fe43c62f56
🌲 Core Philosophy: "Was kostet die Frage eines Kindes?"
In the forest: priceless. In the system: measured. Pedagogically: teaches mindful questioning.
Changes:
- Added .env.template for API configuration
• OpenRouter, Claude API, OpenAI support
• Ollama (local AI) as free alternative
• Qdrant vector database configuration
• Token budget system for mindful learning
• Parental controls (PIN, reports)
- New AI Doktor module in mission selector
• Shows configured APIs and models
• Displays token budget & tracking status
• Lists active AI characters (Waldwächter)
• Links to token logs viewer
• Guides setup if .env missing
- Extended CLAUDE.md with Token Philosophy
• Educational reasoning behind token tracking
• Why it teaches reflection and quality
• Budget system explanation
• Implementation details
Philosophy:
Token tracking isn't restriction - it's mindfulness training.
Just as we teach not to waste water/food/paper, we teach
not to waste computational resources. Children learn to:
- Think before asking
- Value AI's thinking time
- Ask better quality questions
- Research independently first
Result: More thoughtful questions, deeper learning, respect for resources.
81 lines
2.3 KiB
Plaintext
81 lines
2.3 KiB
Plaintext
# ============================================================
|
|
# 🌲 Crumbforest Mission System - Environment Configuration
|
|
# ============================================================
|
|
#
|
|
# Philosophie: "Was kostet die Frage eines Kindes?"
|
|
# Im Wald unbezahlbar - aber Token helfen beim achtsamen Fragen.
|
|
# Jede Frage ist wertvoll, Token-Messung lehrt bewusstes Denken.
|
|
#
|
|
# ============================================================
|
|
|
|
# === API KEYS ===
|
|
|
|
# OpenRouter (Multi-Model Gateway)
|
|
OPENROUTER_API_KEY=""
|
|
OPENROUTER_MODEL="openai/gpt-3.5-turbo" # oder anthropic/claude-3-haiku
|
|
|
|
# Claude API (Anthropic)
|
|
CLAUDE_API_KEY=""
|
|
CLAUDE_MODEL="claude-3-haiku-20240307"
|
|
|
|
# OpenAI
|
|
OPENAI_API_KEY=""
|
|
OPENAI_MODEL="gpt-3.5-turbo"
|
|
|
|
# === LOCAL AI (Optional) ===
|
|
|
|
# Ollama (Local AI - kostenlos!)
|
|
OLLAMA_URL="http://localhost:11434"
|
|
OLLAMA_MODEL="llama3.2:3b"
|
|
USE_OLLAMA="false" # true = nutze Ollama statt API
|
|
|
|
# === VECTOR DATABASE ===
|
|
|
|
# Qdrant (für Maya-Eule Memory)
|
|
QDRANT_URL="http://localhost:6333"
|
|
QDRANT_COLLECTION="crumbforest_memory"
|
|
QDRANT_API_KEY="" # Optional, meist leer bei local
|
|
|
|
# === TOKEN BUDGET (Pädagogisch) ===
|
|
|
|
# Tages-Budget für bewusstes Fragen (in Tokens)
|
|
# 0 = unbegrenzt, >0 = Limit (z.B. 10000 = ~20 Fragen)
|
|
DAILY_TOKEN_BUDGET="0"
|
|
|
|
# Warnung bei hohem Token-Verbrauch
|
|
TOKEN_WARNING_THRESHOLD="1000"
|
|
|
|
# === LOGGING ===
|
|
|
|
# Token-Log speichern?
|
|
ENABLE_TOKEN_TRACKING="true"
|
|
|
|
# Log-Verzeichnis
|
|
LOG_DIR="${HOME}/.crumbforest_logs"
|
|
|
|
# === CHARAKTERE (AI Assistenten) ===
|
|
|
|
# Welche Charaktere sind aktiviert?
|
|
ENABLE_MAYAEULE="true" # Maya-Eule (Weisheit)
|
|
ENABLE_DEEPBIT="true" # Deepbit (Bash-Erklärer)
|
|
ENABLE_BUGSY="true" # Bugsy (Debugging)
|
|
ENABLE_SCHNIPPSI="true" # Schnippsi (Shell-Helfer)
|
|
ENABLE_TOBI="true" # Tobi (JSON/Daten)
|
|
|
|
# === SICHERHEIT ===
|
|
|
|
# Eltern-PIN für Einstellungen (Optional)
|
|
PARENT_PIN=""
|
|
|
|
# Automatische Token-Reports an Eltern?
|
|
SEND_TOKEN_REPORTS="false"
|
|
PARENT_EMAIL=""
|
|
|
|
# ============================================================
|
|
# HINWEISE:
|
|
# - Kopiere diese Datei zu .env und fülle deine Keys ein
|
|
# - .env wird NICHT ins Git committed (siehe .gitignore)
|
|
# - Mindestens einen API-Key ODER Ollama konfigurieren
|
|
# - Qdrant ist optional (für Maya-Eule Memory)
|
|
# ============================================================
|