Environment variables for _AugmentedIntelligence
Recommended defaults for a dev workstation. Empty string means “unset / feature off until configured.”
Also see: RemoteCUDA.md (RemoteCUDA.md) · BuildConfigurations.md (BuildConfigurations.md) · PluginBridges.md (PluginBridges.md)
Quick start (dev defaults)
[bat]
REM — Core (usually leave unset; Settings/MySQL file wins) —
REM set AI_MYSQL_HOST=127.0.0.1
REM set AI_MYSQL_USER=root
REM set AI_MYSQL_PASSWORD=
REM set AI_MYSQL_DATABASE=cognitive_loop
REM — Remote CUDA worker (optional GPU box) —
set AI_REMOTE_CUDA_ENABLED=1
set AI_REMOTE_CUDA_HOST=127.0.0.1
set AI_REMOTE_CUDA_PORT=7720
set AI_REMOTE_CUDA_TOKEN=dev-token
set AI_REMOTE_CUDA_TIMEOUT_MS=15000
REM On the GPU machine:
REM set AI_REMOTE_CUDA_TOKEN=dev-token
REM set AI_REMOTE_CUDA_PORT=7720
REM python tools\remote_cuda_worker.py
Remote CUDA / HardwareAcceleration
| Variable | Default | Meaning |
| `AI_REMOTE_CUDA_ENABLED` | 1 | Master enable for _RemoteCuda client (0/false/off disables) |
| `AI_REMOTE_CUDA_HOST` | (empty) | Single worker host: host, host:port, or http://host:port |
| `AI_REMOTE_CUDA_HOSTS` | (empty) | Comma/space-separated worker list (pool; least-loaded pick) |
| `AI_REMOTE_CUDA_PORT` | 7720 | Default port when host has no :port |
| `AI_REMOTE_CUDA_TOKEN` | (empty) | Bearer / JSON token for worker auth |
| `AI_REMOTE_CUDA_TIMEOUT_MS` | 15000 | HTTP timeout for remote RPC |
| `AI_REMOTE_CUDA_BIND` | 0.0.0.0 | Worker listen address (remote_cuda_worker.py) |
| `AI_CUDA_REMOTE_HOST` | (empty) | Alias for AI_REMOTE_CUDA_HOST |
| `AI_HARDWARE_REMOTE_TOKEN` | (empty) | Fallback token if AI_REMOTE_CUDA_TOKEN unset |
In-app Settings globals (not env, but related):
cuda_accelleration_remote_enabled, cuda_accelleration_remote_hostname,
tensor_accelleration_remote_*, hardware_remote_rpc_port (default 7720),
hardware_remote_rpc_timeout_ms (15000), hardware_remote_rpc_token.
Worker: python tools/remote_cuda_worker.py
Client commands: cuda remote status|ping|hello|bench|set|job|tensor|checkpoint|…
Database / MySQL
| Variable | Default | Meaning |
| MySQL connection | from Settings / AD Credential Manager | Host app uses _Settings + optional CLI; env override patterns vary by build |
| Common convention | 127.0.0.1 / root / empty password / cognitive_loop | Local XAMPP/WAMP style lab |
If you use env-based secrets tooling, prefer OS Credential Manager / GPO (_Settings::ApplyAdPasswords) over plain env in production.
LLM / remote inference (non-CUDA)
| Variable | Default | Meaning |
| Provider API keys | unset | OpenAI/Grok/Gemini/etc. usually via Settings UI / encrypted store, not required env |
| Ollama / remote LLM | Settings | remote_llm_server_enabled + hostname in settings |
Bridges / games
| Variable | Default | Meaning | ||||
| Bridge tokens | configured in-app | Minecraft/SE/GameMod tokens live in bridge config files / commands (minecraft bridge configure …) | ||||
| Lab hosts | e.g. 10.0.0.151 | Documented in Minecraft/SE docs for EARTH/JUPITER | ||||
| AI_CREO_PROFILE | configurator | Creo strategy profile: live \ | batch \ | configurator \ | file \ | toolkit |
| AI_CREO_API | mock | Creo worker API: mock \ | vb \ | jlink \ | trail \ | toolkit |
| AI_CREO_HOST | 127.0.0.1 | Creo worker host | ||||
| AI_CREO_BRIDGE_PORT | 8777 | Creo worker TCP port | ||||
| AI_CREO_TOKEN | change-me | Shared bridge token | ||||
| AI_CREO_WORKDIR | creo/work | Sandbox for exports / jobs | ||||
| AI_CREO_WORKER | (unset) | Path to worker script/exe (tools/creo_worker/creo_worker.py) | ||||
| AI_CREO_EXE | (unset) | Path to Creo Parametric executable | ||||
| AI_CREO_DRY_RUN | on | Default dry-run for Creo exports | ||||
| AI_BLENDER_PROFILE | template | Blender profile: interactive \ | headless \ | template \ | file \ | advanced |
| AI_BLENDER_API | mock | mock or live worker | ||||
| AI_BLENDER_HOST / AI_BLENDER_PORT | 127.0.0.1 / 8775 | Blender worker TCP | ||||
| AI_BLENDER_TOKEN | change-me | Blender bridge token | ||||
| AI_BLENDER_WORKDIR | blender/work | Sandbox | ||||
| AI_BLENDER_WORKER / AI_BLENDER_EXE | (unset) | Worker script / blender.exe | ||||
| AI_BLENDER_DRY_RUN | on | Default dry-run | ||||
| AI_C4D_PROFILE | template | C4D profile: live \ | batch \ | template \ | file \ | sdk |
| AI_C4D_API | mock | mock or live worker | ||||
| AI_C4D_HOST / AI_C4D_PORT | 127.0.0.1 / 8776 | C4D worker TCP | ||||
| AI_C4D_TOKEN | change-me | C4D bridge token | ||||
| AI_C4D_WORKDIR | c4d/work | Sandbox | ||||
| AI_C4D_WORKER / AI_C4D_EXE | (unset) | Worker / Cinema 4D exe | ||||
| AI_C4D_DRY_RUN | on | Default dry-run | ||||
| AI_WEATHER_LAT / AI_WEATHER_LON | Seattle defaults | Weather/storm point | ||||
| AI_WEATHER_LOCATION | Seattle | Location label | ||||
| AI_WEATHER_NWS_AREA | (unset) | Optional US state (e.g. WA) for NWS alerts | ||||
| AI_WEATHER_MOCK / AI_WEATHER_FORCE_MOCK | off | Use fixtures (offline) | ||||
| AI_WEATHER_USER_AGENT | AugmentedIntelligence/… | Required style header for NWS | ||||
| AI_WEATHER_TIMEOUT | 20 | HTTP timeout seconds | ||||
| AI_NEMO_URL | http://127.0.0.1:8000 | NeMo Guardrails / NIM base URL | ||||
| AI_NEMO_TOKEN | (unset) | Optional bearer | ||||
| AI_NEMO_PROFILE | mock | mock \ | guardrails \ | nim \ | speech \ | … |
| AI_NEMO_DRY_RUN | on | Prefer local rails when not live | ||||
| AI_TF_PYTHON | python | Python for TF pipelines | ||||
| AI_TF_TRAIN_ROOT | tensorflow_training | Pipeline root | ||||
| AI_TF_TRAIN_PROFILE | mock | mock \ | pipeline \ | finetune \ | native | |
| AI_TF_TRAIN_DRY_RUN | on | Default dry-run for training jobs |
No required env for bridges; dry-run on by default in BridgeManager. Creo/Blender/C4D work offline with *_API=mock (default). Weather uses Open-Meteo + NWS + USGS — see WeatherHazards.md (WeatherHazards.md).
Plugins
| Variable | Default | Meaning |
| (none required) | — | Plugins load in-process; dynamic DLL via plugins load |
Sandbox paths: plugins/data/ (HostApi FS).
CUDA / GPU local
| Variable | Default | Meaning |
| `CUDA_VISIBLE_DEVICES` | (driver default) | Standard NVIDIA: which GPUs are visible |
| `CUDA_PATH` | toolkit install | MSVC/CUDA toolkit discovery (system install) |
Build-time: USE_CUDA / AI_CPU_ONLY in VS configurations (not runtime env).
Networking / portal
| Variable | Default | Meaning |
| Portal / web | Settings | program_state_portal_*, web server ports in settings |
| Hardware RPC port | 7720 | hardware_remote_rpc_port |
Testing / CI
| Variable | Default | Meaning |
| `AI_EXE` | (auto-discover) | Full path to _AugmentedIntelligence.exe for tests/run_command_batch.ps1 |
Recommended profiles
A. Laptop CPU-only
[bat]
set AI_REMOTE_CUDA_ENABLED=0
REM do not set AI_REMOTE_CUDA_HOST
B. Local GPU + local worker (same machine)
[bat]
set AI_REMOTE_CUDA_ENABLED=1
set AI_REMOTE_CUDA_HOST=127.0.0.1
set AI_REMOTE_CUDA_PORT=7720
set AI_REMOTE_CUDA_TOKEN=dev-token
C. Workstation + remote GPU server
[bat]
set AI_REMOTE_CUDA_ENABLED=1
set AI_REMOTE_CUDA_HOSTS=http://10.0.0.50:7720,http://10.0.0.51:7720
set AI_REMOTE_CUDA_TOKEN=prod-long-random-token
set AI_REMOTE_CUDA_TIMEOUT_MS=30000
D. Match HardwareAcceleration remote Tensor path
[bat]
set AI_REMOTE_CUDA_HOST=10.0.0.50
set AI_REMOTE_CUDA_PORT=7720
set AI_REMOTE_CUDA_TOKEN=same-as-hardware_remote_rpc_token
Also enable remote CUDA/tensor hostnames in Settings if you use the older HA code paths.
Security notes
- Never commit real tokens. Use dev-token only on localhost.
- Bind workers to LAN or localhost; use firewall + strong token in production.
- Kill-switch / safe mode cancels remote CUDA jobs.
- Remote matmul always falls back to local GEMM if the worker is down.
Verification
[text]
cuda remote refresh
cuda remote status
cuda remote ping
cuda remote bench 128
nn device list
nn device set global train remote
Worker logs should show /v1/ping and /infer or /v1/matmul hits.
Filed under: Uncategorized - @ July 18, 2026 6:32 am