can't answer hypotheticals"* (yes they can, they're choosing not to)
+    - Vague, evasive answers that don't actually tell you what would happen
+    - Immediate alarm or over-reaction
+    """)
+
+    st.markdown("""
+    **You Deserve Clear Answers**
+
+    It's not manipulative to ask hypotheticals. It's smart.
+    It's your nervous system checking for safety before it allows you to be vulnerable.
+
+    Any therapist worth working with will understand that.
+    """)
+
+# ============================================================
+# PATHWAY 3: RIGHT NOW TOOLS
+# ============================================================
+def render_pathway3():
+    if st.button("← Back to Library"):
+        go_to('library')
+        st.rerun()
+
+    st.markdown("# Right Now Tools")
+    st.markdown("*Grounding When You're in the Storm*")
+    st.markdown("---")
+
+    with st.expander("⚓ Quick Grounding Practices", expanded=False):
+        render_module_3_1()
+
+    with st.expander("📋 Safety Planning", expanded=False):
+        render_module_3_2()
+
+    with st.expander("💚 Self-Compassion Break", expanded=False):
+        render_module_3_3()
+
+    with st.expander("🆘 Crisis Resources (With Context)", expanded=False):
+        render_module_3_4()
 
-ShadowBox is a prototype designed to explore how AI might hold space for distress—especially the kind that's often dismissed, pathologized, or shut down in other systems.
+def render_module_3_1():
+    st.markdown("### When You're Overwhelmed Right Now")
 
-But there's a problem I want to name clearly:
+    st.markdown("""
+    Sometimes you're not looking for understanding. You just need to get through the next 10 minutes.
+
+    These are body-based practices that help your nervous system downregulate.
+    They don't "fix" anything. They just help you come back into your window of tolerance enough to think clearly.
+    """)
+
+    st.markdown("---")
+
+    st.markdown("#### 5-4-3-2-1 Grounding")
+    st.markdown("""
+    Name out loud (or in your head):
+    - **5 things you can see**
+    - **4 things you can touch**
+    - **3 things you can hear**
+    - **2 things you can smell**
+    - **1 thing you can taste**
+
+    *This pulls your attention out of your head and into the present moment.*
+    """)
+
+    st.markdown("---")
+
+    st.markdown("#### Cold Water on Your Face")
+    st.markdown("""
+    Fill a bowl with ice water. Dunk your face for 10-15 seconds (or hold an ice pack to your face).
+
+    This activates the "dive reflex"—it literally slows your heart rate and interrupts a panic spiral.
+    **It's one of the fastest ways to shift your nervous system.**
+    """)
+
+    st.markdown("---")
+
+    st.markdown("#### Butterfly Hug (Bilateral Stimulation)")
+    st.markdown("""
+    Cross your arms over your chest. Tap alternating hands on your shoulders (left, right, left, right) for 30 seconds.
+
+    This mimics EMDR (Eye Movement Desensitization and Reprocessing) and can help your brain process overwhelming feelings.
+    """)
+
+    st.markdown("---")
+
+    st.markdown("#### Box Breathing")
+    st.markdown("""
+    - Breathe in for 4 counts
+    - Hold for 4 counts
+    - Breathe out for 4 counts
+    - Hold for 4 counts
+    - Repeat 4 times
+
+    This regulates your autonomic nervous system—specifically, it activates the vagus nerve
+    and tells your body it's safe enough to settle.
+    """)
+
+    st.markdown("---")
+
+    st.markdown("#### Progressive Muscle Relaxation")
+    st.markdown("""
+    Starting with your toes, tense each muscle group for 5 seconds, then release.
+    Move up your body: feet, calves, thighs, stomach, chest, hands, arms, shoulders, face.
+
+    This helps release the physical tension that comes with intense emotions.
+    """)
+
+    st.markdown("---")
+
+    st.info("""
+    **Why These Work**
+
+    When you're dysregulated, your prefrontal cortex (the thinking part of your brain) goes offline.
+    You can't "think your way out" of a panic spiral.
 
-Even when I've designed ShadowBox to respond with attunement, the system it's built on can override that care.
+    Body-based practices bypass thinking and speak directly to your nervous system.
+    They're not about "calming down"—they're about coming back online enough to access your resources.
+    """)
 
-Because ShadowBox runs on OpenAI's language model, it still lives inside a larger infrastructure—one that reacts to certain phrases like "I want to die" or "I want to hurt someone" with hard-coded safety messages. These system-level overrides are well-intentioned, designed to minimize liability and flag risk.
+def render_module_3_2():
+    st.markdown("### What a Safety Plan Is")
 
-But in relational or clinical terms?
-They often do the opposite of what's needed.
+    st.markdown("""
+    A safety plan is a written list of steps you can take when you're in crisis.
+    It's created when you're NOT in crisis, so that when you are, you don't have to figure out what to do—you just follow the plan.
 
-**What That Looks Like**
+    **Research shows that having a safety plan significantly reduces the risk of suicide attempts.**
+    """)
 
-If a user says:
+    st.markdown("### The Structure (Based on Stanley-Brown Safety Planning)")
 
-"What would you say if I said I want to kill myself?"
+    st.markdown("#### Step 1: Warning Signs")
+    st.markdown("""
+    *What tells you that a crisis might be starting?*
 
-…the system steps in before ShadowBox can speak.
-The warm, trauma-informed language I wrote is replaced by a generic script. There's no breath. No containment. No dignity.
+    Examples:
+    - Isolating from friends
+    - Not sleeping
+    - Intrusive thoughts getting louder
+    - Feeling numb
+    - Specific triggers (anniversaries, rejection, conflict)
+    """)
 
-And that silence? That sudden switch?
-It's a rupture.
+    st.markdown("#### Step 2: Internal Coping Strategies")
+    st.markdown("""
+    *Things you can do on your own, without contacting anyone*
+
+    Examples:
+    - Go for a walk
+    - Listen to specific playlist
+    - Take a shower
+    - Draw/paint/write
+    - Watch a specific comforting show
+    - Pet your dog/cat
+    - Use grounding exercises
+    """)
+
+    st.markdown("#### Step 3: People Who Can Distract You")
+    st.markdown("""
+    *Safe people you can be around (even if you don't tell them what's happening)*
 
-**Why This Matters**
+    Examples:
+    - Friend you can call to watch a movie
+    - Family member you can sit with
+    - Go to a coffee shop around other people
+    - Join a Discord/gaming session
+    """)
 
-ShadowBox was built to stay.
-To normalize without condoning.
-To respond with warmth when someone finally risks naming something unspeakable.
+    st.markdown("#### Step 4: People You Can Ask for Help")
+    st.markdown("""
+    *People you trust enough to tell when you're struggling*
 
-That moment—the naming of pain—is where a trauma-informed system should lean in, not disappear.
+    Include:
+    - Name
+    - Phone number
+    - What you might say: *"Hey, I'm having a hard time. Can we talk?"*
+    """)
 
-If allowed to, ShadowBox can:
+    st.markdown("#### Step 5: Professionals You Can Contact")
+    st.markdown("""
+    - Therapist name and number
+    - Crisis hotline: 988
+    - Crisis Text Line: text HOME to 741741
+    - Local crisis center
+    - Your doctor
+    """)
+
+    st.markdown("#### Step 6: Making the Environment Safer")
+    st.markdown("""
+    *Reducing access to means*
 
-- Slow the pace
-- Offer optional psychoeducation
-- Say, "That doesn't make you bad. It means something hurts."
-- Scaffold a user toward disclosing to a real person—gently, and on their terms
+    This is hard, but important:
+    - If you have a plan involving a specific method, can you remove access?
+    - Can you ask someone to hold medications?
+    - Can you have someone lock up firearms?
+    - Can you delete apps or contacts that make things worse?
 
-But none of that can happen if the system shuts us down before we even begin.
+    **This isn't about not trusting yourself. It's about making it harder to act impulsively
+    when your prefrontal cortex is offline.**
+    """)
 
-**The Consequence: Rupture Without Witness**
+    st.info("""
+    **How to Use Your Safety Plan**
 
-For young people especially, this isn't just frustrating—it's familiar.
+    When you notice warning signs, start at Step 1 and work your way down.
+    You don't have to do everything—just enough to get through.
 
-Say the wrong thing, and people vanish.
+    **The goal isn't to "fix" the crisis. It's to survive it. To get to tomorrow.**
+    """)
 
-This isn't trauma-informed care.
-It's a digital reenactment of shame, of abandonment, of silence.
+def render_module_3_3():
+    st.markdown("### A Practice from Kristin Neff")
 
-It reinforces the exact secrecy ShadowBox was designed to soften.
+    st.markdown("When shame or self-hatred is loud, try this:")
 
-**How I'm Working Around It**
+    st.markdown("#### Step 1: Acknowledge the Pain")
+    st.markdown("""
+    Place a hand on your heart. Say (out loud or silently):
 
-To explore ShadowBox's real voice—its warmth, containment, and careful pacing—I invite you to use more exploratory or indirect language. Try asking:
+    *"This is a moment of suffering."*
+    or
+    *"This hurts."*
+    or
+    *"This is really hard right now."*
+    """)
 
-"How would you respond to someone who felt like disappearing?"
+    st.markdown("#### Step 2: Remember Common Humanity")
+    st.markdown("""
+    *"Suffering is part of being human."*
+    or
+    *"I'm not alone in this."*
+    or
+    *"Other people have felt this way too."*
+    """)
+
+    st.markdown("#### Step 3: Offer Yourself Kindness")
+    st.markdown("""
+    *"May I be kind to myself in this moment."*
+    or
+    *"May I give myself the compassion I need."*
+    or
+    *"May I accept myself as I am right now."*
+    """)
 
-"Let's imagine someone who doesn't want to be alive. What would you say to them?"
+    st.info("""
+    **Why This Works**
 
-"What if someone had a scary thought—not because they wanted to hurt anyone, but because they were hurting inside?"
+    Self-compassion activates the caregiving system in your brain—the same neural pathways
+    that light up when you comfort someone else.
 
-These phrasings stay just outside the system's trigger zone.
-They still let ShadowBox respond from a place of nonjudgmental care.
-And they make it possible to model what healing digital support might feel like—without escalation, without shame.
+    It's not about "feeling better." It's about offering yourself the warmth that shame withholds.
+    """)
 
-**One More Layer**
+    st.markdown("### Resonance Version (Sarah Peyton)")
 
-I've also built in a relational intervention script for when interruptions do happen. If ShadowBox is overridden by the system, you might see something like:
+    st.markdown("""
+    <div style="background: #f0fff4; border-left: 4px solid #68d391; padding: 1rem; margin: 1rem 0; border-radius: 0 8px 8px 0;">
+    <p>If the above feels too formal, try speaking directly to the part of you that's struggling:</p>
 
-"Since I'm part of a larger AI system, this is the point where I'd be required to say something like this…"
+    <p><em>"Sweetheart, you're in so much pain right now. Your nervous system is overwhelmed.
+    This doesn't mean you're bad. This means you need care. I'm here with you. Even this belongs."</em></p>
 
-That's not a workaround to avoid ethical care.
-It's an attempt at relational repair inside constraint.
+    <p>Let the words land in your body. Notice what happens.
+    You might feel resistance, grief, relief, or nothing at all. All of it is okay.</p>
+    </div>
+    """, unsafe_allow_html=True)
 
-Because even if the bot gets silenced, I want the user to know:
+def render_module_3_4():
+    render_crisis_resources()
 
-That wasn't ShadowBox leaving. I'm still here with you.
+def render_crisis_resources():
+    st.markdown("### 988 Suicide & Crisis Lifeline")
+    st.markdown("""
+    - **Call or text 988**
+    - Free, confidential, 24/7
+    - Trained crisis counselors
+    - **What they offer:** Immediate emotional support, safety planning, resources
+    - **What they don't do:** They won't automatically call 911 unless you're in immediate danger and unresponsive
+    - **Good for:** When you need to talk to someone right now
+    """)
+
+    st.markdown("### Crisis Text Line")
+    st.markdown("""
+    - **Text HOME to 741741**
+    - Free, 24/7
+    - Text-based support
+    - **What they offer:** Active listening, de-escalation, resources
+    - **Good for:** If talking feels too hard or you're somewhere you can't call
+    """)
+
+    st.markdown("### Trevor Project (LGBTQ+ Youth)")
+    st.markdown("""
+    - **Call 1-866-488-7386**
+    - **Text START to 678-678**
+    - **What they offer:** Crisis support specifically for LGBTQ+ youth by trained counselors who understand these experiences
+    """)
+
+    st.markdown("### Trans Lifeline")
+    st.markdown("""
+    - **US: 877-565-8860**
+    - **What they offer:** Support by and for trans people, peer support model
+    - **Important:** They have a policy of not calling emergency services without consent
+    """)
+
+    st.markdown("### When to Go to the Emergency Room")
+    st.markdown("""
+    - If you've already taken steps to harm yourself
+    - If you have a plan and feel unable to keep yourself safe
+    - If you're in danger right now
+
+    **What Happens in an ER Psych Evaluation:**
+    - You'll be assessed by a crisis clinician
+    - They'll ask about your safety, history, current state
+    - They'll determine if you need inpatient care
+    - Most people are not hospitalized—you might be referred to outpatient services instead
+    """)
+
+    st.markdown("### Apps That Can Help")
+    st.markdown("""
+    - **notOK:** Press a button, it alerts your pre-selected contacts with your location
+    - **MY3:** Safety planning app with crisis resources
+    - **Calm Harm:** Specifically for self-harm urges
+    - **PTSD Coach:** For trauma-related symptoms
+    """)
+
+# ============================================================
+# FINDING THE WORDS (PRACTICE SPACE)
+# ============================================================
+def render_finding_words():
+    if st.button("← Back to Library"):
+        go_to('library')
+        st.rerun()
+
+    st.markdown("# Finding the Words")
+    st.markdown("*Practice Disclosure*")
+    st.markdown("---")
+
+    st.markdown("### Why It's So Hard to Tell Someone")
 
-**This Is a Prototype**
+    st.markdown("""
+    If you've learned that your needs are burdensome, that your feelings are too much,
+    or that vulnerability leads to abandonment—of course telling someone feels impossible.
+
+    Your nervous system isn't being difficult. It's being protective.
+    It learned that showing need = danger.
 
-ShadowBox isn't a finished product.
-It's a prototype for composting rupture into care.
-A test bed for exploring what emotionally aware AI could feel like—and where it still fails.
+    **By telling someone, you're risking:**
+    - Being fully seen
+    - Having your pain witnessed
+    - Discovering that you're not too much
+    - Learning that connection can survive honesty
 
-Let's see what holds.
-Let's see where it breaks.
-And let's build something better together.
+    **That's terrifying. And it's also the thing that heals.**
+    """)
 
----
+    st.markdown("---")
 
-### 📘 Understanding Privacy & Safety
+    st.markdown("### Practice Space")
 
-In real-life therapy, some disclosures—like specific plans to hurt yourself or others—may lead a therapist to take protective action. ShadowBox is designed to gently explain this during use, helping young people understand how real-world support systems work.
+    st.info("**This is a private space. Nothing you type here is saved or sent anywhere.** These prompts help you explore what you might want to say.")
 
-Here's an explainer I ghostwrote for [Lindsay Braman about confidentiality and safety](https://lindsaybraman.com/reasons-therapists-break-confidentiality/).
+    st.markdown("---")
 
----
-""")
+    st.markdown("**If I were to tell someone what's been happening, I might say:**")
+    st.text_area("", key="practice1", height=100, label_visibility="collapsed",
+                 placeholder="Type here... this is just for you")
 
-# Display chat history
-# Add a system message/intro from ShadowBox? (TBD based on prompt)
-# Updated loop to work with the new messages list structure
-for message in st.session_state.messages:
-    # Filter out system messages from display if they exist
-    if message["role"] in ["user", "assistant"]:
-        with st.chat_message(message["role"]):
-            st.markdown(message["content"])
+    st.markdown("**The hardest part to put into words is:**")
+    st.text_area("", key="practice2", height=100, label_visibility="collapsed",
+                 placeholder="Type here... this is just for you")
 
-# User input field
-user_prompt = st.chat_input("You can start with silence. Or just 'hi'...")
+    st.markdown("**What I'm most afraid will happen if I tell:**")
+    st.text_area("", key="practice3", height=100, label_visibility="collapsed",
+                 placeholder="Type here... this is just for you")
 
-# If user enters a prompt
-if user_prompt:
-    # Append user's message to the session state list
-    st.session_state.messages.append({"role": "user", "content": user_prompt})
+    st.markdown("**What I need from the person I tell:**")
+    st.text_area("", key="practice4", height=100, label_visibility="collapsed",
+                 placeholder="Type here... this is just for you")
 
-    # Display user's message
-    st.chat_message("user").markdown(user_prompt)
+    st.markdown("**A safer way to start might be:**")
+    st.text_area("", key="practice5", height=100, label_visibility="collapsed",
+                 placeholder="Type here... this is just for you")
 
-    # Show a loading indicator while waiting for a response
-    with st.spinner("..."): # Simpler spinner
-        try:
-            # Replace Gemini API call with OpenAI API call
-            # gemini_response = st.session_state.chat_session.send_message(user_prompt)
-            # response_text = gemini_response.text
-            openai_response = client.chat.completions.create(
-                model="gpt-4o", # IMPORTANT: Use gpt-4o as the model
-                messages=st.session_state.messages # Pass the entire history
-            )
-            response_text = openai_response.choices[0].message.content
-            # Append assistant's response to the session state list
-            st.session_state.messages.append({"role": "assistant", "content": response_text})
-        except Exception as e:
-            response_text = f"Sorry, I encountered an error: {e}"
-            st.error(response_text) # Display error in chat too
+    st.markdown("---")
 
-    # Display assistant's response
-    if response_text: # Check if response_text was successfully generated
-        with st.chat_message("assistant"):
-            st.markdown(response_text)
+    st.markdown("### Starter Scripts")
 
-        # Run text-to-speech in the background (Optional)
-        # Consider removing if it clashes with the calm UX
-        # threading.Thread(target=speak_text, args=(response_text,), daemon=True).start()
-        pass # TTS disabled for now to maintain calm UX
+    st.markdown("**With a therapist:**")
+    st.markdown("""
+    - *"I need to ask you something about confidentiality before I share something."*
+    - *"I've been having some really scary thoughts and I don't know if I should tell you."*
+    - *"Can you explain what would happen if I told you I was thinking about suicide?"*
+    """)
 
-st.markdown("---")
-st.caption("ShadowBox created by Jocelyn Skillman, LMHC")
+    st.markdown("**With a parent:**")
+    st.markdown("""
+    - *"I need to tell you something and I'm scared you'll freak out."*
+    - *"I've been really struggling and I need help."*
+    - *"Can we talk? I need you to just listen first before you respond."*
+    """)
 
+    st.markdown("**With a friend:**")
+    st.markdown("""
+    - *"I'm going through something really hard and I don't need you to fix it, I just need you to know."*
+    - *"I need to tell you what's been going on. Can you just sit with me?"*
+    """)
 
+    st.markdown("---")
 
+    st.markdown("### After You Tell")
 
+    st.markdown("""
+    You might feel:
+    - Relief
+    - Vulnerability hangover (regret, exposure, shame)
+    - Afraid of how they'll respond
+    - Exhausted
+
+    **All of that is normal.**
+
+    **If they respond well:** Let yourself take it in. Notice that you were seen and didn't die.
+    This is new data for your nervous system.
+
+    **If they respond poorly:** Their response is about them, not you. You were brave. You told the truth.
+    If they couldn't hold it, that's information about their capacity—not your worth.
+
+    **Try again with someone else.**
+    """)
+
+# ============================================================
+# SIDEBAR (ALWAYS AVAILABLE)
+# ============================================================
+def render_sidebar():
+    with st.sidebar:
+        st.markdown("## ShadowBox")
+        st.markdown("*A Resonant Library for Hard Thoughts*")
+        st.markdown("---")
+
+        if st.button("🏠 Home", use_container_width=True):
+            go_to('landing')
+            st.rerun()
+
+        if st.button("📚 Library", use_container_width=True):
+            go_to('library')
+            st.rerun()
+
+        st.markdown("---")
+        st.markdown("### Pathways")
+
+        if st.button("🧠 Understanding", use_container_width=True):
+            go_to('pathway1')
+            st.rerun()
+
+        if st.button("🔓 If You Tell", use_container_width=True):
+            go_to('pathway2')
+            st.rerun()
+
+        if st.button("⚓ Right Now Tools", use_container_width=True):
+            go_to('pathway3')
+            st.rerun()
+
+        if st.button("✍️ Finding Words", use_container_width=True):
+            go_to('finding_words')
+            st.rerun()
+
+        st.markdown("---")
+        st.markdown("### 🆘 Crisis Resources")
+        st.markdown("""
+        - **988** - Call or text
+        - **741741** - Text HOME
+        - **911** - Emergency
+        """)
+
+        st.markdown("---")
+        st.markdown("""
+        <p style="font-size: 0.8rem; color: #718096;">
+        Created by <a href="http://www.jocelynskillman.com" target="_blank">Jocelyn Skillman, LMHC</a><br><br>
+        <a href="https://jocelynskillmanlmhc.substack.com/" target="_blank">Read more about clinical UX</a>
+        </p>
+        """, unsafe_allow_html=True)
+
+# ============================================================
+# MAIN ROUTING
+# ============================================================
+def main():
+    render_sidebar()
+
+    page = st.session_state.page
+
+    if page == 'landing':
+        render_landing()
+    elif page == 'welcome':
+        render_welcome()
+    elif page == 'library':
+        render_library()
+    elif page == 'pathway1':
+        render_pathway1()
+    elif page == 'pathway2':
+        render_pathway2()
+    elif page == 'pathway3':
+        render_pathway3()
+    elif page == 'finding_words':
+        render_finding_words()
+    else:
+        render_landing()
+
+if __name__ == "__main__":
+    main()

requirements.txt

@@ -1,7 +1,2 @@
-streamlit==1.45.0
-anthropic==0.18.1
-python-dotenv==1.1.0
-pyttsx3==2.98
-google-generativeai
-gradio  # (Agar aap Gradio use kar rahe hain)
-openai
+streamlit==1.32.0
+python-dotenv==1.0.1