{"repo": "cbetts1/ai-model-200b-", "files": {"config/aura.yaml": {"eedb907466790adaebaa1cb8516279ad2bd43e9d": {"t": 1785674400.6891322, "n": 7341, "h": "fb519f052e53c5ac", "c": "# config/aura.yaml — Main runtime configuration for AURA.\n#\n# Edit this file to configure the model backend, memory path, and features.\n# Do NOT commit secrets (API keys) to version control — use environment\n# variables or a separate secrets.yaml (git-ignored).\n\n# ── Model backend ──────────────────────────────────────────────────────────────\nmodel:\n  # \"echo\"   → offline echo backend for testing (no model server required)\n  # \"local\"  → run inference on this machine (requires llama-cpp-python or Ollama)\n  # \"remote\" → call an OpenAI-compatible HTTP endpoint\n  # \"router\" → try a prioritised list of backends (best for 200B-class quality)\n  # \"free\"   → auto-configure all free 200B-class cloud services from env vars\n  backend: free\n\n  local:\n    # \"llamacpp\" uses llama-cpp-python; \"ollama\" uses the Ollama client + server.\n    engine: llamacpp\n    model_path: models/weights/model.gguf\n    context_length: 4096\n    n_threads: 4         # set to your CPU core count\n\n  remote:\n    # Works with: OpenAI, Ollama (localhost), LM Studio, vLLM, LocalAI, etc.\n    base_url: http://localhost:11434/v1\n    api_key: \"\"          # leave blank for local servers; or use $OPENAI_API_KEY\n    model_name: llama3\n    temperature: 0.7\n    max_tokens: 4096\n\n  # ── Model Router — free 200B-class cloud services ────────────────────────\n  # When backend is \"router\", AURA tries each entry in order and uses the\n  # first one that succeeds.  Set one or more API keys as environment variables\n  # to unlock that backend (they're all free-tier or have free credits):\n  #\n  #   export GROQ_API_KEY=gsk_...         # https://console.groq.com (free)\n  #   export CEREBRAS_API_KEY=csk_...     # https://inference.cerebras.ai (free)\n  #   export OPENROUTER_API_KEY=sk-or-... # https://openrouter.ai (free models)\n  #   export TOGETHER_API_KEY=...         # https://api.together.xyz (free credits)\n  #   export HUGGINGFACE_API_KEY=hf_...   # https://huggingface.co/settings/tokens\n  #\n  # No key = that backend is skipped; if ALL are unset the echo backend is used.\n  router:\n    - name: groq\n      base_url: https://api.groq.com/openai/v1\n      api_key_env: GROQ_API_KEY\n      model_name: llama-3.3-70b-versatile\n      temperature: 0.7\n      max_tokens: 4096\n    - name: cerebras\n      base_url: https://api.cerebras.ai/v1\n      api_key_env: CEREBRAS_API_KEY\n      model_name: llama-3.3-70b\n      temperature: 0.7\n      max_tokens: 4096\n    - name: openrouter\n      base_url: https://openrouter.ai/api/v1\n      api_key_env: OPENROUTER_API_KEY\n      model_name: meta-llama/llama-3.3-70b-instruct:free\n      temperature: 0.7\n      max_tokens: 4096\n    - name: together\n      base_url: https://api.together.xyz/v1\n      api_key_env: TOGETHER_API_KEY\n      model_name: meta-llama/Llama-3.3-70B-Instruct-Turbo\n      temperature: 0.7\n      max_tokens: 4096\n    - name: huggingface\n      base_url: https://api-inference.huggingface.co/v1\n      api_key_env: HUGGINGFACE_API_KEY\n      model_name: microsoft/Phi-4\n      temperature: 0.7\n      max_tokens: 2048\n    - name: echo\n      base_url: \"\"     # always available — offline fallback\n\n  # ── Model scaling presets ─────────────────────────────────────────────────\n  # Choose a preset to automatically configure context length, max_tokens,\n  # and recommended hardware.  Override individual settings above as needed.\n  #\n  # Presets:\n  #   small   — 7B params, 4K context, runs on 8GB RAM / mobile\n  #   medium  — 13-34B params, 8K context, runs on 16GB RAM / laptop\n  #   large   — 70B params, 16K context, runs on 32GB+ RAM / desktop GPU\n  #   max     — 200B+ params, 32K+ context, cloud GPU cluster recommended\n  #             (free via Groq/Cerebras/OpenRouter — no local hardware needed)\n  #\n  # The preset is informational — AURA works with any model size.\n  # Set backend to \"free\" or \"router\" and add your API keys for best results.\n  preset: max\n\n# ── Memory ─────────────────────────────────────────────────────────────────────\nmemory:\n  # Directory where session history files are stored.\n  # Default: ~/.aura/memory/\n  dir: ~/.aura/memory\n\n# ── Identity ───────────────────────────────────────────────────────────────────\nidentity:\n  config_path: config/identity.yaml\n\n# ── HTTP API server ────────────────────────────────────────────────────────────\n# AURA is cloud-native: run `aura serve` to start the API so it can be\n# accessed from web pages, other servers, or any device on the network.\n# The web chat UI is served at the root URL (/).\nserver:\n  host: 0.0.0.0         # \"0.0.0.0\" to listen on all interfaces\n  port: 8000\n  # Set to a non-empty string to require an Authorization: Bearer <token> header.\n  # Use $AURA_API_TOKEN env var in production — never commit real tokens.\n  api_token: \"\"\n\n# ── Admin API ──────────────────────────────────────────────────────────────────\n# Set AURA_ADMIN_TOKEN to a strong random string to enable the admin dashboard\n# and remote-maintenance API at /admin.\n#\n# The admin API provides:\n#   - System metrics (uptime, memory, Python version)\n#   - Model / cloud-backend status and reconnect\n#   - Server log tail\n#   - Runtime config reload\n#   - Graceful restart signalling\n#\n# NEVER commit a real token to version control.  Use the environment variable:\n#   export AURA_ADMIN_TOKEN=$(python3 -c \"import secrets; print(secrets.token_hex(32))\")\nadmin:\n  # Leave empty to disable admin endpoints entirely.\n  # In production always use the env var: $AURA_ADMIN_TOKEN\n  admin_token: \"\"\n\n# ── Features ───────────────────────────────────────────────────────────────────\nfeatures:\n  voice: true           # Web Speech API voice input (browser native)\n  video: false          # set true when video integration is ready\n  screen_share: false   # set true when screen-share integration is ready\n  web_ui: true          # serve the web chat interface at /\n  templates: true       # enable pre-fab template system (20 built-in)\n  streaming: true       # SSE streaming via /v1/chat/stream\n  agent_loop: true      # multi-step agentic tool use\n\n# ── Governance ──────────────────────────────────────────────────────────────────\n# AURA refuses requests that violate legal or ethical boundaries.\n# The governance engine is always active and cannot be disabled.\ngovernance:\n  enabled: true         # always on — cannot be turned off\n\n# ── Storage ────────────────────────────────────────────────────────────────────\n# Pluggable backend for persistent key-value data.\n# \"file\"   → JSON files in ~/.aura/storage/  (default, zero dependencies)\n# \"sqlite\" → SQLite database at ~/.aura/storage/aura.db  (better performance)\nstorage:\n  backend: file\n  dir: ~/.aura/storage\n\n# ── Plugins ────────────────────────────────────────────────────────────────────\n# AURA loads extra tools from .py files in plugin directories.\n# Drop a file with a Tool subclass into any listed directory and it will be\n# auto-discovered on next start.  Plugins cannot override built-in tools.\nplugins:\n  enabled: true\n  dirs:\n    - plugins/            # project-local plugins directory\n    - ~/.aura/plugins/    # user-global plugins directory\n\n# ── Self-Build ─────────────────────────────────────────────────────────────────\n# AURA's self-improvement proposal system.  Proposals are never auto-applied —\n# they wait for explicit human approval.\nself_build:\n  enabled: true\n  store_dir: ~/.aura/self_build\n"}}, "docker-compose.yaml": {"eedb907466790adaebaa1cb8516279ad2bd43e9d": {"t": 1785674400.8244214, "n": 1521, "h": "ba592fafc560f5b3", "c": "# docker-compose.yaml — One-command AURA cloud deployment.\n#\n# Usage:\n#   docker compose up --build\n#\n# To enable free 200B-class AI, set one or more API keys before starting:\n#   export GROQ_API_KEY=gsk_...        # https://console.groq.com (free)\n#   export CEREBRAS_API_KEY=csk_...    # https://inference.cerebras.ai (free)\n#   export OPENROUTER_API_KEY=sk-or-.. # https://openrouter.ai (free models)\n#   export TOGETHER_API_KEY=...        # https://api.together.xyz (free credits)\n#   docker compose up --build\n#\n# AURA will be available at http://localhost:8000\n\nservices:\n  aura:\n    build: .\n    ports:\n      - \"8000:8000\"\n    environment:\n      # Set a bearer token for auth (optional — leave empty for open access)\n      - AURA_API_TOKEN=${AURA_API_TOKEN:-}\n      # Admin dashboard token — generate with:\n      #   python3 -c \"import secrets; print(secrets.token_hex(32))\"\n      - AURA_ADMIN_TOKEN=${AURA_ADMIN_TOKEN:-}\n      # Free 200B-class model API keys (all optional — any one unlocks quality AI)\n      - GROQ_API_KEY=${GROQ_API_KEY:-}\n      - CEREBRAS_API_KEY=${CEREBRAS_API_KEY:-}\n      - OPENROUTER_API_KEY=${OPENROUTER_API_KEY:-}\n      - TOGETHER_API_KEY=${TOGETHER_API_KEY:-}\n      - HUGGINGFACE_API_KEY=${HUGGINGFACE_API_KEY:-}\n    volumes:\n      # Persist conversation memory across container restarts\n      - aura-memory:/root/.aura/memory\n      # Persist storage (notes, key-value data)\n      - aura-storage:/root/.aura/storage\n    restart: unless-stopped\n\nvolumes:\n  aura-memory:\n  aura-storage:\n"}}}}