Environment Variables

10.3. Environment Variables

1. Generated Environment Matrix

Variable	Service	Default	Description
AIRFLOW_SOURCE	airflow	disabled	Deployment source for Airflow: container or disabled (default).
AIRFLOW_PORT	airflow	63070	Host port for the Airflow Web UI + REST API (in-container 8080).
AIRFLOW_DB_USER	airflow	airflow	Postgres role that owns the airflow metadata database on supabase-db. Supabase DB init creates it; Airflow connects with it and AIRFLOW_DB_PASSWORD.
AIRFLOW_DB_PASSWORD	airflow	atlas-db-password	Auto-generated by bootstrapper. Postgres role for the Airflow metadata DB.
AIRFLOW_DB_USER_URI	airflow	airflow	Bootstrapper-synchronized percent-encoded AIRFLOW_DB_USER for DSN userinfo.
AIRFLOW_DB_PASSWORD_URI	airflow	atlas-db-password	Bootstrapper-synchronized percent-encoded AIRFLOW_DB_PASSWORD for DSN userinfo.
AIRFLOW_ATLAS_DB_USER	airflow	atlas_airflow_reader	Read-only login seeded into the postgres_supabase Airflow Connection.
AIRFLOW_ATLAS_DB_PASSWORD	airflow	atlas-db-password	Auto-generated password for Airflow DAGs' read-only Atlas database connection.
AIRFLOW_FERNET_KEY	airflow		Auto-generated. Fernet key for encrypting Connection passwords.
AIRFLOW_SECRET_KEY	airflow		Auto-generated. Mapped to AIRFLOW__API__SECRET_KEY ([api] secret_key) in every Airflow container. Execution API JWTs use AIRFLOW_JWT_SECRET instead.
AIRFLOW_JWT_SECRET	airflow		Auto-generated. Mapped to AIRFLOW__API_AUTH__JWT_SECRET; signs Execution API and /api/v2/ auth-token JWTs. Every Airflow process must share it, or cross-process verification fails with 403 InvalidSignatureError. Kept on warm restart.
AIRFLOW_ADMIN_PASSWORD	airflow		Auto-generated. Admin login. Username is hardcoded 'admin'.
AIRFLOW_WEBSERVER_SCALE	airflow		Set by ./start.sh from AIRFLOW_SOURCE; do not edit. 1 runs airflow-webserver (UI and REST API), 0 does not.
AIRFLOW_SCHEDULER_SCALE	airflow		Set by ./start.sh from AIRFLOW_SOURCE; do not edit. 1 runs airflow-scheduler, 0 does not.
AIRFLOW_DAG_PROCESSOR_SCALE	airflow		Set by ./start.sh from AIRFLOW_SOURCE; do not edit. 1 runs airflow-dag-processor. Airflow 3 needs it: it parses DAG files into the metadata DB, which the scheduler does not do.
AIRFLOW_INIT_SCALE	airflow		Set by ./start.sh from AIRFLOW_SOURCE; do not edit. 1 runs the one-shot airflow-init container at stack start, 0 does not.
AIRFLOW_WEBSERVER_MEMORY_LIMIT	airflow	2g	Container memory limit for airflow-webserver (deploy.resources.limits.memory). The api-server process is Python + Flask/FAB; 2 g is comfortable on a 32 GB host and prevents runaway memory from large DAG list pages.
AIRFLOW_WEBSERVER_CPU_LIMIT	airflow	1.0	Container CPU limit for airflow-webserver (deploy.resources.limits.cpus). 1.0 core is ample for the API/UI server under typical load.
AIRFLOW_SCHEDULER_MEMORY_LIMIT	airflow	2g	Container memory limit for airflow-scheduler (deploy.resources.limits.memory). The LocalExecutor scheduler forks task processes in-process; 2 g caps runaway task forking.
AIRFLOW_SCHEDULER_CPU_LIMIT	airflow	1.0	Container CPU limit for airflow-scheduler (deploy.resources.limits.cpus). 1.0 core is sufficient for LocalExecutor task scheduling on a dev/staging host.
AIRFLOW_DAG_PROCESSOR_MEMORY_LIMIT	airflow	2g	Container memory limit for airflow-dag-processor (deploy.resources.limits.memory). DAG parsing is Python-heavy but short-lived per cycle; 2 g prevents unbounded growth from parsing pathological DAG files.
AIRFLOW_DAG_PROCESSOR_CPU_LIMIT	airflow	1.0	Container CPU limit for airflow-dag-processor (deploy.resources.limits.cpus). 1.0 core covers the periodic DAG file parsing cycle.
AIRFLOW_IMAGE	airflow	apache/airflow:3.3.2	Container image for airflow-webserver.
ASSET_BAKER_SOURCE	asset-baker	disabled	Deployment mode for the Blender HP→LP bake worker. container-cpu runs deterministic Cycles-CPU bakes (GPU-optional, deferred).
ASSET_BAKER_PORT	asset-baker	63052	Host port for the asset-baker API (in-container listen port is 8096).
ASSET_BAKER_SCALE	asset-baker		Set by ./start.sh from ASSET_BAKER_SOURCE; do not edit. 1 runs asset-baker (container-cpu), 0 does not.
ASSET_BAKER_PLATFORM	asset-baker	linux/amd64	Platform the asset-baker image is built and run for. The Blender build is x86_64-only, so arm64 hosts (Apple Silicon) run it under emulation; keep linux/amd64.
ASSET_BAKER_ENDPOINT	asset-baker		In-network URL for the asset-baker API.
ASSET_BAKER_API_TOKEN	asset-baker		Auto-generated bearer token required by all Asset Baker data and processing routes; /health remains unauthenticated.
ASSET_BAKER_ALLOWED_INPUT_BUCKETS	asset-baker		Optional comma- or space-separated reference-route bucket allowlist; blank follows MINIO_BUCKET_ASSET_INPUTS.
ASSET_BAKER_BLENDER_VERSION	asset-baker	4.3.2	Pinned headless Blender version (>= 4.3) downloaded into the asset-baker image. Blender bundles its own Python + numpy for the bake script.
ASSET_BAKER_BLENDER_SHA256	asset-baker	4da1c956673c0485e63054e563ee69198cc8f80d8157dd7592dffc8a6a5592e6	SHA-256 checksum verified before extracting the pinned Blender archive. Update it with ASSET_BAKER_BLENDER_VERSION from Blender's published checksum file.
ASSET_BAKER_ARTIFACT_DIR	asset-baker	/data/artifacts	Local artifact cache used when MinIO upload is disabled or for direct downloads.
ASSET_BAKER_MINIO_ENABLED	asset-baker	true	When true, baked LP GLBs + textures are written to MinIO using content-addressed keys.
ASSET_BAKER_MINIO_BUCKET	asset-baker	asset-baker	MinIO bucket for baked LP GLB + texture outputs.
ASSET_BAKER_MINIO_ENDPOINT	asset-baker		In-network MinIO S3 API endpoint used by the worker.
ASSET_BAKER_MINIO_ACCESS_KEY	asset-baker		Optional MinIO access key override. Empty uses the generated asset-baker scoped MinIO account.
ASSET_BAKER_MINIO_SECRET_KEY	asset-baker		Optional MinIO secret key override. Empty uses the generated asset-baker scoped MinIO account.
ASSET_BAKER_TARGET_TRIS	asset-baker	39000	Default low-poly triangle budget the decimate stage targets (overridable per request).
ASSET_BAKER_TEX_SIZE	asset-baker	2048	Default baked BaseColor/Normal texture resolution (square). Higher resolutions take longer on CPU; overridable per request.
ASSET_BAKER_CANONICAL_SIZE	asset-baker	4.0	Canonical max-dimension (units) each mesh is scaled to before remesh so the relative voxel/ray heuristics are stable regardless of the raw GLB import scale.
ASSET_BAKER_BRIGHTNESS_MIN	asset-baker	0.05	Mean-brightness QA gate: a baked BaseColor whose mean < this is a silently-failed (black) bake and is refused with a non-zero result.
ASSET_BAKER_TIMEOUT_SECONDS	asset-baker	600	Hard wall-clock ceiling for a single bake subprocess; exceeded bakes are killed and reported as timeout.
ASSET_BAKER_MAX_UPLOAD_MB	asset-baker	200	Maximum accepted GLB input size (MiB). Larger uploads/refs are rejected before a Blender process is spawned.
ASSET_BAKER_CONCURRENCY	asset-baker	1	Maximum concurrent bake subprocesses. Default 1 — a single Cycles bake saturates CPU and concurrent bakes are net-negative.
ASSET_BAKER_IMAGE	asset-baker	python:3.12.9-slim	Base image; the Dockerfile downloads a pinned headless Blender (ASSET_BAKER_BLENDER_VERSION) on top.
ASSET_WORKER_SOURCE	asset-worker	disabled	Deployment mode for the glTF post-processing worker.
ASSET_WORKER_PORT	asset-worker	63053	Host port for the asset-worker API (in-container listen port is 8095).
ASSET_WORKER_SCALE	asset-worker		Set by ./start.sh from ASSET_WORKER_SOURCE; do not edit. 1 runs asset-worker (container), 0 does not.
ASSET_WORKER_ENDPOINT	asset-worker		In-network URL for the asset-worker API.
ASSET_WORKER_API_TOKEN	asset-worker		Auto-generated bearer token required by all Asset Worker data and processing routes; /health remains unauthenticated.
ASSET_WORKER_ALLOWED_INPUT_BUCKETS	asset-worker		Optional comma- or space-separated reference-route bucket allowlist; blank follows MINIO_BUCKET_ASSET_INPUTS.
ASSET_WORKER_GLTF_TRANSFORM_VERSION	asset-worker	4.5.1	Expected pinned @gltf-transform/cli lock version; overrides must match package.json and package-lock.json.
ASSET_WORKER_MAX_UPLOAD_MB	asset-worker	200	Maximum GLB upload or MinIO reference size in MiB. Input is streamed and rejected before transformation when this limit is exceeded.
ASSET_WORKER_TIMEOUT_SECONDS	asset-worker	300	Per-command timeout for gltf-transform inspect, validate, and optimize subprocesses.
ASSET_WORKER_CONCURRENCY	asset-worker	1	Maximum concurrent glTF mutation requests; saturated workers reject new work with HTTP 429 before request-body parsing or object-store fetch.
ASSET_WORKER_ARTIFACT_DIR	asset-worker	/data/artifacts	Local artifact cache used when MinIO upload is disabled or for direct downloads.
ASSET_WORKER_MINIO_ENABLED	asset-worker	true	When true, optimized GLBs are written to MinIO using content-addressed keys.
ASSET_WORKER_MINIO_BUCKET	asset-worker	asset-worker	MinIO bucket for optimized GLB outputs.
ASSET_WORKER_MINIO_ENDPOINT	asset-worker		In-network MinIO S3 API endpoint used by the worker.
ASSET_WORKER_MINIO_ACCESS_KEY	asset-worker		Optional MinIO access key override. Empty uses the generated asset-worker scoped MinIO account.
ASSET_WORKER_MINIO_SECRET_KEY	asset-worker		Optional MinIO secret key override. Empty uses the generated asset-worker scoped MinIO account.
ASSET_WORKER_IMAGE	asset-worker	python:3.12.9-slim	Container image for asset-worker.
BACKEND_DB_USER	backend	atlas_backend	Scoped PostgreSQL login for Backend and its Celery worker.
BACKEND_DB_PASSWORD	backend	atlas-db-password	Auto-generated password for the scoped Backend database role.
BACKEND_DB_USER_URI	backend	atlas_backend	Bootstrapper-synchronized percent-encoded BACKEND_DB_USER for DSN userinfo.
BACKEND_DB_PASSWORD_URI	backend	atlas-db-password	Bootstrapper-synchronized percent-encoded BACKEND_DB_PASSWORD for DSN userinfo.
BACKEND_SOURCE	backend	container	Adaptive core service; always container. Single-option.
BACKEND_PORT	backend	63093	Host port that publishes the Backend API (in-container port 8000).
BACKEND_SCALE	backend		Always 1 (BACKEND_SOURCE is single-valued).
MAX_UPLOAD_BYTES	backend	104857600	Maximum accepted /storage/upload body size in bytes before the backend returns 413.
COMFYUI_MAX_IMAGE_BYTES	backend	20971520	Maximum input or generated-image bytes accepted by the Backend's ComfyUI bridge before it aborts the transfer.
COMFYUI_INIT_IMAGE_TRUSTED_ORIGINS	backend		Comma-separated exact HTTPS origins allowed for remote ComfyUI init-image URLs. Blank rejects all remote URLs; data URIs remain available.
COMFYUI_COMPLETION_TIMEOUT_SECONDS	backend	300	End-to-end deadline in seconds for synchronous ComfyUI generation and workflow requests. The Backend poller and bundled Open WebUI tool share this default.
RAG_INGESTION_MAX_FILE_BYTES	backend	104857600	Maximum bytes loaded from any single mounted or MinIO RAG corpus file. Oversize files fail ingestion before unbounded allocation.
RAG_INGESTION_MAX_CORPUS_BYTES	backend	1073741824	Maximum aggregate bytes loaded by one RAG corpus discovery pass across mounted or MinIO files.
RAG_INGESTION_MAX_FILES	backend	10000	Maximum number of files accepted in one mounted or MinIO RAG corpus discovery pass.
RAG_INGESTION_CORPUS_ROOT	backend	/app/corpus	Container path under which operator-provided read-only RAG corpus mounts are resolved. Mount the same path into Backend and Celery worker when asynchronous ingestion is enabled.
CHONKIE_SEMANTIC_EMBEDDING_MODEL	backend	minishlab/potion-base-32M	Non-empty Chonkie embedding model reference used by semantic chunking. Blank values use the bundled default. Shared by Backend and Celery ingestion workers.
LIGHTRAG_PIPELINE_STATUS_TIMEOUT_SECONDS	backend	30	Finite per-request timeout in seconds for LightRAG pipeline-status polling. Values must be greater than 0 and no greater than 3600; invalid values use 30 seconds. Shared by Backend and Celery ingestion workers.
RAG_INGESTION_TTL_SECONDS	backend	604800	Redis retention in seconds for RAG ingestion records and idempotency keys. Values must be from 60 through 31536000 (one year); invalid values use the seven-day default. Shared by Backend and Celery ingestion workers.
RAG_INGESTION_EXECUTION_LEASE_SECONDS	backend	30	Renewable execution-ownership lease for RAG ingestion workers. Must be an integer from 10 through 300; duplicate deliveries retry until the active owner completes, an exact-owner recovery claim succeeds, or the lease expires.
BACKEND_STATE_STORE_MODE	backend	redis	Shared operation-state mode: 'redis' (durable, multi-process default) or 'memory' (explicit single-process testing/ephemeral mode). Redis outages fail stateful routes with HTTP 503; they never trigger an in-memory fallback.
RESEARCH_SESSION_LEASE_SECONDS	backend	300	Maximum age in seconds for pending research work or a running session heartbeat before the Backend atomically marks the abandoned session failed.
RESEARCH_MAX_CONCURRENT	backend	4	Maximum research sessions admitted concurrently per Backend process. Excess start requests fail before database work with HTTP 429.
BACKEND_HEAVY_WORK_CONCURRENCY	backend	4	Maximum /api/chunk and /api/rag/evaluate jobs the Backend runs at once, on a pool of its own so they cannot starve other routes. Further calls get HTTP 503 with Retry-After until a slot frees.
BACKEND_HEAVY_WORK_TIMEOUT_SECONDS	backend	600	Seconds a /api/chunk or /api/rag/evaluate call waits for its result before answering HTTP 504. The job keeps its concurrency slot until its thread actually finishes.
BACKEND_PG_POOL_MIN	backend	1	Minimum connections kept warm in the Backend's shared asyncpg pool. Short-lived DB operations draw from it instead of opening a new connection; long-hold paths (memory vector reconcile) keep dedicated connections.
BACKEND_PG_POOL_MAX	backend	10	Maximum connections the Backend's shared asyncpg pool opens per database URL. Concurrent short-lived DB operations beyond this queue for a free slot rather than over-subscribing Postgres.
BACKEND_STORAGE_ALLOWED_BUCKETS	backend	default	Comma-separated Supabase Storage bucket allowlist accepted by /storage/upload.
BACKEND_MEDIA_INPUT_BUCKET	backend	default	Atlas storage bucket the media gateway hosts image_to_3d inputs in when a provider (e.g. Tripo) rejects data-URI inputs. Objects are written under the media-inputs/ prefix.
BACKEND_MEDIA_INPUT_PUBLIC_BASE_URL	backend		Optional public base URL for hosted media inputs. When set, hosted image_to_3d input URLs are <base>/<bucket>/<key>, so the provider's cloud can fetch them through a reachable ingress. Empty uses the storage client's public URL.
MEDIA_REQUEST_MAX_BYTES	backend	41943040	Maximum POST /media/generate request body size in bytes. Oversized declared and streamed bodies are rejected with HTTP 413 before route parsing.
MEDIA_INPUT_MAX_BYTES	backend	26214400	Maximum decoded image data-URI size in bytes, including conditioned output. Oversized base64 payloads are rejected before decoding.
MEDIA_INPUT_MAX_PIXELS	backend	40000000	Maximum decoded or conditioned image pixel count. Dimensions are checked before full Pillow conversion to bound decompression and canvas memory.
MEDIA_BUDGET_ENABLED	backend	false	Enables media spend accounting and budget enforcement. When off, ordinary operations record no spend rows and enforce no caps; ambiguous FAL submissions still create a recovery row.
MEDIA_BUDGET_STORE	backend	postgres	Ledger/recovery store: 'postgres' (durable, media_spend_ledger table) or 'memory' (process-local, non-durable — testing/ephemeral). Ambiguous FAL submissions use it for recovery even when budget enforcement is disabled.
MEDIA_BUDGET_CURRENCY	backend	USD	Currency label recorded on media spend ledger rows.
MEDIA_BUDGET_DEFAULT_USD	backend		Default per-consumer/project spend cap in USD. Empty = no default cap (ledger records without hard-stop). Per-consumer overrides go in MEDIA_BUDGET_CONSUMER_CAPS.
MEDIA_BUDGET_CONSUMER_CAPS	backend		Optional JSON map of per-scope caps, keyed by 'consumer' or 'consumer:project', e.g. {"rag-showcase":25,"rag-showcase:demo":5}. Overrides MEDIA_BUDGET_DEFAULT_USD for matching scopes. A 'consumer' key still caps each project separately, and user callers choose the project name, so it is not a per-user total.
MEDIA_DISABLED_PROVIDERS	backend		Comma-separated media providers to kill-switch off (e.g. 'fal', 'comfyui'). A disabled provider hard-stops its own submissions (/media/generate, /comfyui/generate and /comfyui/workflow, whether or not MEDIA_BUDGET_ENABLED is set) without downing the gateway or other providers.
MEDIA_BUDGET_ALLOW_UNKNOWN_COST	backend	false	When false (default), a budgeted submission for a model with no known cost is rejected rather than billed an unknown amount (never silently $0). Set true to allow unknown-cost submissions under a cap.
MEDIA_BUDGET_RETENTION_DAYS	backend		Optional retention window in days for media spend ledger rows; empty keeps rows indefinitely (no pruning). When set, an hourly background sweep prunes rows older than the window.
MEDIA_OPERATION_TTL_SECONDS	backend	604800	Redis retention in seconds for hosted media operation state. The default keeps terminal results available for seven days across Backend replicas and restarts.
MEDIA_LEDGER_RECOVERY_BATCH_SIZE	backend	100	Maximum media ledger intents loaded in one bounded Redis recovery page. Must be an integer from 1 through 500.
MEDIA_LEDGER_RECOVERY_MAX_CYCLES	backend	4	Maximum bounded media ledger pages processed per 30-second recovery poll. Must be an integer from 1 through 50; the cursor resumes on the next poll.
BACKEND_CORS_ORIGINS	backend	*	Comma-separated browser origins accepted by the backend CORS middleware. Default preserves local-development permissiveness. Requests through Kong are also limited to local origins plus KONG_CORS_EXTRA_ORIGINS.
BACKEND_CORS_ALLOW_ORIGIN_REGEX	backend		Optional regex accepted by the backend CORS middleware for wildcard subdomain policies.
BACKEND_DEV_RELOAD	backend	false	Opt-in uvicorn --reload for live development. Off by default: with backend_plugins bind-mounted, host-side git churn in the plugin tree would restart or crash-loop the backend. Plugins install at boot, so recreate the backend to apply plugin changes.
BACKEND_KONG_AUTH	backend	disabled	Default gateway authentication for the api.localhost backend route: disabled (default, local development) or key-auth (requires an apikey header). A backend plugin's plugin.yml auth: setting (open or key-auth) overrides it per route prefix.
BACKEND_KONG_API_KEY	backend		Auto-generated key accepted by Kong when BACKEND_KONG_AUTH=key-auth. Send as apikey: <value>.
BACKEND_IDENTITY_AUTH	backend	required	Application identity boundary for backend routes: required (default) accepts Supabase user JWTs or the trusted internal-service token; disabled is an explicit emergency rollback mode.
BACKEND_INTERNAL_API_TOKEN	backend		Auto-generated full-operator bearer token. Reserve for trusted administration; first-party services use narrower caller tokens.
BACKEND_N8N_API_TOKEN	backend		Auto-generated n8n bearer accepted only by the research, memory-automation, legacy ComfyUI, and storage-upload routes used by bundled workflows.
BACKEND_NOTEBOOK_API_TOKEN	backend		Auto-generated bearer token exposed to JupyterHub and accepted only by stateless backend document extraction, chunking, and RAG evaluation routes.
BACKEND_OPEN_WEBUI_API_TOKEN	backend		Auto-generated Open WebUI bearer accepted only by memory and legacy ComfyUI routes used by bundled server-side tools and functions.
BACKEND_PLUGINS_PIP_TIMEOUT_SECONDS	backend	300	Timeout in seconds for pip install of backend plugin requirements (plugin_seam) at startup. The install runs at backend import time, so an unbounded hang would block startup; raise for slow networks.
BACKEND_PLUGINS_SITE_DIR	backend	/tmp/atlas-plugins-site	Writable directory for plugin requirements (pip --target), prepended to sys.path before plugins load; site-packages is root-owned. If it cannot be created, pip installs untargeted. Emptied at each start; a custom directory only once it has the .atlas-plugin-site marker.
LANGMEM_ENABLED	backend	true	Enables the Backend memory service in backend and celery-worker. false makes memory operations fail with "LangMem memory service is disabled".
LANGMEM_MEMORY_NAMESPACE	backend	default	Default memory namespace (passed as LANGMEM_NAMESPACE to backend and celery-worker). A request that names its own namespace overrides it.
LANGMEM_AUTO_CONSOLIDATE	backend	true	Reserved — not yet honored. Memory consolidation currently runs only on demand via POST /memory/consolidate; there is no background scheduler, so changing this value has no effect today.
LANGMEM_CONSOLIDATION_INTERVAL	backend	86400	Reserved — not yet honored (see LANGMEM_AUTO_CONSOLIDATE). Interval in seconds a future auto-consolidation scheduler would use.
LANGMEM_MAX_FACTS_PER_USER	backend	1000	Maximum active memory facts per user in backend and celery-worker. Extraction stops storing new facts at the limit, consolidation deactivates the oldest excess facts, and a restore beyond it fails with 409.
LANGMEM_EXTRACTION_MODEL	backend		LiteLLM model id (empty = highest-priority active content model resolved by model_resolver from YAML catalogs + env).
LANGMEM_EMBEDDING_MODEL	backend		Optional LiteLLM embedding model id. Empty falls back to LITELLM_EMBEDDING_MODEL.
LANGMEM_EMBEDDING_DIM	backend	768	Validated output dimension shared by the selected LangMem embedding model, Backend pgvector queries, and the Supabase memory schema migration. Curated wizard selections derive this from services/*/models.yaml; custom models must declare it explicitly (1..4000, the pgvector halfvec HNSW limit).
GRAPHITI_ENABLED	backend	false	Disabled-by-default backend-only Graphiti temporal graph memory experiment. Does not expose Graphiti to agents or run a Graphiti service.
GRAPHITI_GROUP_ID_PREFIX	backend	atlas	Prefix for backend Graphiti group_id values. Final shape: atlas:<project>:backend:<namespace>:user:<uuid>.
GRAPHITI_DEFAULT_NAMESPACE	backend	langmem	Default backend Graphiti namespace segment used when projecting selected LangMem facts/events.
GRAPHITI_LLM_MODEL	backend		Optional LiteLLM model id for Graphiti extraction. Empty falls back to LANGMEM_EXTRACTION_MODEL, then LITELLM_DEFAULT_MODEL.
GRAPHITI_EMBEDDING_MODEL	backend		Optional LiteLLM embedding model id for Graphiti. Empty falls back to LANGMEM_EMBEDDING_MODEL, then LITELLM_EMBEDDING_MODEL.
GRAPHITI_EXPOSE_TO_AGENTS	backend	false	Must remain false for this backend-only evaluation; Hermes/OpenClaw exposure and Graphiti MCP are deferred.
RAGAS_EVALUATOR_MODEL	backend		Optional LiteLLM model id for Backend Ragas evaluation. Empty falls back to LITELLM_DEFAULT_MODEL.
RAGAS_EMBEDDINGS_MODEL	backend		Optional LiteLLM embedding model id for Backend Ragas answer-relevancy evaluation. Empty falls back to LITELLM_EMBEDDING_MODEL.
RAY_JOB_API_TOKEN	backend		Auto-generated bearer token required by every /api/ray/* route. Send as Authorization: Bearer <value>.
LIGHTRAG_RERANK_ADAPTER_ENABLED	backend	false	Opt-in for the backend rerank adapter route POST /lightrag/rerank. When true and TEI_RERANKER_SOURCE is enabled, LightRAG reranks through the adapter and consumer profiles may set enable_rerank=true. Direct LightRAG-to-TEI rerank is unsupported (payload shapes differ).
LIGHTRAG_RERANK_ADAPTER_TOKEN	backend		Auto-generated bearer token guarding the /lightrag/rerank adapter route. Handed to LightRAG as RERANK_BINDING_API_KEY so the two share one secret. Send as Authorization: Bearer <value>.
LIGHTRAG_RERANK_ADAPTER_TIMEOUT_SECONDS	backend	30	Finite per-request timeout in seconds for /lightrag/rerank calls to TEI. Must be greater than 0 and no greater than 3600; invalid values fail Backend startup.
BACKEND_IMAGE	backend	python:3.12.14-bookworm@sha256:581429e3df12d76e6af4be5ab7d0e7fc2013eb57dc23d2de691411c8efdbb970	Container image for backend.
BACKUP_SOURCE	backup	disabled	Deployment mode: container (enable the runner) or disabled (default).
BACKUP_BUCKET	backup	atlas-backups	Target bucket for backups.
BACKUP_S3_MODE	backup	local	Credential/dependency mode: local uses on-stack MinIO and permits the documented MINIO_ROOT_* upgrade fallback; external requires dedicated BACKUP_S3_* credentials and has no MinIO runtime dependency.
BACKUP_S3_ENDPOINT	backup	http://minio:9000	S3 endpoint as a strict HTTP(S) origin: canonical DNS/IPv4 or bracketed IPv6, optional canonical port, and no userinfo/path/query/fragment. The default is on-stack MinIO.
BACKUP_S3_ACCESS_KEY	backup		Dedicated S3 access key. Required for external endpoints; local MinIO may fall back to MINIO_ROOT_USER for upgrade compatibility.
BACKUP_S3_SECRET_KEY	backup		Dedicated S3 secret key. Required for external endpoints; local MinIO may fall back to MINIO_ROOT_PASSWORD for upgrade compatibility.
BACKUP_S3_REGION	backup	us-east-1	S3 signing and bucket-creation region.
BACKUP_S3_SESSION_TOKEN	backup		Optional S3 security token for temporary credentials.
BACKUP_S3_TLS_VERIFY	backup	true	Verify HTTPS certificates. false is valid only with an https endpoint and explicitly enables the MinIO client's insecure certificate mode.
BACKUP_S3_ALIAS_URL	backup		Deprecated pre-M11 endpoint variable accepted only as a migration fallback when BACKUP_S3_ENDPOINT remains at its local default.
BACKUP_COMMAND_TIMEOUT_SECONDS	backup	900	Positive per-command deadline in seconds for package install, database, archive, and S3 operations.
BACKUP_RESTORE_GLOBAL_TIMEOUT_SECONDS	backup	28800	Finite active-restore deadline; must exceed BACKUP_COMMAND_TIMEOUT_SECONDS. Hard-kill grace covers three bounded commands plus 60 seconds for compensation and exact lock cleanup.
BACKUP_MANIFEST_HMAC_KEY	backup		Required operator-supplied 64-hex HMAC-SHA-256 key for trusted backup manifests; preserve it outside the backup bucket for disaster recovery.
BACKUP_DEPLOYMENT_ID	backup		Required stable operator-supplied deployment identity bound into every trusted backup manifest.
BACKUP_MAX_POSTGRES_DUMP_BYTES	backup	10737418240	Finite maximum PostgreSQL dump size accepted by both backup production and bounded restore download.
BACKUP_RESTORE_MAX_CANDIDATES	backup	100	Maximum newest completion markers inspected by latest restore; use an exact BACKUP_TIMESTAMP for older backups.
BACKUP_DATABASES	backup	true	Capture Neo4j and Weaviate using their consistency-safe contracts. Run through services/backup/run-consistent-backup.sh so Neo4j is quiesced first.
BACKUP_DATABASE_QUIESCE_TIMEOUT_SECONDS	backup	120	Bounded host-side stop, start, and database snapshot status deadline.
BACKUP_MAX_DATABASE_ARCHIVE_BYTES	backup	53687091200	Maximum bytes accepted for each Neo4j or Weaviate snapshot archive during backup and restore.
BACKUP_LOCAL_SNAPSHOT_RETENTION_COUNT	backup	3	Completed local Neo4j/Weaviate native snapshot sets retained after authenticated publication (1-100); pruning runs only while the database services are quiesced.
BACKUP_LOCAL_ROLLBACK_RETENTION_COUNT	backup	1	Validated offline rollback volumes retained per database after a successful restore cutover (1-20).
BACKUP_SCALE	backup		0 (never long-running); the runner is invoked on demand via docker compose run.
BACKUP_IMAGE	backup	postgres:17.11-alpine@sha256:b0f9560a2de083e2cc7382e75f808c7381a32852a7ec49117deedb300e552b24	Digest-pinned build base for pg_dump/pg_restore; its major version must be >= the supabase-db server (17.x), or pg_dump aborts. The image bakes exact OpenSSL and the pgsty mc RELEASE.2026-09-16T00-00-00Z, copied from the digest-pinned pgsty/mc image. The entrypoint verifies its checksum and version for amd64/arm64.
BLENDER_MCP_SOURCE	blender-mcp	disabled	Host-installed Blender MCP bridge source. Disabled by default because Blender MCP can execute generated Python code inside Blender.
BLENDER_MCP_HOST	blender-mcp	localhost	Hostname written into the BLENDER_MCP_ENDPOINT hint for MCP clients. It does not change where the socket binds; BLENDER_MCP_BIND and BLENDER_MCP_ALLOW_REMOTE set the managed bind.
BLENDER_MCP_LOCALHOST_PORT	blender-mcp	9876	Host port for the Blender MCP add-on socket. This is a host-tool port hint, not a Kong or container port.
BLENDER_MCP_ENDPOINT	blender-mcp		Resolved endpoint hint for MCP clients when BLENDER_MCP_SOURCE is a host source (localhost / managed-localhost).
BLENDER_MCP_STATE_DIR	blender-mcp	~/.atlas/blender-mcp	State directory for the managed-localhost source: pinned add-on, generated headless launcher, pid and log files. ./start.sh blender-mcp remove deletes it.
BLENDER_MCP_INSTANCES	blender-mcp	1	Size of the managed-localhost pool, 1 to 16 (consumer manifest blender_mcp.instances). Instance 0 uses BLENDER_MCP_STATE_DIR and BLENDER_MCP_LOCALHOST_PORT; instance i uses <state dir>/instances/i and port BLENDER_MCP_LOCALHOST_PORT+i. Excess instances stop at the next start.
BLENDER_MCP_BIND	blender-mcp	127.0.0.1	Bind address for the managed headless bridge. Loopback-only by default — execute_code runs arbitrary Python inside Blender; any other value is refused unless BLENDER_MCP_ALLOW_REMOTE=true.
BLENDER_MCP_ALLOW_REMOTE	blender-mcp	false	Explicit second opt-in required to bind the managed bridge beyond loopback. Leave false unless you own the network boundary.
BLENDER_MCP_BLENDER_PATH	blender-mcp		Explicit Blender binary for the managed source. Empty = auto-detect (/Applications/Blender.app on macOS, blender on PATH elsewhere). Atlas manages the MCP bridge, not the Blender application — install Blender yourself.
BLENDER_MCP_ADDON_REF	blender-mcp	6641189231caf3752302ae20591bc87fda85fc4e	Pinned ahujasid/blender-mcp ref whose addon.py the managed source provisions. Move together with BLENDER_MCP_ADDON_SHA256.
BLENDER_MCP_ADDON_SHA256	blender-mcp	bba60831f5f89a74deda0294b131668a086cf46eb35a6a01abbd0d21d9e92630	sha256 of the pinned addon.py; the manager refuses to install a mismatching download (it will later execute arbitrary Python).
BLENDER_MCP_ADDON_FILE	blender-mcp		Escape hatch: use a local add-on file instead of the pinned download (no sha verification; preflight warns).
CELERY_SOURCE	celery	disabled	Deployment mode: container (backend worker + Flower) or disabled.
CELERY_WORKER_SCALE	celery		Set by ./start.sh from CELERY_SOURCE; do not edit. 1 runs celery-worker, 0 does not.
FLOWER_SCALE	celery		Set by ./start.sh from CELERY_SOURCE; do not edit. 1 runs the Flower monitoring UI, 0 does not.
FLOWER_PORT	celery	63071	Host port for Flower (in-container 5555).
CELERY_BROKER_URL	celery		Redis broker URL; Atlas uses Redis database 4 for Celery.
CELERY_RESULT_BACKEND	celery		Redis result backend URL; Atlas uses Redis database 4 for Celery task state.
CELERY_QUEUE	celery	atlas	Default Celery queue consumed by the Atlas backend worker.
CELERY_WORKER_CONCURRENCY	celery	2	Positive worker process concurrency for backend async jobs. Invalid values fail Backend and worker startup.
CELERY_WORKER_PREFETCH_MULTIPLIER	celery	1	Positive prefetch multiplier. The default reduces head-of-line blocking for long tasks; invalid values fail startup.
CELERY_TASK_TIME_LIMIT_SECONDS	celery	900	Positive hard per-task limit. Must exceed the soft limit and remain below the broker visibility timeout.
CELERY_TASK_SOFT_TIME_LIMIT_SECONDS	celery	840	Positive soft per-task limit. Must be strictly less than the hard limit.
CELERY_BROKER_VISIBILITY_TIMEOUT_SECONDS	celery	3600	Positive Redis broker visibility timeout, strictly greater than the task hard limit. The worker raises it to 300 s past max(rag_ingestion hard limit, global hard limit + 60) when larger, so running or delayed tasks are not re-delivered.
RAG_INGESTION_TASK_SOFT_TIME_LIMIT_SECONDS	celery		Positive soft limit for the rag_ingestion task alone; other tasks keep CELERY_TASK_SOFT_TIME_LIMIT_SECONDS. Empty means the larger of 3840 and that value. An ingestion whose graph targets' timeout_seconds sum to at least this is refused at submission.
RAG_INGESTION_TASK_TIME_LIMIT_SECONDS	celery		Positive hard limit for the rag_ingestion task alone; must exceed the soft limit. Empty (default) means the larger of 3900 and CELERY_TASK_TIME_LIMIT_SECONDS.
FLOWER_IMAGE	celery	mher/flower:2.0.1	Container image for flower.
CHATTERBOX_PORT	chatterbox	63059	Host port for the Chatterbox container (in-container listen port is 4123).
CHATTERBOX_LOCALHOST_PORT	chatterbox	63044	Host port for the chatterbox-localhost source variant. URL is derived at compose-render time as http://host.docker.internal:63044.
CHATTERBOX_SCALE	chatterbox		Computed by bootstrapper/services/service_config.py::_generate_tts_provider_config — set to 1 when TTS_PROVIDER_SOURCE=chatterbox-container-gpu.
CHATTERBOX_IMAGE	chatterbox	travisvn/chatterbox-tts-api:gpu@sha256:8c0b379172d48b49d59d8f29a35fa23c0dc27955dba4d3ab41a77b2d71b4711f	Rolling :gpu tag, pinned by digest. To refresh, run docker pull travisvn/chatterbox-tts-api:gpu and replace the digest.
CLOUD_OPENAI_SOURCE	cloud-providers	disabled	Options: enabled, disabled.
OPENAI_API_KEY	cloud-providers		OpenAI API key used by LiteLLM when CLOUD_OPENAI_SOURCE=enabled.
OPENAI_USER_MODELS	cloud-providers		Comma-separated OpenAI model names the wizard activated.
CLOUD_ANTHROPIC_SOURCE	cloud-providers	disabled	Options: enabled, disabled.
ANTHROPIC_API_KEY	cloud-providers		Anthropic API key used by LiteLLM when CLOUD_ANTHROPIC_SOURCE=enabled.
ANTHROPIC_USER_MODELS	cloud-providers		Comma-separated Anthropic model names the wizard activated.
CLOUD_OPENROUTER_SOURCE	cloud-providers	disabled	Options: enabled, disabled.
OPENROUTER_API_KEY	cloud-providers		OpenRouter API key used by LiteLLM when CLOUD_OPENROUTER_SOURCE=enabled.
OPENROUTER_USER_MODELS	cloud-providers		Comma-separated OpenRouter model names the wizard activated.
LITELLM_OPENAI_ENABLED	cloud-providers		Derived from CLOUD_OPENAI_SOURCE.
LITELLM_ANTHROPIC_ENABLED	cloud-providers		Derived from CLOUD_ANTHROPIC_SOURCE.
LITELLM_OPENROUTER_ENABLED	cloud-providers		Derived from CLOUD_OPENROUTER_SOURCE.
LITELLM_ENABLED_PROVIDERS	cloud-providers		Comma-separated lowercase list of enabled cloud providers.
CLOUDFLARED_SOURCE	cloudflared	disabled	Deployment mode: container (run the tunnel) or disabled (default).
CLOUDFLARE_TUNNEL_TOKEN	cloudflared		Cloudflare named-tunnel token (Zero Trust > Networks > Tunnels). When CLOUDFLARED_SOURCE=container this must be set or the daemon exits. The tunnel's public hostname is configured in the Cloudflare dashboard to point at http://kong-api-gateway:8000.
CLOUDFLARED_SCALE	cloudflared		Set to 1 when CLOUDFLARED_SOURCE=container, 0 when disabled.
CLOUDFLARED_IMAGE	cloudflared	cloudflare/cloudflared:2026.6.1	Cloudflare Tunnel daemon. Pin a dated tag; bump deliberately.
COMFYUI_SOURCE	comfyui	container-cpu	Deployment source for ComfyUI: container-cpu (default), container-gpu, localhost, managed-localhost-mps or disabled. A consumer manifest may set auto to pick by host capability.
COMFYUI_INIT_SOURCE	comfyui	container	Init-container selector listed in the SOURCE matrix (container or disabled). No code reads it: COMFYUI_INIT_SCALE follows COMFYUI_SOURCE.
COMFYUI_PORT	comfyui	63054	Host port that publishes the comfyui container (in-container port 18188).
COMFYUI_LOCALHOST_PORT	comfyui	8000	Host port for the localhost source variant. URL is derived at compose-render time as http://host.docker.internal:8000.
COMFYUI_MPS_LOCALHOST_PORT	comfyui	8188	Fixed host port the Atlas-managed Apple-Silicon/MPS ComfyUI listens on (source=managed-localhost-mps). Containers reach it as http://host.docker.internal:8188. Named with the _LOCALHOST_ host-port convention so the topology slot allocator leaves the fixed default intact.
COMFYUI_MPS_REF	comfyui	v0.27.0	Pinned upstream ComfyUI git ref the managed MPS host process checks out (mirrors COMFYUI_REF for the container path). A release tag or full commit SHA.
COMFYUI_MPS_STATE_DIR	comfyui	~/.atlas/comfyui-mps	Atlas-owned host state directory for the managed MPS ComfyUI: the pinned checkout, its venv, and the pid/log/status files. Never a container volume.
COMFYUI_MPS_MODELS_PATH	comfyui	~/Documents/ComfyUI/models	Host models directory the managed MPS process reuses via extra_model_paths, so weights are not duplicated. Shared with COMFYUI_LOCAL_MODELS_PATH by default. Under managed-localhost-mps, the host provisioner also downloads COMFYUI_USER_MODELS here at start.
COMFYUI_MPS_MIN_MEMORY_GB	comfyui	16	Minimum unified-memory headroom (GiB) the managed MPS preflight requires before launching ComfyUI; below this the preflight warns.
COMFYUI_MPS_TORCH_PIN	comfyui	torch==2.11.0 torchvision==0.26.0 torchaudio==2.11.0	Space-separated pip specifiers for Torch, torchvision and torchaudio that the managed MPS install applies. The pin keeps fresh installs reproducible against COMFYUI_MPS_REF; bump both together. macOS/arm64 Torch wheels include Metal/MPS.
COMFYUI_MPS_LISTEN	comfyui	127.0.0.1	Bind address of the managed MPS ComfyUI process. Keep 127.0.0.1 on Docker Desktop/macOS. On Linux engines, host.docker.internal resolves to a bridge address that cannot reach loopback, so set 0.0.0.0. Health probes use 127.0.0.1 for 0.0.0.0 or ::.
COMFYUI_BASE_URL	comfyui	http://comfyui:18188	In-container default. Runtime value is COMFYUI_ENDPOINT computed by hook.
COMFYUI_KONG_URL	comfyui	http://kong-api-gateway:8000/comfyui	Passed to the comfyui container as SUPABASE_URL for the inert Supabase output upload. Default http://kong-api-gateway:8000/comfyui.
COMFYUI_ARGS	comfyui	--listen	ComfyUI launch arguments passed to the AI-Dock image verbatim. The manifest default is --listen; compose falls back to --listen --cpu when the value is unset.
COMFYUI_AUTO_UPDATE	comfyui	true	AI-Dock startup update flag. Atlas keeps this true so COMFYUI_REF pins the upstream ComfyUI core even when the base image tag lags.
COMFYUI_REF	comfyui	v0.27.0	Pinned upstream ComfyUI git ref passed to AI-Dock. Use a release tag or full commit SHA when deliberately changing the ComfyUI core.
COMFYUI_PLATFORM	comfyui	linux/amd64	Compose platform for the comfyui image (default linux/amd64).
COMFYUI_UPLOAD_TO_SUPABASE	comfyui	true	Inert placeholder. Nothing in the ai-dock image, Atlas provisioning or the backend reads it, so ComfyUI does not upload outputs to Supabase. Kept for a future output-upload custom node.
COMFYUI_STORAGE_BUCKET	comfyui	comfyui-images	Reserved bucket name for the (currently inert) COMFYUI_UPLOAD_TO_SUPABASE placeholder; unused until an output-upload node is wired.
COMFYUI_MODELS_PATH	comfyui	/opt/ComfyUI/models	Not interpolated by any compose file: comfyui-init sets its own COMFYUI_MODELS_PATH=/models, and the comfyui container mounts models at /opt/ComfyUI/models.
COMFYUI_LOCAL_MODELS_PATH	comfyui	~/Documents/ComfyUI/models	Host models directory of an existing ComfyUI install. With COMFYUI_SOURCE=localhost, ./start.sh warns when it is missing. The comfyui container mounts it read-only at /host_models, which nothing reads.
COMFYUI_USER_MODELS	comfyui		Comma-separated catalog model names picked in the wizard or with --comfyui-models. Read by services/comfyui/init/scripts/download_models.sh, and by the managed-MPS host provisioner, which downloads the same set into COMFYUI_MPS_MODELS_PATH.
COMFYUI_CUSTOM_MODELS_FILE	comfyui	/custom-models.yaml	Path to the ComfyUI custom-models sidecar YAML (user-authored models not in the curated catalog). Read by comfyui_resolver when generating the download manifest, and by the wizard's model picker.
COMFYUI_CUSTOM_NODES_FILE	comfyui	/custom-nodes.yaml	Path list (os.pathsep-joined) of pinned custom-node allowlist files. The shipped services/comfyui/custom-nodes.yaml is always included; consumer manifests append custom_nodes.comfyui entries, which are always active. The bootstrapper maps selected models' requires_custom_node values through it into active-custom-nodes.tsv.
COMFYUI_INIT_SCALE	comfyui		Set by ./start.sh; do not edit. 1 runs the comfyui-init model download for container-* sources, 0 for localhost, managed-localhost-mps and disabled.
COMFYUI_SCALE	comfyui		Set by ./start.sh from COMFYUI_SOURCE; do not edit. 1 runs the comfyui container for container-cpu and container-gpu, 0 otherwise.
COMFYUI_ENDPOINT	comfyui		Resolved per COMFYUI_SOURCE. Read by backend, celery, open-webui, jupyterhub and downstream overlays; hermes receives it as COMFYUI_INTERNAL_URL. Consumer pattern: MY_URL: ${MY_URL:-${COMFYUI_ENDPOINT:-http://comfyui:18188}}
COMFYUI_DEPLOY_RESOURCES	comfyui	~	Derived by ./start.sh from the source's runtime deploy block ('~' when empty); do not set. No compose file reads it.
COMFYUI_MEMORY_LIMIT	comfyui	40g	Hard container-memory ceiling for ComfyUI (deploy.resources.limits.memory). The limit does not reserve memory; 40 g permits large multi-file bundles such as Krea 2 while smaller workloads consume only what they need.
COMFYUI_CPU_LIMIT	comfyui	2.0	CPU limit for the comfyui container (deploy.resources.limits.cpus). The 2.0 default stops CPU-mode inference from monopolising the host. GPU mode is GPU-bound, so the cap rarely matters.
COMFYUI_IMAGE	comfyui	ghcr.io/ai-dock/comfyui:v2-cpu-22.04-v0.2.7	AI-Dock base image; COMFYUI_AUTO_UPDATE + COMFYUI_REF pin the upstream ComfyUI core at startup.
COMFYUI_INIT_IMAGE	comfyui	alpine:3.24.2@sha256:294b683cb724975bec92580e1e685676bd4b50bda910ddb8c51d4cabeaec77e6	Container image for comfyui-init.
CRAWL4AI_SOURCE	crawl4ai	disabled	Deployment source for Crawl4AI: container or disabled (default).
CRAWL4AI_PORT	crawl4ai	63050	Host port that publishes the crawl4ai container API (in-container port 11235).
CRAWL4AI_API_TOKEN	crawl4ai		Bearer token required by Crawl4AI's Docker API; auto-generated by the bootstrapper.
CRAWL4AI_TIMEOUT_SECONDS	crawl4ai	30	Local Deep Researcher timeout for Crawl4AI extraction calls.
CRAWL4AI_MAX_CHARS	crawl4ai	60000	Maximum markdown characters Local Deep Researcher keeps from each Crawl4AI result.
CRAWL4AI_ALLOW_INTERNAL_URLS	crawl4ai	false	Keep false unless deliberately crawling private/internal URLs from inside the Atlas network.
CRAWL4AI_SCALE	crawl4ai		Set by ./start.sh from CRAWL4AI_SOURCE; do not edit. 1 runs crawl4ai, 0 does not.
CRAWL4AI_ENDPOINT	crawl4ai		Set by ./start.sh; do not edit. http://crawl4ai:11235 when CRAWL4AI_SOURCE=container, else empty. Read by Local Deep Researcher and n8n.
CRAWL4AI_IMAGE	crawl4ai	unclecode/crawl4ai:0.9.0	Container image for crawl4ai.
DOC_PROCESSOR_SOURCE	docling	disabled	Deployment source for the document processor: disabled (default), docling-localhost (existing host Docling) or docling-container-gpu (NVIDIA).
DOC_PROCESSOR_PORT	docling	63051	Host port that publishes the docling-gpu container API (in-container port 8000).
DOCLING_GPU_DEVICE	docling	cuda	Torch device for the docling-gpu container, passed as DOCLING_DEVICE (default cuda).
DOCLING_OUTPUT_FORMAT	docling	markdown	Default conversion output format when a request sets none: markdown (default), html, json or doctags.
DOCLING_USE_OCR	docling	auto	Default OCR mode when a request sets none: auto (default), always or never.
DOCLING_TABLE_MODE	docling	accurate	Default table-structure mode when a request sets none: accurate (default) or fast.
DOCLING_MAX_FILE_SIZE	docling	52428800	Positive maximum upload bytes for Docling GPU, localhost, and adapter APIs. Request bodies are capped before multipart parsing with 1 MiB framing overhead; invalid values fail startup and oversized requests return 413.
DOCLING_UPLOAD_TIMEOUT_SECONDS	docling	120	Total seconds (1-3600) allowed to receive one Docling or adapter upload body before a 408 response releases admission capacity.
DOCLING_CONCURRENCY	docling	1	Maximum concurrent Docling conversions per provider process. Default 1 prevents duplicate model loads and GPU memory contention.
DOCLING_API_TOKEN	docling		Auto-generated bearer credential required by Docling routes except /health.
DOCLING_AUTH_MODE	docling	required	Provider authentication mode: required (default) or disabled for an explicit emergency/local rollback.
DOCLING_CORS_ORIGINS	docling		Optional comma-separated browser origin allowlist. Empty disables CORS; wildcard is invalid while authentication is required.
DOCLING_INFERENCE_TIMEOUT_SECONDS	docling	900	Finite conversion and lazy-load deadline in seconds (1-3600). A timeout returns 504 and then exits the provider process for restart.
DOCLING_LOCALHOST_BIND_HOST	docling	127.0.0.1	Native Docling listen address. Set another interface explicitly only when remote access is intended.
DOCLING_ADAPTER_MAX_JOBS	docling	2	Maximum outstanding LightRAG adapter jobs, reserved before multipart parsing.
DOCLING_ADAPTER_RESULT_TTL_SECONDS	docling	900	Seconds a completed LightRAG adapter result may remain before verified artifact cleanup begins; the job slot stays occupied and cleanup is retried if deletion fails.
DOCLING_ADAPTER_MAX_RESULT_BYTES	docling	104857600	Maximum ZIP result bytes streamed from Docling to adapter storage before the job fails safely.
DOCLING_ADAPTER_UPSTREAM_MAX_ATTEMPTS	docling	3	Total bounded Docling request attempts when the provider returns HTTP 429.
DOCLING_ADAPTER_DOWNLOAD_TIMEOUT_SECONDS	docling	300	Positive maximum seconds for one result transmission; expiry starts verified artifact cleanup, and the adapter slot is released only after deletion succeeds.
DOCLING_ADAPTER_TMPFS_SIZE	docling	512m	Size of the adapter's bounded temporary-storage mount. It must cover MAX_JOBS times the larger of upload + 1 MiB multipart overhead or upload + result limit, plus at least 64 MiB staging headroom. Startup fails if capacity is smaller.
DOCLING_ADAPTER_SCALE	docling		Derived adapter replica count; one only when both LightRAG and Docling are enabled.
DOCLING_ADAPTER_UPSTREAM_ENDPOINT	docling		Authenticated internal bundle route selected for the adapter; never exposed to LightRAG.
DOCLING_ENABLE_FORMULAS	docling	true	Enables formula enrichment in the Docling pipeline: true (default) or false; other values fail.
DOCLING_ENABLE_CODE_BLOCKS	docling	true	Enables code-block enrichment in the Docling pipeline: true (default) or false; other values fail.
DOCLING_CHUNK_SIZE	docling	512	Default chunk size for chunked conversions. Must be a positive integer.
DOCLING_CHUNK_OVERLAP	docling	50	Default chunk overlap. Must be non-negative and at most half of DOCLING_CHUNK_SIZE; invalid values fail startup.
DOCLING_GPU_SCALE	docling		Set by ./start.sh from DOC_PROCESSOR_SOURCE; do not edit. 1 runs docling-gpu for docling-container-gpu, 0 otherwise.
DOCLING_ENDPOINT	docling		Set by ./start.sh from DOC_PROCESSOR_SOURCE; do not edit. Docling URL (empty when disabled) read by backend, celery, jupyterhub and n8n.
DOCLING_LOCALHOST_PORT	docling	18159	Host port for the localhost source variant. URL is derived at compose-render time as http://host.docker.internal:18159.
DOCLING_GPU_IMAGE	docling	pytorch/pytorch:2.13.0-cuda12.6-cudnn9-runtime@sha256:6acf597eeb8e376a96580dde4952f37cc017fef732bb40bfc73f28f25e3f64b4	Used as BASE_IMAGE in doc-processor/gpu/Dockerfile.
DOCLING_ADAPTER_IMAGE	docling	python:3.12.14-slim	Minimal base image for the isolated LightRAG compatibility adapter.
FAL_SOURCE	fal	disabled	Enables fal.ai as the backend provider for simple media generation routes.
FAL_API_KEY	fal		fal.ai API key. Required only when FAL_SOURCE=enabled. The backend also exposes it to fal-client as FAL_KEY.
FAL_MODEL	fal	fal-ai/flux/dev	Default fal.ai model endpoint used by the media gateway for text-to-image generation.
FAL_IMAGE_TO_3D_MODEL	fal	fal-ai/trellis	Default fal.ai endpoint id for the media gateway's image_to_3d modality. Must resolve to a verified registry entry (Hunyuan3D / TRELLIS / Tripo / Rodin); TRELLIS is the MIT-licensed default.
FAL_MODEL_LICENSE	fal	fal/provider-terms	License or terms marker returned in normalized media operation responses when provider-specific model licensing is not more specific.
FAL_TIMEOUT_SECONDS	fal	120	Finite Backend timeout budget for FAL media submit/poll operations and the compatibility route. Must be greater than zero and at most 3600 seconds.
FAL_OUTPUT_FORMAT	fal	jpeg	Requested output format for compatible fal.ai image models: jpeg or png.
FAL_ENABLE_SAFETY_CHECKER	fal	true	Whether to request the provider-side safety checker for compatible fal.ai image models.
PROJECT_NAME	globals	atlas	Docker Compose project name: the prefix for containers, volumes and networks, and the docker compose -p namespace. start.sh and stop.sh both read it. ./start.sh --project <name> (or -p) writes it here. Use a unique name for a submodule consumer.
BASE_PORT	globals	63000	Base port. Every service's *_PORT derives from this + a fixed offset.
HOST_BIND_IP	globals	127.0.0.1:	Published-port interface prefix (trailing colon). Default 127.0.0.1: is host-only; 0.0.0.0: is for deliberate remote access. Existing .env files with HOST_BIND_IP= are upgraded by backfill. Empty binds all interfaces except Ray, Zeppelin, mcp-servers, Supabase DB, Docling, Parakeet. LAN Supabase routes: KONG_SUPABASE_EXTRA_HOSTS.
HOST_GATEWAY_IP	globals		Set by ./start.sh for the container runtime: host-gateway on Docker, the literal gateway IP on Podman. If you write .env without ./start.sh, use host-gateway on Docker, or the real bridge IP on rootless setups.
BRAND_NAME	globals	Atlas	Wizard branding: the brand name shown by the wizard and --no-tui banner. Empty uses the Atlas default.
BRAND_TAGLINE	globals	A self-hosted, source-configurable, multi-disciplinary engineering platform — gen-AI, ML, and data.	Wizard branding: the tagline shown by the wizard and --no-tui banner. Empty uses the Atlas default.
BRAND_VERSION	globals	0.1.0	Wizard branding: the version shown by the wizard and --no-tui banner. Empty uses the Atlas default.
BRAND_AUTHOR	globals	Kaveh Razavi	Wizard branding: the author shown by the wizard and --no-tui banner. Empty uses the Atlas default.
BRAND_AUTHOR_EMAIL	globals	kaveh.razavi@gmail.com	Wizard branding: the author email shown by the wizard and --no-tui banner. Empty uses the Atlas default.
BRAND_LICENSE	globals	Apache License 2.0	Wizard branding: the license shown by the wizard and --no-tui banner. Empty uses the Atlas default.
BRAND_REPO_URL	globals	https://github.com/thekaveh/atlas	Wizard branding: the repository URL shown by the wizard and --no-tui banner. Empty uses the Atlas default.
BRAND_LOGO_FILE	globals		Optional path to a block-art logo that replaces the built-in ATLAS lockup in the wizard and --no-tui banner. Empty keeps ATLAS. Keep it 6 rows tall. File format: docs/quick-start/interactive-setup-wizard.md (Brand Customization).
COMPOSE_PROFILES	globals		Docker Compose profile activation list. Rebuilt from scratch on every ./start.sh run from the active STT/TTS/doc-processor sources (speaches-cpu/-gpu, parakeet-gpu, chatterbox-gpu, docling-gpu/doc-gpu). Do not edit by hand — the bootstrapper owns this value.
NVIDIA_VISIBLE_DEVICES	globals	all	Host-side GPU visibility. Forwarded into every NVIDIA-runtime container.
PROD_ENV_CPUS	globals	2	Reserved default container CPU cap (cores). No compose fragment reads it yet.
PROD_ENV_MEM_LIMIT	globals	8g	Reserved default container memory cap. No compose fragment reads it yet.
PROD_ENV_BACKEND_CPUS	globals	1	Reserved backend CPU cap (cores). No compose fragment reads it yet; the backend runs unbounded.
PROD_ENV_BACKEND_MEM_LIMIT	globals	2g	Reserved backend memory cap. No compose fragment reads it yet; the backend runs unbounded.
PROD_ENV_N8N_CPUS	globals	1	Reserved n8n CPU cap (cores). No compose fragment reads it yet; n8n runs unbounded.
PROD_ENV_N8N_MEM_LIMIT	globals	2g	Reserved n8n memory cap. No compose fragment reads it yet; n8n runs unbounded.
PROD_ENV_COMFYUI_CPUS	globals	2	CPU limit (cores) for ComfyUI with COMFYUI_SOURCE=container-gpu. The CPU variant uses COMFYUI_CPU_LIMIT instead.
PROD_ENV_COMFYUI_MEM_LIMIT	globals	4g	Memory limit for ComfyUI with COMFYUI_SOURCE=container-gpu. The CPU variant uses COMFYUI_MEMORY_LIMIT instead.
LOG_MAX_SIZE	globals	10m	Per-container json-file log max size (Docker logging option).
LOG_MAX_FILE	globals	3	Per-container json-file log file count (Docker logging option).
BOOTSTRAPPER_PORT_LAYOUT_VERSION	globals	5	Env-migration sentinel. A missing value or one below 5 runs the chained migrations up to v5. They cover port layout, URL to PORT, ComfyUI model set, stale Ollama model cleanup and Weaviate backup-filesystem. Fresh .env files are v5.
GRAFANA_SOURCE	grafana	disabled	Deployment source for Grafana: container or disabled.
GRAFANA_PORT	grafana	63009	Host port for Grafana's HTTP server (in-container 3000).
GRAFANA_ADMIN_USERNAME	grafana	admin	Grafana admin login. Override in .env if needed.
GRAFANA_ADMIN_PASSWORD	grafana		Auto-generated on first bootstrap and saved to .env. Grafana applies it only when it creates its database on an empty grafana-data volume. To rotate, run grafana cli admin reset-admin-password in the container, then update .env.
GRAFANA_ENDPOINT	grafana		Set by ./start.sh from GRAFANA_SOURCE; do not edit. http://grafana:3000 for the container source, empty when disabled.
GRAFANA_SCALE	grafana		Set by ./start.sh from GRAFANA_SOURCE; do not edit. 1 runs grafana, 0 does not.
GRAFANA_IMAGE	grafana	grafana/grafana:11.4.3	Container image for grafana.
HERMES_SOURCE	hermes	container	Deployment source for Hermes: container, localhost (an existing host Hermes) or disabled.
HERMES_API_PORT	hermes	63072	Host port for the Hermes Agent OpenAI-compatible API (in-container listen port is 8642).
HERMES_DASHBOARD_PORT	hermes	63073	Reserved compatibility allocation; no container dashboard host publish. Kong reaches container port 9119.
HERMES_MEMORY_LIMIT	hermes	4g	Memory limit of the hermes container.
HERMES_CPU_LIMIT	hermes	2.0	CPU limit (cores) of the hermes container.
HERMES_API_KEY	hermes		Auto-generated by bootstrapper. Hermes's API server bearer key, passed to Hermes as API_SERVER_KEY; LiteLLM's hermes-agent row reads the same value as HERMES_API_KEY.
HERMES_DEFAULT_MODEL	hermes		Empty = hermes-init picks from LiteLLM /v1/models priority list.
HERMES_CONTEXT_LENGTH	hermes	65536	Context length hermes-init writes into config.yaml for the default model. Hermes requires a model context window of at least 64K.
HERMES_DASHBOARD_ENABLED	hermes	true	Requested dashboard setting. Atlas forwards it only for container Hermes with configured container Kong. Empty retains the true default. Set false to remove the Chat tab.
HERMES_DASHBOARD_EFFECTIVE	hermes		Auto-managed dashboard flag from requested setting and launch sources. Container-facing HERMES_DASHBOARD defaults false before Atlas computes it. This does not check gateway health.
HERMES_DASHBOARD_TUI	hermes	1	Inert: Hermes v2026.6.19 always embeds the dashboard Chat tab (PTY terminal) and ignores this value; disable the dashboard (HERMES_DASHBOARD_ENABLED=false) to remove it.
HERMES_DASHBOARD_INSECURE	hermes	true	True permits unauthenticated backend-network access behind Kong. False requires an OAuth provider with a provisioned client ID (HERMES_DASHBOARD_OAUTH_CLIENT_ID or dashboard.oauth.client_id). Atlas passes through HERMES_DASHBOARD_OAUTH_CLIENT_ID but does not provision clients. False without a provider: dashboard refuses startup and restarts.
HERMES_DASHBOARD_OAUTH_CLIENT_ID	hermes		Provisioned upstream OAuth client ID, passed through unchanged. A nonempty value takes precedence over dashboard.oauth.client_id. Atlas does not provision clients.
HERMES_UID	hermes	10000	Container user — match a host UID to read hermes-data from outside.
HERMES_GID	hermes	10000	Container group ID; hermes-init chowns hermes-data to HERMES_UID:HERMES_GID.
HERMES_LOCALHOST_PORT	hermes	8642	Host port for the Hermes API in the localhost source variant. Consumed by HERMES_ENDPOINT (in-container clients call this). Defaults to Hermes' native API port.
HERMES_LOCALHOST_DASHBOARD_PORT	hermes	9119	Host port of the Hermes dashboard for HERMES_SOURCE=localhost; the hermes.localhost Kong route targets it. Default: the native dashboard port. The wizard edits only HERMES_LOCALHOST_PORT, so set this one in .env.
HERMES_INIT_SOURCE	hermes	container	Auto-managed by hermes. Container/disabled.
HERMES_INIT_SCALE	hermes		Set by ./start.sh from HERMES_SOURCE; do not edit. 1 runs hermes-init only for the container source, else 0.
HERMES_SCALE	hermes		Set by ./start.sh from HERMES_SOURCE; do not edit. 1 runs the hermes container only for the container source, else 0.
HERMES_ENDPOINT	hermes		Resolved per HERMES_SOURCE. Consumed by n8n, jupyterhub, LiteLLM model_list; also env-pre-wired for backend + openclaw (not yet exercised).
TTS_INTERNAL_URL	hermes		Set by ./start.sh from TTS_PROVIDER_SOURCE; do not edit. hermes-init renders the TTS block of config.yaml from it; empty omits the block.
STT_INTERNAL_URL	hermes		Set by ./start.sh from STT_PROVIDER_SOURCE; do not edit. hermes-init renders the STT block of config.yaml from it; empty omits the block.
STT_INTERNAL_API_KEY	hermes		Parakeet bearer token for the rendered server-side STT block; non-secret compatibility placeholder for other enabled STT engines.
COMFYUI_INTERNAL_URL	hermes		Set by ./start.sh from COMFYUI_ENDPOINT; do not edit. hermes-init renders the ComfyUI host skill from it; empty skips the skill.
SEARXNG_INTERNAL_URL	hermes		Set by ./start.sh from SEARXNG_SOURCE; do not edit. hermes-init renders the web-search block of config.yaml from it; empty omits the block.
LIGHTRAG_INTERNAL_URL	hermes		Set by ./start.sh from LIGHTRAG_ENDPOINT; do not edit. hermes-init receives it, but config.yaml has no LightRAG block; Hermes reaches LightRAG through LiteLLM.
HERMES_IMAGE	hermes	nousresearch/hermes-agent:v2026.6.19	Container image for hermes.
HERMES_INIT_IMAGE	hermes	alpine:3.24.2@sha256:294b683cb724975bec92580e1e685676bd4b50bda910ddb8c51d4cabeaec77e6	Container image for hermes-init.
ICEBERG_REST_SOURCE	iceberg-rest	disabled	Deployment source for the Iceberg REST catalog: container or disabled. The container source requires MINIO_SOURCE to be enabled.
ICEBERG_REST_PORT	iceberg-rest	63022	Host port for the Iceberg REST Catalog API (in-container 8181).
ICEBERG_DB_USER	iceberg-rest	iceberg	Scoped Postgres login for the catalog JDBC store (database iceberg). Supabase db-init (05-scoped-roles.sh) creates it.
ICEBERG_DB_PASSWORD	iceberg-rest	atlas-db-password	Auto-generated by bootstrapper. Postgres role for the Iceberg JDBC catalog DB.
ICEBERG_REST_POSTGRES_JDBC_VERSION	iceberg-rest	42.7.12	PostgreSQL JDBC driver version layered into the local Iceberg REST image build.
ICEBERG_REST_POSTGRES_JDBC_SHA512	iceberg-rest	3759e7160591863e5100361298943df94488a6e4ee03936d20723638142fe8038c9a6a47fa8ee7e424f4ef09bb351edc89dfe2ae4acdfc0f92699a8b00196c5c	SHA-512 digest required for the PostgreSQL JDBC driver selected above. Update both values together.
ICEBERG_REST_SCALE	iceberg-rest		Set by ./start.sh from ICEBERG_REST_SOURCE; do not edit. 1 runs iceberg-rest, 0 does not.
ICEBERG_REST_INIT_SCALE	iceberg-rest		Set by ./start.sh from ICEBERG_REST_SOURCE; do not edit. 1 runs iceberg-rest-init, 0 does not.
ICEBERG_REST_IMAGE	iceberg-rest	apache/iceberg-rest-fixture:1.10.1	Base image for the local Atlas build. Atlas layers the PostgreSQL JDBC driver onto the Apache fixture image.
ICEBERG_REST_INIT_IMAGE	iceberg-rest	postgres:15.19-alpine	Container image for iceberg-rest-init.
JENKINS_SOURCE	jenkins	disabled	Deployment source for Jenkins: container or disabled. The container source requires MINIO_SOURCE to be enabled.
JENKINS_PORT	jenkins	63090	Host port for the Jenkins Web UI (in-container 8080).
JENKINS_SCALE	jenkins		Set by ./start.sh from JENKINS_SOURCE; do not edit. 1 runs the Jenkins controller, 0 does not.
JENKINS_ADMIN_USER	jenkins	admin	Jenkins admin login created by JCasC.
JENKINS_ADMIN_PASSWORD	jenkins		Auto-generated on first bootstrap by KeyGenerator.generate_jenkins_admin_password().
JENKINS_MEMORY_LIMIT	jenkins	3g	Container memory limit for Jenkins controller and local Maven builds.
JENKINS_CPU_LIMIT	jenkins	2.0	Container CPU limit for Jenkins controller and local Maven builds.
JENKINS_IMAGE	jenkins	jenkins/jenkins:lts-jdk21@sha256:c1e4c349365f6d16d88595b2c5f7e8ff39b8ae1d061f62420bac193b4b9616d0	Jenkins LTS with JDK 21; Atlas builds a local derived image with Maven, MinIO mc, and plugins. The default pins lts-jdk21 to a reviewed digest, because the bare tag moves on each new LTS line. To refresh, run docker pull jenkins/jenkins:lts-jdk21 and set the new digest.
JUPYTER_DB_USER	jupyterhub	atlas_jupyter	Read-only PostgreSQL login exposed to curated notebooks.
JUPYTER_DB_PASSWORD	jupyterhub	atlas-db-password	Auto-generated password for JupyterHub's read-only database role.
JUPYTER_DB_USER_URI	jupyterhub	atlas_jupyter	Bootstrapper-synchronized percent-encoded JUPYTER_DB_USER for DSN userinfo.
JUPYTER_DB_PASSWORD_URI	jupyterhub	atlas-db-password	Bootstrapper-synchronized percent-encoded JUPYTER_DB_PASSWORD for DSN userinfo.
JUPYTERHUB_SOURCE	jupyterhub	container	Deployment source for the JupyterLab notebook server: container or disabled.
JUPYTERHUB_PORT	jupyterhub	63094	Host port for the JupyterLab server (in-container 8888).
JUPYTERHUB_TOKEN	jupyterhub		Optional. Empty = generate on start.
JUPYTER_ALLOW_ORIGIN	jupyterhub		Allowed browser origin for Jupyter Server websocket upgrades. Empty keeps Jupyter's same-origin check; token clients such as VS Code do not need '*'. Set one exact origin only for a browser page on another origin.
SPARK_REMOTE	jupyterhub	sc://spark-connect:15002	Spark Connect endpoint used by bundled notebooks; override for a remote or managed Spark Connect server.
MCP_SERVERS_URL	jupyterhub		Streamable HTTP URL for the Curated MCP Servers endpoint. The bootstrapper sets http://mcp-servers:8000/mcp only for MCP_SERVERS_SOURCE=container and clears it when disabled.
JUPYTERHUB_SCALE	jupyterhub		Set by ./start.sh from JUPYTERHUB_SOURCE; do not edit. 0 when disabled, else 1.
JUPYTERHUB_IMAGE	jupyterhub	quay.io/jupyter/datascience-notebook:2026-08-24@sha256:e5029672ab8a861345f117dc466a4dab91ad7c299f3c7c853b9b193860b16aaf	Container image for jupyterhub.
KONG_API_GATEWAY_SOURCE	kong	container	Single-option (container only).
KONG_HTTP_PORT	kong	63000	Host port for the Kong HTTP proxy (in-container 8000). Every *.localhost route uses it.
KONG_HTTPS_PORT	kong	63001	Host port for the Kong HTTPS proxy (in-container 8443).
KONG_CORS_EXTRA_ORIGINS	kong		Comma-separated exact browser origins (scheme://host[:port]) Kong allows cross-origin, beyond *.localhost, localhost and 127.0.0.1. Values are normalised; wildcards and paths are ignored with a warning. Read at Kong config generation.
KONG_SUPABASE_EXTRA_HOSTS	kong		Comma-separated extra lowercase Host names (bare hostname or IPv4) the Supabase API routes answer on, beyond the local defaults. Use for a tunnel, LAN address or custom API_EXTERNAL_URL host. Other Hosts get 404. Read at Kong config generation.
DASHBOARD_USERNAME	kong	kong_admin	Username of Kong's Basic-auth dashboard consumer, which gates Supabase Studio, /pg/ and other dashboard routes such as hermes.localhost.
DASHBOARD_PASSWORD	kong	kong_password	Auto-generated on first ./start.sh when left at the placeholder (rotate-when-absent, like MINIO_ROOT_PASSWORD). Gates the Kong-routed Studio dashboard via basic-auth.
KONG_API_GATEWAY_IMAGE	kong	kong:3.9.3	Container image for kong-api-gateway.
LABEL_STUDIO_SOURCE	label-studio	disabled	Deployment source for Label Studio: container or disabled. The container source requires MINIO_SOURCE to be enabled.
LABEL_STUDIO_PORT	label-studio	63091	Host port for the Label Studio UI/API (in-container 8080).
LABEL_STUDIO_ENDPOINT	label-studio		In-network URL for Atlas services and notebooks.
LABEL_STUDIO_API_URL	label-studio		Notebook/backend API URL; empty when Label Studio is disabled.
LABEL_STUDIO_SCALE	label-studio		Set by ./start.sh from LABEL_STUDIO_SOURCE; do not edit. 1 runs label-studio, 0 does not.
LABEL_STUDIO_INIT_SCALE	label-studio		Set by ./start.sh from LABEL_STUDIO_SOURCE; do not edit. 1 runs label-studio-init, 0 does not.
LABEL_STUDIO_DB_NAME	label-studio	label_studio	Postgres database for Label Studio. Supabase db-init (05-scoped-roles.sh) creates it.
LABEL_STUDIO_DB_NAME_URI	label-studio	label_studio	Bootstrapper-synchronized percent-encoded LABEL_STUDIO_DB_NAME for DSN path components.
LABEL_STUDIO_DB_USER	label-studio	label_studio	Scoped Postgres login that owns LABEL_STUDIO_DB_NAME. Supabase db-init (05-scoped-roles.sh) creates it.
LABEL_STUDIO_DB_PASSWORD	label-studio	atlas-db-password	Auto-generated Postgres password for the dedicated Label Studio database role.
LABEL_STUDIO_USERNAME	label-studio	admin@atlas.local	Initial Label Studio admin email passed to upstream USERNAME.
LABEL_STUDIO_PASSWORD	label-studio		Auto-generated initial Label Studio admin password.
LABEL_STUDIO_USER_TOKEN	label-studio		Auto-generated legacy API token for notebook/API smoke paths.
LABEL_STUDIO_SECRET_KEY	label-studio		Django SECRET_KEY used by the Label Studio container.
LABEL_STUDIO_IMAGE	label-studio	heartexlabs/label-studio:1.23.0	Container image for label-studio.
LABEL_STUDIO_INIT_IMAGE	label-studio	alpine:3.24.2	Container image for label-studio-init.
LANGFUSE_SOURCE	langfuse	disabled	Deployment source for Langfuse (web, worker and ClickHouse): container or disabled. The container source requires MINIO_SOURCE to be enabled.
LANGFUSE_PORT	langfuse	63005	Host port for the Langfuse web UI/API (in-container 3000).
LANGFUSE_ENDPOINT	langfuse		In-network URL used by LiteLLM when Langfuse tracing is enabled.
LANGFUSE_WEB_SCALE	langfuse		Set by ./start.sh from LANGFUSE_SOURCE; do not edit. 1 runs langfuse-web, 0 does not.
LANGFUSE_WORKER_SCALE	langfuse		Set by ./start.sh from LANGFUSE_SOURCE; do not edit. 1 runs langfuse-worker, 0 does not.
LANGFUSE_CLICKHOUSE_SCALE	langfuse		Set by ./start.sh from LANGFUSE_SOURCE; do not edit. 1 runs langfuse-clickhouse, 0 does not.
LANGFUSE_INIT_SCALE	langfuse		Set by ./start.sh from LANGFUSE_SOURCE; do not edit. 1 runs langfuse-init, 0 does not.
LANGFUSE_DB_NAME	langfuse	langfuse	Postgres database for Langfuse. Supabase db-init (05-scoped-roles.sh) creates it.
LANGFUSE_DB_NAME_URI	langfuse	langfuse	Bootstrapper-synchronized percent-encoded LANGFUSE_DB_NAME for DSN path components.
LANGFUSE_DB_USER	langfuse	langfuse	Scoped Postgres login that owns LANGFUSE_DB_NAME. Supabase db-init (05-scoped-roles.sh) creates it.
LANGFUSE_DB_PASSWORD	langfuse	atlas-db-password	Auto-generated password for the owner of Langfuse's dedicated database.
LANGFUSE_DB_USER_URI	langfuse	langfuse	Bootstrapper-synchronized percent-encoded LANGFUSE_DB_USER for DSN userinfo.
LANGFUSE_DB_PASSWORD_URI	langfuse	atlas-db-password	Bootstrapper-synchronized percent-encoded LANGFUSE_DB_PASSWORD for DSN userinfo.
LANGFUSE_SALT	langfuse		Auto-generated by the bootstrapper on first run.
LANGFUSE_ENCRYPTION_KEY	langfuse		64 hex chars. Auto-generated by the bootstrapper on first run.
LANGFUSE_NEXTAUTH_SECRET	langfuse		Auto-generated by the bootstrapper on first run.
LANGFUSE_PUBLIC_KEY	langfuse		Initial Langfuse project public key used by LiteLLM tracing.
LANGFUSE_SECRET_KEY	langfuse		Initial Langfuse project secret key used by LiteLLM tracing.
LANGFUSE_INIT_ORG_ID	langfuse	atlas	Organisation ID that langfuse-web headless init creates on first start.
LANGFUSE_INIT_ORG_NAME	langfuse	Atlas	Organisation name that langfuse-web headless init creates on first start.
LANGFUSE_INIT_PROJECT_ID	langfuse	atlas-default	Project ID that langfuse-web headless init creates on first start, with LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY as its API keys.
LANGFUSE_INIT_PROJECT_NAME	langfuse	Atlas Default	Project name that langfuse-web headless init creates on first start.
LANGFUSE_INIT_USER_EMAIL	langfuse	admin@atlas.localhost	Login email of the initial Langfuse user that headless init creates on first start.
LANGFUSE_INIT_USER_NAME	langfuse	Atlas Admin	Display name of the initial Langfuse user that headless init creates on first start.
LANGFUSE_INIT_USER_PASSWORD	langfuse		Auto-generated by the bootstrapper on first run.
LANGFUSE_CLICKHOUSE_USER	langfuse	clickhouse	ClickHouse user for langfuse-clickhouse, also used by langfuse-web and langfuse-worker.
LANGFUSE_CLICKHOUSE_PASSWORD	langfuse		Auto-generated by the bootstrapper on first run.
LANGFUSE_IMAGE	langfuse	langfuse/langfuse:3.115.0	Container image for langfuse-web.
LANGFUSE_WORKER_IMAGE	langfuse	langfuse/langfuse-worker:3.115.0	Container image for langfuse-worker.
LANGFUSE_CLICKHOUSE_IMAGE	langfuse	clickhouse/clickhouse-server:25.8.33.6	Container image for langfuse-clickhouse.
LANGFUSE_INIT_IMAGE	langfuse	alpine:3.24.2	Container image for langfuse-init.
LIGHTRAG_DB_USER	lightrag	atlas_lightrag	Scoped PostgreSQL owner for LightRAG PGVectorStorage relations.
LIGHTRAG_DB_PASSWORD	lightrag	atlas-db-password	Auto-generated password for the scoped LightRAG database role.
LIGHTRAG_DB_USER_URI	lightrag	atlas_lightrag	Bootstrapper-synchronized percent-encoded LIGHTRAG_DB_USER for DSN userinfo.
LIGHTRAG_DB_PASSWORD_URI	lightrag	atlas-db-password	Bootstrapper-synchronized percent-encoded LIGHTRAG_DB_PASSWORD for DSN userinfo.
LIGHTRAG_SOURCE	lightrag	disabled	Deployment source for LightRAG: container, localhost (an existing host LightRAG) or disabled.
LIGHTRAG_API_PORT	lightrag	63074	Host port for LightRAG (in-container 9621).
LIGHTRAG_LOCALHOST_PORT	lightrag	63068	Host port for the localhost source variant.
LIGHTRAG_API_KEY	lightrag		Auto-generated bearer key. Forwarded to LiteLLM as the model api_key.
LIGHTRAG_TOKEN_SECRET	lightrag		Auto-generated JWT signing key for LightRAG's /login endpoint. Without this, LightRAG falls back to a hardcoded default key (security risk).
LIGHTRAG_WORKERS	lightrag	2	Passed as WORKERS but inert: the image starts lightrag.api.lightrag_server under a single uvicorn process (WORKERS is read only by lightrag-gunicorn).
LIGHTRAG_MEMORY_LIMIT	lightrag	6g	Memory limit of the lightrag container.
LIGHTRAG_CPU_LIMIT	lightrag	2.0	CPU limit (cores) of the lightrag container.
LIGHTRAG_LLM_BINDING	lightrag	openai	Base LightRAG LLM binding, through LiteLLM at LIGHTRAG_LLM_BINDING_HOST by default. A role without its own binding, host or key inherits it. Exception: with a direct-provider base, a role whose model declares catalog request_defaults stays on LiteLLM.
LIGHTRAG_LLM_BINDING_HOST	lightrag	http://litellm:4000/v1	Base LightRAG LLM endpoint (LLM_BINDING_HOST). Default: the in-network LiteLLM.
LIGHTRAG_LLM_MODEL	lightrag		Empty = lightrag-init resolves from LITELLM_DEFAULT_MODEL.
LIGHTRAG_EXTRACT_LLM_MODEL	lightrag		Optional LightRAG EXTRACT role model. Empty = inherit LLM_MODEL.
LIGHTRAG_KEYWORD_LLM_MODEL	lightrag		Optional LightRAG KEYWORD role model. Empty = inherit LLM_MODEL.
LIGHTRAG_QUERY_LLM_MODEL	lightrag		Optional LightRAG QUERY role model. Empty = inherit LLM_MODEL.
LIGHTRAG_EXTRACT_LLM_BINDING	lightrag		Optional EXTRACT role provider binding. Empty = inherit LLM_BINDING.
LIGHTRAG_KEYWORD_LLM_BINDING	lightrag		Optional KEYWORD role provider binding. Empty = inherit LLM_BINDING.
LIGHTRAG_QUERY_LLM_BINDING	lightrag		Optional QUERY role provider binding. Empty = inherit LLM_BINDING.
LIGHTRAG_EXTRACT_LLM_BINDING_HOST	lightrag		Optional EXTRACT role provider endpoint. Empty = inherit LLM_BINDING_HOST.
LIGHTRAG_KEYWORD_LLM_BINDING_HOST	lightrag		Optional KEYWORD role provider endpoint. Empty = inherit LLM_BINDING_HOST.
LIGHTRAG_QUERY_LLM_BINDING_HOST	lightrag		Optional QUERY role provider endpoint. Empty = inherit LLM_BINDING_HOST.
LIGHTRAG_EXTRACT_LLM_BINDING_API_KEY	lightrag		Optional EXTRACT role API key. Unset means the LiteLLM master key, but only if the role's effective host is the in-network LiteLLM. A role with its own binding or host elsewhere must set it, or the container exits.
LIGHTRAG_KEYWORD_LLM_BINDING_API_KEY	lightrag		Optional KEYWORD role API key. Unset means the LiteLLM master key, but only if the role's effective host is the in-network LiteLLM. A role with its own binding or host elsewhere must set it, or the container exits.
LIGHTRAG_QUERY_LLM_BINDING_API_KEY	lightrag		Optional QUERY role API key. Unset means the LiteLLM master key, but only if the role's effective host is the in-network LiteLLM. A role with its own binding or host elsewhere must set it, or the container exits.
LIGHTRAG_EXTRACT_OLLAMA_LLM_NUM_PREDICT	lightrag	4096	EXTRACT output cap in tokens when EXTRACT is bound to native Ollama; other bindings ignore it. Without it Ollama output is unbounded. Must be an integer. Cached extractions keep any truncation, so clear the LightRAG LLM cache after raising it.
LIGHTRAG_EXTRACT_OLLAMA_LLM_NUM_CTX	lightrag	16384	EXTRACT context window in tokens when EXTRACT is bound to native Ollama; other bindings ignore it. The default fits the extraction prompt, gleaning pass, merge summaries and output cap. Overrides Ollama's VRAM-based default and OLLAMA_CONTEXT_LENGTH. Must be an integer.
LIGHTRAG_EXTRACT_MAX_ASYNC_LLM	lightrag		Optional EXTRACT role LLM concurrency. Empty = inherit MAX_ASYNC_LLM.
LIGHTRAG_KEYWORD_MAX_ASYNC_LLM	lightrag		Optional KEYWORD role LLM concurrency. Empty = inherit MAX_ASYNC_LLM.
LIGHTRAG_QUERY_MAX_ASYNC_LLM	lightrag		Optional QUERY role LLM concurrency. Empty = inherit MAX_ASYNC_LLM.
LIGHTRAG_EXTRACT_LLM_TIMEOUT	lightrag		Optional EXTRACT role LLM timeout in seconds. Empty = inherit LLM_TIMEOUT.
LIGHTRAG_KEYWORD_LLM_TIMEOUT	lightrag		Optional KEYWORD role LLM timeout in seconds. Empty = inherit LLM_TIMEOUT.
LIGHTRAG_QUERY_LLM_TIMEOUT	lightrag		Optional QUERY role LLM timeout in seconds. Empty = inherit LLM_TIMEOUT.
LIGHTRAG_QUERY_ENABLE_RERANK	lightrag	false	LightRAG query default for reranking. Atlas keeps this false by default because direct LightRAG→TEI rerank payloads are incompatible without an adapter.
LIGHTRAG_QUERY_TOP_K	lightrag	10	LightRAG query TOP_K. Atlas defaults this to a concrete integer because LightRAG v1.5.4 crashes when this env var is present but empty.
LIGHTRAG_QUERY_CHUNK_TOP_K	lightrag	5	LightRAG query CHUNK_TOP_K. Atlas defaults this to a concrete integer because LightRAG v1.5.4 crashes when this env var is present but empty.
LIGHTRAG_QUERY_MAX_TOTAL_TOKENS	lightrag	12000	LightRAG query MAX_TOTAL_TOKENS. Atlas defaults this to a concrete integer because LightRAG v1.5.4 crashes when this env var is present but empty.
LIGHTRAG_EMBEDDING_BINDING	lightrag	openai	LightRAG embedding binding (EMBEDDING_BINDING). Default openai, through LiteLLM.
LIGHTRAG_EMBEDDING_BINDING_HOST	lightrag	http://litellm:4000/v1	LightRAG embedding endpoint (EMBEDDING_BINDING_HOST). Default: the in-network LiteLLM.
LIGHTRAG_EMBEDDING_MODEL	lightrag		Empty = lightrag-init resolves from LITELLM_EMBEDDING_MODEL.
LIGHTRAG_EMBEDDING_DIM	lightrag		Empty = auto (lightrag-init probes the embedding model's true dimension). Set ONLY to force a value — a wrong dim writes a mismatched PGVector index and every insert fails.
LIGHTRAG_VLM_PROCESS_ENABLE	lightrag	true	Passed as VLM_PROCESS_ENABLE: process images and figures with a vision LLM during document parsing.
LIGHTRAG_KV_STORAGE	lightrag	RedisKVStorage	LightRAG KV storage class. Default RedisKVStorage needs Redis; set a local class such as JsonKVStorage to run without it.
LIGHTRAG_VECTOR_STORAGE	lightrag	PGVectorStorage	LightRAG vector storage class. Default PGVectorStorage needs Supabase; set a local class such as NanoVectorDBStorage to run without it.
LIGHTRAG_GRAPH_STORAGE	lightrag	Neo4JStorage	LightRAG graph storage class. Default Neo4JStorage needs Neo4j; set a local class such as NetworkXStorage to run without it.
LIGHTRAG_DOC_STATUS_STORAGE	lightrag	RedisDocStatusStorage	LightRAG document-status storage class. Default RedisDocStatusStorage needs Redis; set a local class such as JsonDocStatusStorage to run without it.
LIGHTRAG_SCALE	lightrag		Set by ./start.sh from LIGHTRAG_SOURCE; do not edit. 1 runs the lightrag container only for the container source, else 0.
LIGHTRAG_INIT_SCALE	lightrag		Set by ./start.sh from LIGHTRAG_SOURCE; do not edit. 1 runs lightrag-init only for the container source, else 0.
LIGHTRAG_ENDPOINT	lightrag		Consumed by hermes/n8n/backend runtime_adaptive and LiteLLM model_list.
LIGHTRAG_RERANK_BINDING_HOST	lightrag		Set by ./start.sh; do not edit. The backend rerank adapter URL when LIGHTRAG_RERANK_ADAPTER_ENABLED=true and TEI is enabled; empty otherwise.
LIGHTRAG_RERANK_BINDING	lightrag		Always null by default. Direct LightRAG→TEI rerank payloads are incompatible without an adapter, and LightRAG crashes on an empty value.
LIGHTRAG_DOCLING_ENDPOINT	lightrag		Set by ./start.sh; do not edit. The docling-lightrag-adapter URL when LightRAG runs as a container and the adapter runs; empty otherwise.
LIGHTRAG_PG_URI	lightrag		Set by ./start.sh; do not edit. PostgreSQL DSN for LightRAG's scoped role on supabase-db; empty when Supabase is disabled.
LIGHTRAG_NEO4J_URI	lightrag		Set by ./start.sh; do not edit. Source-aware Neo4j Bolt URI; empty when Neo4j is disabled.
LIGHTRAG_NEO4J_USERNAME	lightrag		Set by ./start.sh; do not edit. Neo4j user for LightRAG graph storage; empty when Neo4j is disabled.
LIGHTRAG_NEO4J_PASSWORD	lightrag		Set by ./start.sh from GRAPH_DB_PASSWORD; do not edit. Empty when Neo4j is disabled.
LIGHTRAG_REDIS_URI	lightrag		Set by ./start.sh; do not edit. Redis URI (db 2) for LightRAG KV and document-status storage; empty when Redis is disabled.
LIGHTRAG_IMAGE	lightrag	ghcr.io/hkuds/lightrag:v1.5.4	Container image for lightrag.
LIGHTRAG_INIT_IMAGE	lightrag	alpine:3.24.2@sha256:294b683cb724975bec92580e1e685676bd4b50bda910ddb8c51d4cabeaec77e6	Container image for lightrag-init.
LITELLM_SOURCE	litellm	container	Locked. LiteLLM is mandatory; only one source value supported.
LITELLM_PORT	litellm	63040	Host port for the LiteLLM gateway API and UI (in-container 4000).
LITELLM_INIT_SCALE	litellm	1	Init container always runs (single value).
LITELLM_MASTER_KEY	litellm		Auto-generated by bootstrapper on first run. Doubles as the admin-dashboard password.
LITELLM_UI_USERNAME	litellm	admin	Admin-dashboard login username at http://litellm.localhost:${KONG_HTTP_PORT}/ui/. Password is LITELLM_MASTER_KEY.
LITELLM_CACHE_TTL	litellm	3600	Lifetime in seconds of a cached LiteLLM response in Redis. Embeddings, reranks and transcriptions are cached, and chat with LITELLM_CACHE_CHAT=true. LiteLLM's own default is 60s. A non-numeric or non-positive value falls back to 3600.
LITELLM_CACHE_NAMESPACE	litellm	litellm.cache	Key prefix for LiteLLM's Redis response cache. Redis db 0 also holds the n8n and Langfuse queues and the backend media store. The prefix keeps gateway cache keys separately scannable and droppable.
LITELLM_CACHE_CHAT	litellm	false	Also cache chat, text completion and Responses API calls. Off by default, because a cached reply silently replaces a fresh hermes-agent or lightrag passthrough run, such as a query over newly ingested documents.
LITELLM_DB_NAME	litellm	litellm	Postgres database for LiteLLM. Supabase db-init (05-scoped-roles.sh) creates it.
LITELLM_DB_NAME_URI	litellm	litellm	Bootstrapper-synchronized percent-encoded LITELLM_DB_NAME for DSN path components.
LITELLM_DB_USER	litellm	litellm	Scoped Postgres login that owns LITELLM_DB_NAME. Supabase db-init (05-scoped-roles.sh) creates it.
LITELLM_DB_PASSWORD	litellm	atlas-db-password	Auto-generated password for the owner of LiteLLM's dedicated database.
LITELLM_DB_USER_URI	litellm	litellm	Bootstrapper-synchronized percent-encoded LITELLM_DB_USER for DSN userinfo.
LITELLM_DB_PASSWORD_URI	litellm	atlas-db-password	Bootstrapper-synchronized percent-encoded LITELLM_DB_PASSWORD for DSN userinfo.
LITELLM_EMBEDDING_MODEL	litellm	ollama/nomic-embed-text	Default embedding model used by weaviate-init when computing collection vectorizers.
LITELLM_DEFAULT_MODEL	litellm	ollama/qwen3.8:latest	Optional default chat model passed to LangMem (backend).
LITELLM_VISION_MODEL	litellm	ollama/qwen3.8:latest	Default vision-capable model (resolved from the model catalog; forward-looking — no consumer routes on it yet).
LITELLM_BASE_URL	litellm		In-network URL consumers point at (backend, n8n, jupyterhub, open-webui).
LITELLM_IMAGE	litellm	ghcr.io/berriai/litellm:v1.83.14-stable.patch.2	Container image for litellm.
LLM_GRAPH_BUILDER_SOURCE	llm-graph-builder	disabled	Deployment source for Neo4j LLM Graph Builder: container (pinned source build) or disabled. The container source requires NEO4J_GRAPH_DB_SOURCE=container.
LLM_GRAPH_BUILDER_REF	llm-graph-builder	4a412f4688cf4096976045c019edc0a7f6ddcb6b	Pinned upstream neo4j-labs/llm-graph-builder git ref used by the source-build containers.
LLM_GRAPH_BUILDER_PORT	llm-graph-builder	63095	Host port for the Graph Builder frontend (in-container 8080).
LLM_GRAPH_BUILDER_MODEL_ID	llm-graph-builder	atlas_litellm	Model id shown in the Graph Builder UI. Keep atlas_litellm: compose only exports LLM_MODEL_CONFIG_ATLAS_LITELLM, and Graph Builder reads LLM_MODEL_CONFIG_<ID>, so another id breaks extraction and chat.
LLM_GRAPH_BUILDER_LLM_MODEL	llm-graph-builder		LiteLLM model alias used behind atlas_litellm. Empty = use LITELLM_DEFAULT_MODEL, then gpt-4o-mini (needs an OpenAI key).
LLM_GRAPH_BUILDER_NEO4J_DATABASE	llm-graph-builder	neo4j	Neo4j database name for Graph Builder data. Use a dedicated database on editions that support it.
LLM_GRAPH_BUILDER_DIFFBOT_API_KEY	llm-graph-builder		Optional Diffbot API key for upstream Diffbot extraction/model features. Atlas' default LiteLLM path does not require it.
LLM_GRAPH_BUILDER_GCP_LOG_METRICS_ENABLED	llm-graph-builder	false	Enable the upstream Google Cloud logging integration. Requires project id and an ADC credential file.
LLM_GRAPH_BUILDER_GCS_FILE_CACHE	llm-graph-builder	false	Store uploaded and failed files in GCS instead of local container storage. Requires the full GCP and bucket configuration.
LLM_GRAPH_BUILDER_GCP_PROJECT_ID	llm-graph-builder		Google Cloud project id used by optional GCS caching and Cloud logging.
LLM_GRAPH_BUILDER_GCS_UPLOAD_BUCKET	llm-graph-builder		GCS bucket for uploaded files when LLM_GRAPH_BUILDER_GCS_FILE_CACHE=true.
LLM_GRAPH_BUILDER_GCS_FAILED_BUCKET	llm-graph-builder		GCS bucket for failed files when LLM_GRAPH_BUILDER_GCS_FILE_CACHE=true.
LLM_GRAPH_BUILDER_GCP_CREDENTIALS_FILE	llm-graph-builder		Absolute host path to a Google ADC JSON file. Mounted read-only only for the Graph Builder backend.
LLM_GRAPH_BUILDER_REACT_APP_SOURCES	llm-graph-builder	local,wiki,web	Comma-separated upstream source picker values compiled into the frontend.
LLM_GRAPH_BUILDER_CHAT_MODES	llm-graph-builder	vector,graph_vector,graph,fulltext,graph_vector_fulltext,entity_vector,global_vector	Comma-separated chat modes compiled into the frontend.
LLM_GRAPH_BUILDER_TIME_PER_PAGE	llm-graph-builder	50	Frontend build argument VITE_TIME_PER_PAGE (upstream setting). Changing it requires an image rebuild.
LLM_GRAPH_BUILDER_CHUNK_SIZE	llm-graph-builder	5242880	Frontend build argument VITE_CHUNK_SIZE (upstream setting). Changing it requires an image rebuild.
LLM_GRAPH_BUILDER_LARGE_FILE_SIZE	llm-graph-builder	5242880	Frontend build argument VITE_LARGE_FILE_SIZE (upstream setting). Changing it requires an image rebuild.
LLM_GRAPH_BUILDER_BATCH_SIZE	llm-graph-builder	2	Frontend build argument VITE_BATCH_SIZE (upstream setting). Changing it requires an image rebuild.
LLM_GRAPH_BUILDER_CHUNK_OVERLAP	llm-graph-builder	20	Frontend build argument VITE_CHUNK_OVERLAP (upstream setting). Changing it requires an image rebuild.
LLM_GRAPH_BUILDER_TOKENS_PER_CHUNK	llm-graph-builder	200	Frontend build argument VITE_TOKENS_PER_CHUNK (upstream setting). Changing it requires an image rebuild.
LLM_GRAPH_BUILDER_CHUNK_TO_COMBINE	llm-graph-builder	1	Frontend build argument VITE_CHUNK_TO_COMBINE (upstream setting). Changing it requires an image rebuild.
LLM_GRAPH_BUILDER_MAX_TOKEN_CHUNK_SIZE	llm-graph-builder	10000	Passed to the Graph Builder backend as MAX_TOKEN_CHUNK_SIZE (upstream setting).
LLM_GRAPH_BUILDER_KNN_MIN_SCORE	llm-graph-builder	0.94	Passed to the Graph Builder backend as KNN_MIN_SCORE (upstream setting).
LLM_GRAPH_BUILDER_UPDATE_GRAPH_CHUNKS_PROCESSED	llm-graph-builder	20	Passed to the Graph Builder backend as UPDATE_GRAPH_CHUNKS_PROCESSED (upstream setting).
LLM_GRAPH_BUILDER_ENTITY_EMBEDDING	llm-graph-builder	false	Passed to the Graph Builder backend as ENTITY_EMBEDDING (upstream setting).
LLM_GRAPH_BUILDER_ENABLE_USER_AGENT	llm-graph-builder	true	Passed to the Graph Builder backend as ENABLE_USER_AGENT (upstream setting).
LLM_GRAPH_BUILDER_USER_AGENT	llm-graph-builder	Atlas-LLM-Graph-Builder	Passed to the Graph Builder backend as USER_AGENT (upstream setting).
LLM_GRAPH_BUILDER_TRACK_USER_USAGE	llm-graph-builder	false	Passed to the Graph Builder backend as TRACK_USER_USAGE (upstream setting).
LLM_GRAPH_BUILDER_AUTHENTICATION_REQUIRED	llm-graph-builder	false	Passed to the Graph Builder backend as AUTHENTICATION_REQUIRED (upstream setting): upstream backend authentication; Atlas keeps it false and relies on Kong Basic Auth.
LLM_GRAPH_BUILDER_BACKEND_SCALE	llm-graph-builder		Set by ./start.sh from LLM_GRAPH_BUILDER_SOURCE; do not edit. 1 runs llm-graph-builder-backend, 0 does not.
LLM_GRAPH_BUILDER_FRONTEND_SCALE	llm-graph-builder		Set by ./start.sh from LLM_GRAPH_BUILDER_SOURCE; do not edit. 1 runs llm-graph-builder-frontend, 0 does not.
LLM_GRAPH_BUILDER_ENDPOINT	llm-graph-builder		Internal frontend URL.
LLM_GRAPH_BUILDER_BACKEND_ENDPOINT	llm-graph-builder		Internal backend API URL.
LLM_GRAPH_BUILDER_LITELLM_MODEL_CONFIG	llm-graph-builder		Model config consumed as LLM_MODEL_CONFIG_ATLAS_LITELLM by the upstream backend.
LOCAL_DEEP_RESEARCHER_SOURCE	local-deep-researcher	container	Deployment source for Local Deep Researcher: container or disabled.
LOCAL_DEEP_RESEARCHER_PORT	local-deep-researcher	63097	Host port for the Local Deep Researcher LangGraph API (in-container 2024).
LOCAL_DEEP_RESEARCHER_REF	local-deep-researcher	38f769f84380f2065de76021ac7c5215f88aa39e	Full upstream local-deep-researcher commit SHA materialized into the managed repo volume. Update deliberately after compatibility validation; runtime startup never follows a branch.
LOCAL_DEEP_RESEARCHER_LANGGRAPH_CLI_VERSION	local-deep-researcher	0.4.31	Exact langgraph-cli[inmem] version used to serve the pinned source tree.
LOCAL_DEEP_RESEARCHER_UPSTREAM_LOCK_SHA256	local-deep-researcher	26fc35ac377836de6628e5f7b180944c4d4bd50a5e9f0200bd6e663f20e35c1a	SHA-256 digest of uv.lock at LOCAL_DEEP_RESEARCHER_REF; startup rejects a source tree whose lock does not match.
LOCAL_DEEP_RESEARCHER_LOOPS	local-deep-researcher	3	Maximum web research loops per run, passed as MAX_WEB_RESEARCH_LOOPS.
LOCAL_DEEP_RESEARCHER_SEARCH_API	local-deep-researcher	searxng	Web search backend, passed as SEARCH_API. Default searxng.
LOCAL_DEEP_RESEARCHER_WORKERS	local-deep-researcher	3	Reserved: passed to the container, but the entrypoint does not pass it to langgraph dev. One server process owns the in-memory research state.
LOCAL_DEEP_RESEARCHER_FULL_PAGE_MODE	local-deep-researcher	disabled	Full-page extraction mode: disabled, builtin, or crawl4ai. The crawl4ai mode requires CRAWL4AI_SOURCE=container.
FETCH_FULL_PAGE	local-deep-researcher		Set by ./start.sh from LOCAL_DEEP_RESEARCHER_FULL_PAGE_MODE; do not edit. true for builtin or crawl4ai, false for disabled.
LOCAL_DEEP_RESEARCHER_SCALE	local-deep-researcher		Set by ./start.sh from LOCAL_DEEP_RESEARCHER_SOURCE; do not edit. 0 when disabled, else 1.
LOCAL_DEEP_RESEARCHER_IMAGE	local-deep-researcher	python:3.11.15-slim	Container image for local-deep-researcher.
LOKI_SOURCE	loki	disabled	Deployment source for Loki, the local log store: container or disabled.
LOKI_SCALE	loki		Set by ./start.sh from LOKI_SOURCE; do not edit. 1 runs loki, 0 does not.
LOKI_ENDPOINT	loki		Internal Loki API endpoint consumed by Grafana.
LOKI_RETENTION_PERIOD	loki	24h	Short local retention period for Loki compactor-managed log deletion.
LOKI_IMAGE	loki	grafana/loki:3.7.0	Container image for loki.
MCP_POSTGRES_DB_USER	mcp-servers	atlas_mcp	Read-only PostgreSQL login for the curated MCP query tool.
MCP_POSTGRES_DB_PASSWORD	mcp-servers	atlas-db-password	Auto-generated password for the MCP read-only database role.
MCP_SERVERS_SOURCE	mcp-servers	disabled	Curated MCP package source. Disabled by default because MCP tools can reach stack internals.
MCP_SERVERS_PORT	mcp-servers	63078	Host port for the curated MCP Streamable HTTP endpoint (container port 8000).
MCP_SERVERS_SCALE	mcp-servers		Replicas: 1 when MCP_SERVERS_SOURCE=container, else 0.
MCP_POSTGRES_MAX_ROWS	mcp-servers	50	Maximum rows returned by the Postgres MCP query tool.
MCP_NEO4J_MAX_ROWS	mcp-servers		Maximum rows returned by the Neo4j MCP query tool. Empty falls back to MCP_POSTGRES_MAX_ROWS (default 50).
MCP_SEARXNG_MAX_RESULTS	mcp-servers	5	Maximum SearXNG results returned by the web-search MCP tool.
MCP_TOOL_TIMEOUT_SECONDS	mcp-servers	15	Timeout for upstream database/search calls made by MCP tools.
MCP_SERVERS_IMAGE	mcp-servers	python:3.12.14-slim	Used as build-arg BASE_IMAGE in build/Dockerfile.
MINIO_SOURCE	minio	container	MinIO deployment source: container or disabled.
MINIO_PORT	minio	63020	Host port for the MinIO S3 API (in-container listen port is 9000).
MINIO_CONSOLE_PORT	minio	63021	Host port for the MinIO admin console (in-container listen port is 9001).
MINIO_ROOT_USER	minio	minioadmin	MinIO root username. minio-init, airflow and backup also receive it; consumer services use their own scoped access keys.
MINIO_ROOT_PASSWORD	minio		Auto-generated by bootstrapper KeyGenerator.generate_minio_root_password().
MINIO_REGION	minio	us-east-1	S3 region MinIO reports. Clients such as iceberg-rest, jupyterhub, langfuse, label-studio and the asset services use it as their S3 region.
MINIO_INIT_SOURCE	minio	container	Auto-managed by minio. Container/disabled.
MINIO_SCALE	minio		Set by ./start.sh from MINIO_SOURCE; do not edit. 1 runs the minio container, 0 does not.
MINIO_INIT_SCALE	minio		Set by ./start.sh from MINIO_SOURCE; do not edit. 1 runs the minio-init container, 0 does not.
MINIO_ENDPOINT	minio		In-network URL (S3 API).
MINIO_PUBLIC_ENDPOINT	minio		Host-side URL (for browser redirects).
MINIO_PUBLIC_CONSOLE_ENDPOINT	minio		Derived by ./start.sh; do not set. Host URL of the MinIO console (http://localhost:${MINIO_CONSOLE_PORT}), empty when disabled. Passed to MinIO as MINIO_BROWSER_REDIRECT_URL.
MINIO_EXTRA_CONSUMERS	minio		Optional space-separated parent-owned MinIO entries using CONSUMER:BUCKET_VAR:ACCESS_VAR:SECRET_VAR[:RW_BUCKET_VAR,...[:RO_BUCKET_VAR,...]] grammar for services/_user overlays.
MINIO_BUCKET_ASSET_INPUTS	minio	raw-assets	Shared input bucket for the Asset Worker and Asset Baker scoped service accounts.
MINIO_ASSET_INGEST_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for writing shared raw asset inputs.
MINIO_ASSET_INGEST_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for writing shared raw asset inputs.
MINIO_ASSET_WORKER_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for Asset Worker input and output buckets.
MINIO_ASSET_WORKER_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for Asset Worker input and output buckets.
MINIO_ASSET_BAKER_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for Asset Baker input and output buckets.
MINIO_ASSET_BAKER_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for Asset Baker input and output buckets.
MINIO_BUCKET_COMFYUI	minio	comfyui	Bucket minio-init creates for the comfyui service account.
MINIO_COMFYUI_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for the comfyui bucket.
MINIO_COMFYUI_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for the comfyui bucket.
MINIO_BUCKET_BACKEND	minio	backend	Bucket minio-init creates for the backend service account.
MINIO_BACKEND_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for the backend bucket.
MINIO_BACKEND_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for the backend bucket.
MINIO_BUCKET_N8N	minio	n8n	Bucket minio-init creates for the n8n service account.
MINIO_N8N_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for the n8n bucket.
MINIO_N8N_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for the n8n bucket.
MINIO_BUCKET_JUPYTER	minio	jupyter	Bucket minio-init creates for the jupyter service account, which can also read the Iceberg lakehouse bucket.
MINIO_JUPYTER_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for the jupyter bucket.
MINIO_JUPYTER_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for the jupyter bucket.
MINIO_BUCKET_SPARK_HISTORY	minio	spark-history	Spark event-log bucket used by Zeppelin and the Spark History Server.
MINIO_SPARK_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for Spark event logs and lakehouse workflow buckets.
MINIO_SPARK_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for Spark event logs and lakehouse workflow buckets.
MINIO_BUCKET_DOCLING	minio	docling	Bucket minio-init creates for the docling service account.
MINIO_DOCLING_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for the docling bucket.
MINIO_DOCLING_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for the docling bucket.
MINIO_BUCKET_LANGFUSE	minio	langfuse	Bucket minio-init creates for the langfuse service account. Langfuse uses it for event, media and batch-export uploads.
MINIO_LANGFUSE_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for the langfuse bucket.
MINIO_LANGFUSE_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for the langfuse bucket.
MINIO_BUCKET_MLFLOW	minio	mlflow	Bucket minio-init creates for the mlflow service account. MLflow stores run artifacts in it.
MINIO_MLFLOW_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for the mlflow bucket.
MINIO_MLFLOW_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for the mlflow bucket.
MINIO_BUCKET_LABEL_STUDIO	minio	label-studio	Bucket minio-init creates for the label-studio service account. Label Studio uses it as its storage bucket.
MINIO_LABEL_STUDIO_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for the label-studio bucket.
MINIO_LABEL_STUDIO_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for the label-studio bucket.
MINIO_BUCKET_ICEBERG_LAKEHOUSE	minio	lakehouse	Iceberg warehouse bucket. The iceberg and spark service accounts can write it; the jupyter account can read it.
MINIO_BUCKET_ICEBERG_JARS	minio	jars	Jar bucket for Spark jobs. The spark and iceberg service accounts can write it; Airflow and Jenkins receive its name.
MINIO_BUCKET_ICEBERG_CHECKPOINTS	minio	checkpoints	Checkpoint bucket. The spark and iceberg service accounts can write it; Zeppelin and JupyterHub receive its name.
MINIO_BUCKET_ICEBERG_LANDING	minio	landing	Landing bucket. The spark and iceberg service accounts can write it; Airflow, Zeppelin and JupyterHub receive its name.
MINIO_ICEBERG_ACCESS_KEY	minio		Auto-generated scoped MinIO access key for the Iceberg lakehouse buckets.
MINIO_ICEBERG_SECRET_KEY	minio		Auto-generated scoped MinIO secret key for the Iceberg lakehouse buckets.
MINIO_IMAGE	minio	pgsty/silo:RELEASE.2026-09-16T00-00-00Z@sha256:635197cb9f36d01bee221d34d1c7d7960f6a95c48b0b6c01d99cd13bdae51a46	Silo, the Pigsty-maintained AGPL fork of the archived MinIO community server, pinned by tag and multi-arch digest. It keeps the MINIO_* environment, /minio/* routes, mc admin API and on-disk format. Its entrypoint maps server to silo server, so the compose command is unchanged. It carries the bounded S3 Select CSV reader fix for CVE-2026-39414. Review upstream security advisories before bumping.
MINIO_INIT_IMAGE	minio	pgsty/mc:RELEASE.2026-09-16T00-00-00Z@sha256:cfc83108c3abb371f8fb84d99c1fdc88f8c237e022409b0081fb7c0a3be634dd	Container image for minio-init.
MLFLOW_SOURCE	mlflow	disabled	MLflow deployment source: container or disabled. container requires MINIO_SOURCE=container for artifact storage.
MLFLOW_PORT	mlflow	63092	Host port for the MLflow UI/API (in-container 5000).
MLFLOW_ENDPOINT	mlflow		In-network URL for Atlas services and notebooks.
MLFLOW_TRACKING_URI	mlflow		Tracking URI injected into JupyterHub when MLflow is enabled.
MLFLOW_SCALE	mlflow		Set by ./start.sh from MLFLOW_SOURCE; do not edit. 1 runs the mlflow container, 0 does not.
MLFLOW_INIT_SCALE	mlflow		Set by ./start.sh from MLFLOW_SOURCE; do not edit. 1 runs the mlflow-init container, 0 does not.
MLFLOW_DB_NAME	mlflow	mlflow	Postgres database in supabase-db that holds the MLflow tracking and model-registry stores.
MLFLOW_DB_NAME_URI	mlflow	mlflow	Bootstrapper-synchronized percent-encoded MLFLOW_DB_NAME for DSN path components.
MLFLOW_DB_USER	mlflow	mlflow	Restricted Postgres login that owns the MLflow database; created by 05-scoped-roles.sh.
MLFLOW_DB_PASSWORD	mlflow	atlas-db-password	Auto-generated Postgres password for the dedicated MLflow database role.
MLFLOW_DB_USER_URI	mlflow	mlflow	Bootstrapper-synchronized percent-encoded MLFLOW_DB_USER for DSN userinfo.
MLFLOW_DB_PASSWORD_URI	mlflow	atlas-db-password	Bootstrapper-synchronized percent-encoded MLFLOW_DB_PASSWORD for DSN userinfo.
MLFLOW_IMAGE	mlflow	ghcr.io/mlflow/mlflow:v3.16.1	Container image for mlflow.
MLFLOW_INIT_IMAGE	mlflow	alpine:3.24.2	Container image for mlflow-init.
N8N_SOURCE	n8n	container	n8n deployment source: container or disabled. Drives the n8n, n8n-worker and n8n-init scales.
N8N_DB_USER	n8n	atlas_n8n	Owner of the isolated n8n schema in Supabase PostgreSQL.
N8N_DB_PASSWORD	n8n	atlas-db-password	Auto-generated password for the scoped n8n database role.
N8N_PORT	n8n	63075	Host port for the n8n editor and webhooks (container port 5678).
N8N_ENCRYPTION_KEY	n8n	your-random-encryption-key	Encryption key for n8n credentials and workflow secrets; auto-generated by the bootstrapper when left as the placeholder.
N8N_API_KEY	n8n		n8n public-API key (n8n CE issues these only via the UI: Settings → n8n API). Consumed by the backend's /workflows endpoints; while empty those endpoints currently return backend errors from failed upstream n8n calls.
N8N_HOST	n8n	n8n.localhost	Not interpolated: the n8n compose fragment sets N8N_HOST=n8n.localhost directly.
N8N_PROTOCOL	n8n	http	Protocol n8n uses in the URLs it generates (N8N_PROTOCOL in the n8n and n8n-worker containers).
N8N_EXECUTIONS_MODE	n8n	queue	n8n EXECUTIONS_MODE. The default queue runs workflow executions on n8n-worker through Redis.
N8N_INIT_SOURCE	n8n	container	Init-container selector listed in the SOURCE values reference. N8N_INIT_SCALE follows N8N_SOURCE, so n8n-init runs whenever n8n runs.
N8N_INIT_NODES	n8n	n8n-nodes-comfyui@0.0.9,@ksc1234/n8n-nodes-comfyui-image-to-image@1.0.2	Exact community package specs installed before n8n starts. Atlas' default ComfyUI set uses the committed package-lock; n8n 2.28.2 already supplies first-party MCP nodes. Custom values must use name@x.y.z for every comma-separated package.
N8N_SEED_HTTP_TIMEOUT_MS	n8n	10000	Absolute deadline in milliseconds for each consumer workflow seeder HTTP operation; must be positive.
N8N_SEED_COMMAND_TIMEOUT_MS	n8n	120000	Deadline in milliseconds for each consumer workflow n8n CLI operation; must be positive.
N8N_SEED_MAX_RESPONSE_BYTES	n8n	1048576	Maximum bytes accepted from one consumer workflow seeder HTTP response; must be positive.
N8N_SCALE	n8n		Set by ./start.sh from N8N_SOURCE; do not edit. 1 runs the n8n container, 0 does not.
N8N_WORKER_SCALE	n8n		Set by ./start.sh from N8N_SOURCE; do not edit. 1 runs the n8n-worker container, 0 does not.
N8N_INIT_SCALE	n8n		Set by ./start.sh from N8N_SOURCE; do not edit. 1 runs the n8n-init container, 0 does not.
N8N_IMAGE	n8n	n8nio/n8n:2.28.2	Container image for n8n.
N8N_INIT_IMAGE	n8n	n8nio/n8n:2.28.2	Container image for n8n-init.
NEO4J_GRAPH_DB_SOURCE	neo4j	container	Neo4j deployment source. Drives NEO4J_SCALE and NEO4J_URI.
GRAPH_DB_PORT	neo4j	63023	Host port for Neo4j bolt (container port 7687).
GRAPH_DB_DASHBOARD_PORT	neo4j	63024	Host port for Neo4j browser dashboard (container port 7474).
GRAPH_DB_USER	neo4j	neo4j	Neo4j username. Backend and jupyterhub receive it as NEO4J_USER, and airflow-init seeds it into neo4j_default. Keep it equal to the user in GRAPH_DB_AUTH.
GRAPH_DB_PASSWORD	neo4j	neo4j_password	Neo4j password; auto-generated by the bootstrapper when left as the placeholder.
GRAPH_DB_AUTH	neo4j	neo4j/neo4j_password	Neo4j auth in username/password format.
NEO4J_URI	neo4j		Computed from NEO4J_GRAPH_DB_SOURCE. Read by backend/jupyterhub.
NEO4J_SCALE	neo4j		Replicas: 1 for container, 0 otherwise.
NEO4J_LOCALHOST_HTTP_PORT	neo4j	7474	Host HTTP port for the neo4j-localhost source variant (browser UI + REST). Kong route target is derived as http://host.docker.internal:7474.
NEO4J_LOCALHOST_BOLT_PORT	neo4j	7687	Host Bolt port for the neo4j-localhost source variant (driver protocol). NEO4J_URI is derived as bolt://host.docker.internal:7687. The wizard's inline-textbox edits THIS port (HTTP is overridable via .env hand-edit).
NEO4J_MEMORY_LIMIT	neo4j	2g	Memory limit for neo4j-graph-db (deploy.resources.limits.memory). 2g is conservative for a 32 GB host. Single-node dev and staging workloads typically stay well under it.
NEO4J_CPU_LIMIT	neo4j	1.5	Container CPU limit for neo4j-graph-db (deploy.resources.limits.cpus). 1.5 cores is sufficient for single-node Bolt + HTTP query serving.
NEO4J_GRAPH_DB_IMAGE	neo4j	neo4j:5.26.31	Used as build-arg BASE_IMAGE in graph-db/Dockerfile.
LLM_PROVIDER_SOURCE	ollama	ollama-container-cpu	Where LiteLLM's Ollama upstream runs: ollama-container-cpu, ollama-container-gpu, ollama-localhost (host daemon) or none (cloud providers only). Also accepts auto in a consumer manifest.
OLLAMA_USER_MODELS	ollama	qwen3.8:latest,nomic-embed-text,qwen3-embedding:0.6b	Comma-separated Ollama model names to activate/pull. Pre-filled by env_assembler from the catalog's default-active Ollama set (see services/ollama/models.yaml); the wizard overwrites it with the user's selection.
OLLAMA_CUSTOM_MODELS	ollama		Free-text extra Ollama models; a consumer's model_sidecars.ollama lands here. The ollama-pull init container pulls them for container-* sources, and ./start.sh pulls them onto the host daemon for ollama-localhost.
OLLAMA_PULL_STALL_TIMEOUT_SECONDS	ollama	120	Seconds an ollama-pull download may stay below 1 KiB/s before it is aborted and retried. A download that keeps progressing has no wall-clock cutoff.
OLLAMA_AUTO_IMPORT_LOCAL_MODELS	ollama	true	When LLM_PROVIDER_SOURCE=ollama-localhost, litellm-init queries the upstream's /api/tags and auto-imports every model found, unioning with OLLAMA_USER_MODELS. Set false to keep strict wizard-only control.
OLLAMA_DEPLOY_RESOURCES	ollama	~	Reserved for future GPU deploy block. Currently '~' (null).
OLLAMA_SCALE	ollama		Set by ./start.sh from LLM_PROVIDER_SOURCE; do not edit. 1 runs the ollama container (container-* sources), 0 does not.
OLLAMA_PULL_SCALE	ollama		Set by ./start.sh from LLM_PROVIDER_SOURCE; do not edit. 1 runs the ollama-pull init container for container-* sources, else 0.
OLLAMA_NVIDIA_VISIBLE_DEVICES	ollama		Set by ./start.sh from LLM_PROVIDER_SOURCE; do not edit. all for ollama-container-gpu, else null. Passed to the ollama container as NVIDIA_VISIBLE_DEVICES.
OLLAMA_ENDPOINT	ollama		Auto-managed; do not set. The per-source Ollama URL (empty for none). ./start.sh writes it into LITELLM_OLLAMA_UPSTREAM; no container reads OLLAMA_ENDPOINT directly.
LITELLM_OLLAMA_UPSTREAM	ollama		Ollama URL the LiteLLM gateway forwards Ollama-routed requests to. Owned by ollama because the value flips with LLM_PROVIDER_SOURCE; consumed by litellm-init when rendering its config.yaml.
OLLAMA_LOCALHOST_PORT	ollama	11434	Host port for the ollama-localhost source variant. URL is derived at compose-render time as http://host.docker.internal:11434.
OLLAMA_NUM_PARALLEL	ollama	8	Ollama parallel request slots per model. Multi-agent consumers (8+ concurrent requests) need more than 1; Ollama's default of 1 serializes them. Applies to the container-* sources; for ollama-localhost the host daemon owns it.
OLLAMA_MAX_LOADED_MODELS	ollama	2	Maximum models Ollama keeps resident at once. Raise it for workloads that swap between several models; lower it to bound memory. Applies to the container-* sources; for ollama-localhost the host daemon owns it.
OLLAMA_KV_CACHE_TYPE	ollama	q8_0	KV cache quantization: f16 (full precision, Ollama's default), q8_0 (about half the memory, negligible quality cost) or q4_0 (about a quarter, measurable cost). Needs OLLAMA_FLASH_ATTENTION active. Container-* sources only; ollama-localhost uses the host daemon's setting.
OLLAMA_FLASH_ATTENTION	ollama	1	Enables Ollama's flash-attention path. Current Ollama enables it automatically where supported. Atlas pins it because OLLAMA_KV_CACHE_TYPE quantization has no effect without it. Set 0 to force it off.
OLLAMA_KEEP_ALIVE	ollama		Time Ollama keeps a model loaded after its last use (5m, 1h, or -1 for forever); empty uses Ollama's 5m default. Container-* sources only. -1 keeps every loaded model in RAM until reverted and the daemon restarts.
OLLAMA_MODELS_RESIDENT_MIN	ollama		Advisory, ollama-localhost only: how many models a workload needs resident at once. ./start.sh doctor reports a failure when the daemon's OLLAMA_MAX_LOADED_MODELS is lower. Empty declares no expectation. A LightRAG run with separate extract, embed and keyword models needs 3.
OLLAMA_PARALLEL_MIN	ollama		Advisory, ollama-localhost only: the concurrent-request floor a consumer needs. ./start.sh doctor warns when the daemon's OLLAMA_NUM_PARALLEL is lower. Empty declares no expectation. Ollama's default of 1 slot silently serializes a multi-agent consumer.
LLM_PROVIDER_IMAGE	ollama	ollama/ollama:0.30.11	Container image for ollama.
OLLAMA_PULL_IMAGE	ollama	alpine:3.24.2@sha256:294b683cb724975bec92580e1e685676bd4b50bda910ddb8c51d4cabeaec77e6	Container image for ollama-pull.
OPEN_WEB_UI_SOURCE	open-webui	container	Open WebUI deployment source: container or disabled.
OPEN_WEBUI_DB_USER	open-webui	atlas_open_webui	Scoped PostgreSQL owner for Open WebUI migration objects.
OPEN_WEBUI_DB_PASSWORD	open-webui	atlas-db-password	Auto-generated password for the scoped Open WebUI database role.
OPEN_WEBUI_DB_USER_URI	open-webui	atlas_open_webui	Bootstrapper-synchronized percent-encoded OPEN_WEBUI_DB_USER for DSN userinfo.
OPEN_WEBUI_DB_PASSWORD_URI	open-webui	atlas-db-password	Bootstrapper-synchronized percent-encoded OPEN_WEBUI_DB_PASSWORD for DSN userinfo.
OPEN_WEB_UI_STT_API_KEY	open-webui		Parakeet bearer token when a Parakeet source is selected; a non-secret compatibility placeholder for other enabled STT engines.
OPEN_WEB_UI_PORT	open-webui	63096	Host port for the Open WebUI web interface (container port 8080).
OPEN_WEB_UI_SECRET_KEY	open-webui	secret	Open WebUI session/signing secret; auto-generated by the bootstrapper when left as the placeholder.
OPEN_WEB_UI_ADMIN_EMAIL	open-webui	admin@localhost	Email of the admin account that open-webui-init creates.
OPEN_WEB_UI_ADMIN_PASSWORD	open-webui	admin	Initial Open WebUI admin password; auto-generated by the bootstrapper when left as the placeholder.
OPEN_WEB_UI_ADMIN_NAME	open-webui	Admin	Display name of the admin account that open-webui-init creates.
OPEN_WEB_UI_WEBSOCKET_SUPPORT	open-webui	true	Open WebUI ENABLE_WEBSOCKET_SUPPORT. The websocket manager uses Redis.
OPEN_WEB_UI_REDIS_DB	open-webui	2	Redis database index for Open WebUI's websocket manager. LightRAG shares db 2 by default.
OPEN_WEB_UI_MODEL_CACHE_TTL	open-webui	300	Seconds Open WebUI caches the model list (MODELS_CACHE_TTL).
OPEN_WEB_UI_ENABLE_IMAGE_GENERATION	open-webui	true	Set by bootstrapper: true unless COMFYUI_SOURCE=disabled, so Open WebUI hides its ComfyUI image toggle when there is no ComfyUI. Open WebUI keeps the value saved in its database once an admin edits the Images settings.
OPEN_WEB_UI_TTS_ENGINE	open-webui		Set by bootstrapper when speaches/chatterbox are enabled.
OPEN_WEB_UI_STT_ENGINE	open-webui		Set by ./start.sh: openai when STT_PROVIDER_SOURCE is not disabled, else empty. Passed as AUDIO_STT_ENGINE.
OPEN_WEB_UI_TTS_MODEL	open-webui		Model name Open WebUI sends to /v1/audio/speech. Empty = engine default.
OPEN_WEB_UI_TTS_VOICE	open-webui		Voice id. Empty = engine default.
OPEN_WEB_UI_TTS_API_URL	open-webui		Set by ./start.sh from TTS_ENDPOINT plus /v1, or empty. Passed as AUDIO_TTS_OPENAI_API_BASE_URL.
OPEN_WEB_UI_STT_API_URL	open-webui		Set by ./start.sh from STT_ENDPOINT plus /v1, or empty. Passed as AUDIO_STT_OPENAI_API_BASE_URL.
OPEN_WEB_UI_SCALE	open-webui		Set by ./start.sh from OPEN_WEB_UI_SOURCE; do not edit. 1 runs the open-web-ui container, 0 does not.
OPEN_WEB_UI_INIT_SCALE	open-webui		Set by ./start.sh from OPEN_WEB_UI_SOURCE; do not edit. 1 runs the open-webui-init container, 0 does not.
OPEN_WEB_UI_IMAGE	open-webui	dyrnq/open-webui:v0.6.32	Container image for open-web-ui.
OPEN_WEB_UI_INIT_IMAGE	open-webui	python:3.12.14-slim-bookworm@sha256:0f5b26b9518d002b6173fd61daad821fa340635ebfec5bba471013f9ca114579	Container image for open-webui-init.
OPENCLAW_SOURCE	openclaw	disabled	OpenClaw deployment source: container, localhost (host install on OPENCLAW_LOCALHOST_PORT) or disabled.
OPENCLAW_GATEWAY_PORT	openclaw	63076	Host port for the OpenClaw gateway API and dashboard (container port 18789).
OPENCLAW_BRIDGE_PORT	openclaw	63077	Host port for the OpenClaw bridge (container port 18790).
OPENCLAW_GATEWAY_TOKEN	openclaw		Auto-generated bearer token for the OpenClaw gateway API and dashboard; kept once set. The gateway binds to the LAN interface and refuses to do so without a token. Left empty, it would mint an unreadable per-start token.
OPENCLAW_LOCALHOST_PORT	openclaw	63065	Host port for the localhost source variant (its own slot, separate from OPENCLAW_GATEWAY_PORT). URL is derived at compose-render time as http://host.docker.internal:63065.
OPENCLAW_INIT_SOURCE	openclaw	container	Init-container selector listed in the SOURCE values reference. OPENCLAW_INIT_SCALE follows OPENCLAW_SOURCE, so openclaw-init runs only for container.
OPENCLAW_ANTHROPIC_API_KEY	openclaw		Optional per-OpenClaw Anthropic override.
OPENCLAW_OPENAI_API_KEY	openclaw		Optional per-OpenClaw OpenAI override. Leave empty to keep OpenClaw routed through LiteLLM instead of bypassing the stack gateway.
OPENCLAW_SCALE	openclaw		Set by ./start.sh from OPENCLAW_SOURCE; do not edit. 1 runs the openclaw-gateway container for container, else 0.
OPENCLAW_INIT_SCALE	openclaw		Set by ./start.sh from OPENCLAW_SOURCE; do not edit. 1 runs the openclaw-init container for container, else 0.
OPENCLAW_ENDPOINT	openclaw		Resolved per source by ./start.sh: the container DNS name, a host.docker.internal URL, or empty when disabled. No compose fragment injects it into a container; consumers wire it themselves.
OPENCLAW_IMAGE	openclaw	ghcr.io/openclaw/openclaw:2026.6.10	Container image for openclaw-gateway.
OPENCLAW_INIT_IMAGE	openclaw	alpine:3.24.2@sha256:294b683cb724975bec92580e1e685676bd4b50bda910ddb8c51d4cabeaec77e6	Container image for openclaw-init.
OTEL_COLLECTOR_SOURCE	otel-collector	disabled	OpenTelemetry Collector deployment source: container or disabled.
OTEL_COLLECTOR_SCALE	otel-collector		Set by ./start.sh from OTEL_COLLECTOR_SOURCE; do not edit. 1 runs the otel-collector container, 0 does not.
OTEL_COLLECTOR_ENDPOINT	otel-collector		Preferred internal OTLP HTTP endpoint for Atlas services.
OTEL_COLLECTOR_OTLP_HTTP_ENDPOINT	otel-collector		Internal OTLP HTTP endpoint (http://otel-collector:4318) when enabled.
OTEL_COLLECTOR_OTLP_GRPC_ENDPOINT	otel-collector		Internal OTLP gRPC endpoint (http://otel-collector:4317) when enabled.
ATLAS_OTEL_ENABLED	otel-collector		Boolean string consumed by backend and LiteLLM tracing toggles.
OTEL_COLLECTOR_IMAGE	otel-collector	otel/opentelemetry-collector-contrib:0.154.0	Container image for otel-collector.
STT_PROVIDER_SOURCE	parakeet	speaches-container-cpu	Speech-to-text engine: speaches-container-cpu, speaches-container-gpu, parakeet-container-gpu, parakeet-localhost, whisper-cpp-localhost or disabled. Drives STT_ENDPOINT and the engine scales.
STT_PROVIDER_PORT	parakeet	63055	Wizard display slot. Real per-engine ports live on their own manifests (PARAKEET_*, SPEACHES_*, etc.).
STT_PROVIDER_SCALE	parakeet		Set by ./start.sh; do not edit. 1 when a container engine serves speech-to-text, else 0. The wizard uses it to colour the STT row.
PARAKEET_MODEL	parakeet	nvidia/parakeet-tdt-0.6b-v3	Model the Parakeet GPU container loads (NeMo model ID), passed as PARAKEET_MODEL.
PARAKEET_GPU_DEVICE	parakeet	cuda	Torch device for the Parakeet GPU container, passed as PARAKEET_DEVICE.
PARAKEET_MAX_UPLOAD_BYTES	parakeet	104857600	Maximum audio upload size in bytes for the Parakeet GPU and localhost APIs. The request body is capped before multipart parsing, with 1 MiB added for framing. Larger requests return 413. An invalid value stops startup.
PARAKEET_UPLOAD_TIMEOUT_SECONDS	parakeet	120	Total seconds (1-3600) allowed to receive one Parakeet upload body before a 408 response releases admission capacity.
PARAKEET_CONCURRENCY	parakeet	1	Maximum concurrent Parakeet inference calls per provider process. Default 1 prevents model thread-safety and GPU memory contention.
PARAKEET_API_TOKEN	parakeet		Auto-generated bearer credential required by Parakeet routes except /health.
PARAKEET_AUTH_MODE	parakeet	required	Provider authentication mode: required (default) or disabled for an explicit emergency/local rollback.
PARAKEET_CORS_ORIGINS	parakeet		Optional comma-separated browser origin allowlist. Empty disables CORS; wildcard is invalid while authentication is required.
PARAKEET_INFERENCE_TIMEOUT_SECONDS	parakeet	900	Finite model-load and transcription deadline in seconds (1-3600). A timeout returns 504 and then exits the provider process for restart.
PARAKEET_LOCALHOST_BIND_HOST	parakeet	127.0.0.1	Native Parakeet listen address. Set another interface explicitly only when remote access is intended.
PARAKEET_LOCALHOST_PORT	parakeet	63042	Host port for the parakeet-localhost source variant. URL is derived at compose-render time as http://host.docker.internal:63042.
WHISPER_CPP_LOCALHOST_PORT	parakeet	63042	Host port for the whisper-cpp-localhost source variant, same as parakeet because the two modes are mutually exclusive. URL is derived at compose-render time as http://host.docker.internal:63042.
HUGGING_FACE_HUB_TOKEN	parakeet		Shared by parakeet, speaches, chatterbox, docling for model downloads. Consumed by ${HUGGING_FACE_HUB_TOKEN} interpolation in those services' compose blocks; owned here as its historic home.
PARAKEET_GPU_SCALE	parakeet		Set by ./start.sh from STT_PROVIDER_SOURCE; do not edit. 1 runs the parakeet-gpu container for parakeet-container-gpu, else 0.
STT_ENDPOINT	parakeet		Resolved per STT_PROVIDER_SOURCE. Consumed by n8n, backend, jupyterhub, open-webui, hermes-init (the latter as STT_INTERNAL_URL — owned by services/hermes/).
PARAKEET_GPU_IMAGE	parakeet	nvcr.io/nvidia/pytorch:26.06-py3	Used as BASE_IMAGE in stt-provider/gpu/Dockerfile.
PROMETHEUS_SOURCE	prometheus	disabled	Prometheus deployment source: container or disabled. Also turns the node-exporter, cAdvisor, postgres-exporter and redis-exporter sidecars on or off.
PROMETHEUS_PORT	prometheus	63006	Host port for Prometheus's HTTP UI + API (in-container 9090).
NODE_EXPORTER_PORT	prometheus	63007	Host port for node-exporter (in-container 9100). Host-level metrics: CPU, memory, disk, network.
CADVISOR_PORT	prometheus	63008	Host port for cAdvisor (in-container 8080). Per-container metrics: CPU, memory, IO.
PROMETHEUS_RETENTION_DAYS	prometheus	7	TSDB retention in days. User-configurable at wizard time via the inline secondary numeric input. Default 7.
PROMETHEUS_ENDPOINT	prometheus		Resolved per PROMETHEUS_SOURCE — http://prometheus:9090 for container, empty for disabled. Consumed by Grafana's datasource provisioning.
PROMETHEUS_SCALE	prometheus		Set by ./start.sh from PROMETHEUS_SOURCE; do not edit. 1 runs the prometheus container, 0 does not.
NODE_EXPORTER_SCALE	prometheus		Set by ./start.sh from PROMETHEUS_SOURCE; do not edit. 1 runs the node-exporter container, 0 does not.
CADVISOR_SCALE	prometheus		Set by ./start.sh from PROMETHEUS_SOURCE; do not edit. 1 runs the cadvisor container, 0 does not.
PROMETHEUS_IMAGE	prometheus	prom/prometheus:v2.55.1	Container image for prometheus.
NODE_EXPORTER_IMAGE	prometheus	prom/node-exporter:v1.11.1	Container image for node-exporter.
CADVISOR_IMAGE	prometheus	gcr.io/cadvisor/cadvisor:v0.55.1	Container image for cadvisor.
RAY_SOURCE	ray	disabled	Ray deployment source: ray-container-cpu, ray-container-gpu or disabled. Drives RAY_HEAD_SCALE, RAY_WORKER_SCALE, RAY_IMAGE and RAY_ADDRESS.
RAY_DASHBOARD_PORT	ray	63002	Loopback-only host port for the Ray dashboard + native REST job-submission API (in-container 8265). Use the authenticated Kong route for remote access.
RAY_GCS_PORT	ray	63003	Loopback-only host port for the Ray GCS (in-container 6379). Distinct from the project's Redis cache container; Ray bundles its own GCS process.
RAY_CLIENT_PORT	ray	63004	Loopback-only host port for the unauthenticated Ray client server (in-container 10001). Used only by trusted host Python: ray.init('ray://localhost:<port>').
RAY_WORKER_COUNT	ray	2	Number of ray-worker containers; 0 runs the head only. The wizard and --ray-worker-count accept 0-64; the wizard default is 2.
RAY_HEAD_SCALE	ray		Set by ./start.sh from RAY_SOURCE; do not edit. 1 runs the ray-head container for the container sources, else 0.
RAY_WORKER_SCALE	ray		Resolved by the bootstrapper from RAY_WORKER_COUNT when source is ray-container-{cpu,gpu}; 0 otherwise.
RAY_ADDRESS	ray		Resolved per source: ray://ray-head:10001 for container sources, empty for disabled. Injected into the backend + jupyterhub containers via their compose environment blocks.
RAY_DASHBOARD_URL	ray		Optional override for the Ray dashboard URL the backend's job client uses. Empty = derive from RAY_ADDRESS (http://ray-head:8265).
RAY_HEAD_MEMORY_LIMIT	ray	4g	Memory limit for ray-head (deploy.resources.limits.memory). The head runs the GCS, dashboard and object directory; 4g is safe on a 32 GB host. The shm_size=8gb /dev/shm allocation is separate from this cap.
RAY_HEAD_CPU_LIMIT	ray	2.0	Container CPU limit for ray-head (deploy.resources.limits.cpus). GCS + dashboard coordination rarely needs more than 1 core; 2.0 leaves room for brief spikes.
RAY_WORKER_MEMORY_LIMIT	ray	4g	Container memory limit per ray-worker (deploy.resources.limits.memory). Workers run tasks and actors; 4 g per replica is a modest OOM fence on a 32 GB host with the default 2 workers. Raise for large in-memory datasets.
RAY_WORKER_CPU_LIMIT	ray	2.0	Container CPU limit per ray-worker (deploy.resources.limits.cpus). 2.0 cores per worker. Raise if workers need more parallelism; lower if the host is saturated.
RAY_IMAGE	ray	rayproject/ray:2.56.0	CPU build. Same image used for both head and CPU workers.
RAY_GPU_IMAGE	ray	rayproject/ray:2.56.0-gpu	GPU build with CUDA runtime. Selected by ray-container-gpu source.
REDIS_SOURCE	redis	container	Single-option source: Redis always runs as a container. It is an env entry, not a sources block, so the wizard shows no one-option dropdown. The source validator expects REDIS_SOURCE=container.
REDIS_MAXMEMORY	redis	0	Redis memory cap, e.g. 512mb. 0 means unlimited: Redis grows until the container is OOM-killed and loses in-flight queue state. A cap lets REDIS_MAXMEMORY_POLICY shed recomputable data instead. Size it to about 75% of the container memory budget.
REDIS_MAXMEMORY_POLICY	redis	volatile-lru	volatile-lru evicts TTL keys, including LiteLLM cache and ordinary media records (seven-day default). Budget-tracked in-flight records, unresolved submissions and pending ledger intents persist until settlement. Without TTL keys, writes fail with OOM.
REDIS_PORT	redis	63025	Host port. Container port is always 6379.
REDIS_PASSWORD	redis	redis_password	Used by Redis consumers, including n8n queues, litellm cache, open-webui websocket storage, and backend state and queues. The default is for fresh-install convenience only; rotate before any deployment.
REDIS_URL	redis	redis://:${REDIS_PASSWORD}@redis:6379/0	Default connection string (database 0, also n8n's queue and LiteLLM's cache). Consumers that need another DB index build their own URL: open-webui and lightrag share /2, jupyterhub uses /3. Compose substitutes REDIS_PASSWORD.
REDIS_EXPORTER_PORT	redis	63026	Host port for redis-exporter (in-container 9121). Sidecar gated by PROMETHEUS_SOURCE.
REDIS_EXPORTER_SCALE	redis		Set by _generate_prometheus_config() to 1 when PROMETHEUS_SOURCE=container, 0 otherwise.
REDIS_IMAGE	redis	redis:7.2.14-alpine	Bumping the redis version: change this default and rerun ./start.sh.
REDIS_EXPORTER_IMAGE	redis	oliver006/redis_exporter:v1.86.0	Prometheus exporter sidecar. Scales 1↔0 with PROMETHEUS_SOURCE (gated by _generate_prometheus_config).
REDPANDA_SOURCE	redpanda	disabled	Redpanda deployment source: container or disabled. Drives the Redpanda, redpanda-init and Console scales plus REDPANDA_BROKERS and SPARK_KAFKA_BOOTSTRAP_SERVERS.
REDPANDA_KAFKA_PORT	redpanda	63010	Host port for the Redpanda Kafka API external listener (in-container 19092).
REDPANDA_CONSOLE_PORT	redpanda	63011	Host port for Redpanda Console (in-container 8080).
REDPANDA_DEMO_TOPICS	redpanda	atlas_stream_events	Comma-separated Kafka topics created idempotently by redpanda-init.
REDPANDA_SCALE	redpanda		Set by ./start.sh from REDPANDA_SOURCE; do not edit. 1 runs the redpanda broker, 0 does not.
REDPANDA_INIT_SCALE	redpanda		Set by ./start.sh from REDPANDA_SOURCE; do not edit. 1 runs the redpanda-init container, 0 does not.
REDPANDA_CONSOLE_SCALE	redpanda		Set by ./start.sh from REDPANDA_SOURCE; do not edit. 1 runs the redpanda-console container, 0 does not.
REDPANDA_BROKERS	redpanda		In-network Kafka bootstrap servers for Atlas services.
SPARK_KAFKA_BOOTSTRAP_SERVERS	redpanda		Spark Structured Streaming Kafka bootstrap servers; blank when Redpanda is disabled.
REDPANDA_IMAGE	redpanda	docker.redpanda.com/redpandadata/redpanda:v26.1.12	Pinned to the latest Redpanda 26.1.x maintenance release verified for multi-arch Docker images.
REDPANDA_CONSOLE_IMAGE	redpanda	docker.redpanda.com/redpandadata/console:v3.8.0	Pinned to the latest Redpanda Console 3.8.x release.
SEARXNG_SOURCE	searxng	container	SearXNG deployment source: container or disabled.
SEARXNG_PORT	searxng	63056	Host port for the SearXNG web UI and JSON API (container port 8080). Also sets SEARXNG_BASE_URL.
SEARXNG_SECRET	searxng		Auto-generated by bootstrapper.
SEARXNG_SCALE	searxng		Set by ./start.sh from SEARXNG_SOURCE; do not edit. 1 runs the searxng container, 0 does not.
SEARXNG_IMAGE	searxng	searxng/searxng:2026.6.28-357662d86	Container image for searxng.
SPARK_SOURCE	spark	disabled	Deployment source for the Spark standalone cluster (master, workers, history, Spark Connect, spark-init): container or disabled (default).
SPARK_MASTER_UI_PORT	spark	63027	Host port for the Spark Master Web UI (in-container 8080).
SPARK_HISTORY_PORT	spark	63028	Host port for the Spark History Server (in-container 18080).
SPARK_WORKER_COUNT	spark	2	Number of spark-worker replicas alongside the master. 1-8.
SPARK_MASTER_SCALE	spark		Set by ./start.sh from SPARK_SOURCE; do not edit. 1 runs spark-master, 0 does not.
SPARK_WORKER_SCALE	spark		Resolved by the bootstrapper from SPARK_WORKER_COUNT when source=container; 0 otherwise.
SPARK_HISTORY_SCALE	spark		Set by ./start.sh from SPARK_SOURCE; do not edit. 1 runs the Spark History Server, 0 does not.
SPARK_INIT_SCALE	spark		Set by ./start.sh from SPARK_SOURCE; do not edit. 1 runs spark-init, which creates the spark-history bucket; 0 does not.
SPARK_CONNECT_SCALE	spark		Resolved by the bootstrapper from SPARK_SOURCE; spark-connect runs as a sidecar to spark-master.
SPARK_CONNECT_CORES_MAX	spark	1	Maximum standalone-cluster cores the long-lived Spark Connect server may hold. Keep below total worker cores so Airflow SparkSubmit and Zeppelin standalone workloads can run alongside Connect; raise only when the worker pool has enough headroom.
SPARK_MASTER_MEMORY_LIMIT	spark	2g	Container memory limit for spark-master (deploy.resources.limits.memory). The standalone master process manages cluster state and requires minimal heap — 2 g is generous on a 32 GB host.
SPARK_MASTER_CPU_LIMIT	spark	1.0	Container CPU limit for spark-master (deploy.resources.limits.cpus). Master is lightweight; 1.0 core is ample.
SPARK_WORKER_MEMORY_LIMIT	spark	4g	Container memory limit per spark-worker (deploy.resources.limits.memory). Workers run executor JVMs; 4 g per replica is a conservative cap on a 32 GB host with the default 2 workers (8 g total). Raise for memory-intensive jobs.
SPARK_WORKER_CPU_LIMIT	spark	2.0	Container CPU limit per spark-worker (deploy.resources.limits.cpus). 2.0 cores per worker leaves headroom for other services on the host.
SPARK_IMAGE	spark	apache/spark:4.1.2	Container image for spark-master.
SPEACHES_GPU_IMAGE	speaches	ghcr.io/speaches-ai/speaches:0.9.0-rc.3-cuda	CUDA build of the same image. When the speaches-gpu profile wins, the bootstrapper writes this value into SPEACHES_IMAGE (the compose interpolation target).
SPEACHES_TTS_MODEL	speaches	speaches-ai/Kokoro-82M-v1.0-ONNX	Model id Open WebUI sends in /v1/audio/speech requests. It must be the Speaches Kokoro ONNX executor id (speaches-ai/Kokoro-82M-v1.0-ONNX). The HuggingFace 'hexgrad/Kokoro-82M' id is invalid for Speaches and returns 404.
SPEACHES_STT_MODEL	speaches	Systran/faster-distil-whisper-large-v3	Inert: Speaches preloads from a hard-coded PRELOAD_MODELS list, and Open WebUI requests whisper-1. Changing this value has no effect.
SPEACHES_PORT	speaches	63060	Host port for the Speaches container (in-container listen port is 8000).
SPEACHES_SCALE	speaches		Computed by bootstrapper/services/service_config.py::_generate_stt_provider_config (or _generate_tts_provider_config) — set to 1 when STT_PROVIDER_SOURCE OR TTS_PROVIDER_SOURCE picks a speaches-* variant.
SPEACHES_IMAGE	speaches	ghcr.io/speaches-ai/speaches:0.9.0-rc.3-cpu	CPU build (default). When the speaches-gpu profile wins, the bootstrapper rewrites this var from SPEACHES_GPU_IMAGE at generation time.
SUPABASE_DB_SOURCE	supabase	container	Single-option (container only). Kept for forward-compat with legacy source-validator.
SUPABASE_DB_PORT	supabase	63012	Host port for supabase-db PostgreSQL (in-container 5432), bound to HOST_BIND_IP (loopback by default).
SUPABASE_DB_USER	supabase	supabase_admin	Postgres superuser name, passed to supabase-db as POSTGRES_USER and used by supabase-db-init.
SUPABASE_DB_PASSWORD	supabase	password	Supabase postgres admin password; auto-generated by the bootstrapper when left as the placeholder.
SUPABASE_DB_NAME	supabase	postgres	Database name, passed to supabase-db as POSTGRES_DB and used by supabase-db-init, pg-meta and Realtime.
SUPABASE_DB_NAME_URI	supabase	postgres	Bootstrapper-synchronized percent-encoded SUPABASE_DB_NAME for DSN path components.
SUPABASE_DB_APP_USER	supabase	app_user	Reserved for a future least-privilege application-tier database user; not created by init scripts today.
SUPABASE_DB_APP_PASSWORD	supabase	app_password	Reserved future app-user password; auto-generated by the bootstrapper when left as the placeholder.
SUPABASE_AUTH_DB_USER	supabase	supabase_auth_admin	Login name of the Supabase Auth (GoTrue) schema owner. supabase-db-init creates it if missing and sets its password.
SUPABASE_AUTH_DB_PASSWORD	supabase	atlas-db-password	Auto-generated login password for the upstream auth schema owner.
SUPABASE_AUTH_DB_USER_URI	supabase	supabase_auth_admin	Bootstrapper-synchronized percent-encoded SUPABASE_AUTH_DB_USER for DSN userinfo.
SUPABASE_AUTH_DB_PASSWORD_URI	supabase	atlas-db-password	Bootstrapper-synchronized percent-encoded SUPABASE_AUTH_DB_PASSWORD for DSN userinfo.
SUPABASE_STORAGE_DB_USER	supabase	supabase_storage_admin	Login name of the Supabase Storage schema owner. supabase-db-init creates it if missing and sets its password.
SUPABASE_STORAGE_DB_PASSWORD	supabase	atlas-db-password	Auto-generated login password for the upstream storage schema owner.
SUPABASE_STORAGE_DB_USER_URI	supabase	supabase_storage_admin	Bootstrapper-synchronized percent-encoded SUPABASE_STORAGE_DB_USER for DSN userinfo.
SUPABASE_STORAGE_DB_PASSWORD_URI	supabase	atlas-db-password	Bootstrapper-synchronized percent-encoded SUPABASE_STORAGE_DB_PASSWORD for DSN userinfo.
SUPABASE_API_DB_USER	supabase	authenticator	Login name of the PostgREST's role-switching authenticator. supabase-db-init creates it if missing and sets its password.
SUPABASE_API_DB_PASSWORD	supabase	atlas-db-password	Auto-generated login password for PostgREST's role-switching authenticator.
SUPABASE_API_DB_USER_URI	supabase	authenticator	Bootstrapper-synchronized percent-encoded SUPABASE_API_DB_USER for DSN userinfo.
SUPABASE_API_DB_PASSWORD_URI	supabase	atlas-db-password	Bootstrapper-synchronized percent-encoded SUPABASE_API_DB_PASSWORD for DSN userinfo.
SUPABASE_REALTIME_DB_USER	supabase	atlas_realtime	Login name of the replication-scoped Realtime role. supabase-db-init creates it if missing and sets its password.
SUPABASE_REALTIME_DB_PASSWORD	supabase	atlas-db-password	Auto-generated password for the replication-scoped Realtime role.
SUPABASE_META_DB_USER	supabase	atlas_meta	Login name of the dashboard metadata role used by pg-meta and Studio. supabase-db-init creates it if missing and sets its password.
SUPABASE_META_DB_PASSWORD	supabase	atlas-db-password	Auto-generated password for the dashboard metadata role used by Meta and Studio.
SUPABASE_META_CRYPTO_KEY	supabase		Auto-generated key shared by Studio (as PG_META_CRYPTO_KEY) and Postgres Meta (as CRYPTO_KEY) to encrypt the connection header Studio sends Meta. Empty falls back to the images' public SAMPLE_KEY constant.
SUPABASE_META_DB_USER_URI	supabase	atlas_meta	Bootstrapper-synchronized percent-encoded SUPABASE_META_DB_USER for DSN userinfo.
SUPABASE_META_DB_PASSWORD_URI	supabase	atlas-db-password	Bootstrapper-synchronized percent-encoded SUPABASE_META_DB_PASSWORD for DSN userinfo.
SUPABASE_STUDIO_DB_USER	supabase	atlas_studio_readonly	Read-only SQL-editor identity used by Studio; it shares Studio's single password interface with Meta but has no Meta admin authority.
SUPABASE_STUDIO_DB_USER_URI	supabase	atlas_studio_readonly	Bootstrapper-synchronized percent-encoded SUPABASE_STUDIO_DB_USER for Studio's constructed DSN.
SUPABASE_DB_INIT_SOURCE	supabase	container	Container only: the post-init SQL runner every Supabase-backed service waits for. Other values are rejected.
POSTGRES_EXPORTER_PORT	supabase	63013	Host port for postgres-exporter (in-container 9187). Sidecar gated by PROMETHEUS_SOURCE.
POSTGRES_EXPORTER_SCALE	supabase		Set by _generate_prometheus_config() to 1 when PROMETHEUS_SOURCE=container, 0 otherwise.
POSTGRES_EXPORTER_DB_USER	supabase	atlas_metrics	Login name of the pg_monitor-only role postgres-exporter uses. supabase-db-init creates it if missing and sets its password.
POSTGRES_EXPORTER_DB_PASSWORD	supabase	atlas-db-password	Auto-generated password for the pg_monitor-only exporter role.
POSTGRES_EXPORTER_DB_USER_URI	supabase	atlas_metrics	Bootstrapper-synchronized percent-encoded POSTGRES_EXPORTER_DB_USER for DSN userinfo.
POSTGRES_EXPORTER_DB_PASSWORD_URI	supabase	atlas-db-password	Bootstrapper-synchronized percent-encoded POSTGRES_EXPORTER_DB_PASSWORD for DSN userinfo.
SUPABASE_META_SOURCE	supabase	container	Deployment source for pg-meta (supabase-meta): container (default) or disabled.
SUPABASE_META_SCALE	supabase		Set by ./start.sh from SUPABASE_META_SOURCE; do not edit. 1 runs the container, 0 (disabled) does not.
SUPABASE_META_PORT	supabase	63014	Reserved port slot for pg-meta; not published on the host. Reach it through Kong at /pg/.
SUPABASE_STORAGE_SOURCE	supabase	container	Deployment source for Supabase Storage (supabase-storage): container (default) or disabled.
SUPABASE_STORAGE_SCALE	supabase		Set by ./start.sh from SUPABASE_STORAGE_SOURCE; do not edit. 1 runs the container, 0 (disabled) does not.
SUPABASE_STORAGE_PORT	supabase	63015	Host port for Supabase Storage (in-container 5000), bound to HOST_BIND_IP (loopback by default).
STORAGE_REGION	supabase	local	Region name passed to Supabase Storage as REGION.
STORAGE_FILE_SIZE_LIMIT	supabase	52428800	Maximum upload size in bytes for Supabase Storage (FILE_SIZE_LIMIT). The default is 50 MiB.
STORAGE_BACKEND	supabase	file	Supabase Storage backend. The default file stores objects in the storage volume at /var/lib/storage.
PROJECT_REF	supabase	default	Project reference passed to Supabase Storage as PROJECT_REF.
TENANT_ID	supabase	stub	Tenant id passed to Supabase Storage as TENANT_ID.
SUPABASE_AUTH_SOURCE	supabase	container	Deployment source for Supabase Auth, GoTrue (supabase-auth): container (default) or disabled.
SUPABASE_AUTH_SCALE	supabase		Set by ./start.sh from SUPABASE_AUTH_SOURCE; do not edit. 1 runs the container, 0 (disabled) does not.
SUPABASE_AUTH_PORT	supabase	63016	Reserved port slot for Supabase Auth (GoTrue); not published on the host. Reach it through Kong at /auth/v1.
SUPABASE_JWT_SECRET	supabase		Auto-generated by bootstrapper/generate_supabase_keys.py on first run.
SUPABASE_ANON_KEY	supabase		Auto-generated.
SUPABASE_SERVICE_KEY	supabase		Auto-generated.
SUPABASE_API_SOURCE	supabase	container	Deployment source for PostgREST (supabase-api): container (default) or disabled.
SUPABASE_API_SCALE	supabase		Set by ./start.sh from SUPABASE_API_SOURCE; do not edit. 1 runs the container, 0 (disabled) does not.
SUPABASE_API_PORT	supabase	63017	Host port for PostgREST (in-container 3000), bound to HOST_BIND_IP (loopback by default).
SUPABASE_REALTIME_SOURCE	supabase	container	Deployment source for Supabase Realtime (supabase-realtime): container (default) or disabled.
SUPABASE_REALTIME_SCALE	supabase		Set by ./start.sh from SUPABASE_REALTIME_SOURCE; do not edit. 1 runs the container, 0 (disabled) does not.
SUPABASE_REALTIME_PORT	supabase	63018	Host port for Supabase Realtime (in-container 4000), bound to HOST_BIND_IP (loopback by default).
SUPABASE_STUDIO_SOURCE	supabase	container	Deployment source for Supabase Studio (supabase-studio): container (default) or disabled.
SUPABASE_STUDIO_SCALE	supabase		Set by ./start.sh from SUPABASE_STUDIO_SOURCE; do not edit. 1 runs the container, 0 (disabled) does not.
SUPABASE_STUDIO_PORT	supabase	63019	Reserved port slot for Supabase Studio; not published on the host. Reach it through Kong.
SUPABASE_DB_IMAGE	supabase	supabase/postgres:17.6.1.139	Container image for supabase-db.
SUPABASE_DB_INIT_IMAGE	supabase	postgres:15.19-alpine	Container image for supabase-db-init.
SUPABASE_META_IMAGE	supabase	supabase/postgres-meta:v0.96.6	Container image for supabase-meta.
SUPABASE_STORAGE_IMAGE	supabase	supabase/storage-api:v1.61.5	Container image for supabase-storage.
SUPABASE_AUTH_IMAGE	supabase	supabase/gotrue:v2.191.0	Container image for supabase-auth.
SUPABASE_API_IMAGE	supabase	postgrest/postgrest:v12.2.10	Container image for supabase-api.
SUPABASE_REALTIME_IMAGE	supabase	supabase/realtime:v2.112.0	Container image for supabase-realtime.
SUPABASE_STUDIO_IMAGE	supabase	supabase/studio:2026.06.22-sha-2207d7f	Container image for supabase-studio.
POSTGRES_EXPORTER_IMAGE	supabase	prometheuscommunity/postgres-exporter:v0.19.1	Prometheus exporter sidecar. Scales 1↔0 with PROMETHEUS_SOURCE (gated by _generate_prometheus_config).
SUPAVISOR_DB_ADMIN_USER	supavisor	atlas_supavisor	Scoped owner for Supavisor's dedicated management database.
SUPAVISOR_DB_ADMIN_PASSWORD	supavisor	atlas-db-password	Auto-generated password for Supavisor management migrations and auth lookup.
SUPAVISOR_DB_ADMIN_USER_URI	supavisor	atlas_supavisor	Bootstrapper-synchronized percent-encoded SUPAVISOR_DB_ADMIN_USER for DSN userinfo.
SUPAVISOR_DB_ADMIN_PASSWORD_URI	supavisor	atlas-db-password	Bootstrapper-synchronized percent-encoded SUPAVISOR_DB_ADMIN_PASSWORD for DSN userinfo.
SUPAVISOR_SOURCE	supavisor	disabled	Deployment mode: container (transaction pooler) or disabled/direct Postgres.
SUPAVISOR_SCALE	supavisor		1 when SUPAVISOR_SOURCE=container, otherwise 0.
SUPAVISOR_TENANT_ID	supavisor	atlas	Supavisor tenant suffix. Clients connect as <db_user>.<tenant>; only ASCII letters, digits, underscores, and hyphens are accepted because dot is the upstream user/tenant delimiter.
SUPAVISOR_DEFAULT_POOL_SIZE	supavisor	20	Default upstream Postgres pool size for the Atlas tenant.
SUPAVISOR_MAX_CLIENT_CONN	supavisor	100	Maximum client connections accepted by the Atlas tenant.
SUPAVISOR_DB_POOL_SIZE	supavisor	5	Supavisor metadata database pool size.
SUPAVISOR_DATABASE_URL	supavisor		Generated Postgres URL consumed by backend/Celery; points at Supavisor when enabled and direct Supabase Postgres when disabled.
SUPAVISOR_DB_HOST	supavisor		Generated n8n Postgres host; supavisor when enabled, supabase-db when disabled.
SUPAVISOR_DB_PORT_VALUE	supavisor		Generated n8n Postgres port; 6543 when enabled, 5432 when disabled.
SUPAVISOR_DB_USER	supavisor		Generated n8n Postgres user; appends .${SUPAVISOR_TENANT_ID} when Supavisor is enabled.
SUPAVISOR_SECRET_KEY_BASE	supavisor		Supavisor Phoenix secret; auto-generated by the bootstrapper when blank.
SUPAVISOR_VAULT_ENC_KEY	supavisor		Supavisor vault encryption key; must be exactly 32 bytes and is auto-generated when blank.
SUPAVISOR_API_JWT_SECRET	supavisor		Supavisor tenant API JWT secret; auto-generated by the bootstrapper when blank.
SUPAVISOR_METRICS_JWT_SECRET	supavisor		Supavisor metrics JWT secret; auto-generated by the bootstrapper when blank.
SUPAVISOR_IMAGE	supavisor	supabase/supavisor:2.9.5	Container image for supavisor.
TEI_RERANKER_SOURCE	tei-reranker	disabled	Deployment source for the TEI reranker: container-cpu, container-gpu, localhost or disabled (default).
TEI_RERANKER_PORT	tei-reranker	63041	Host port for the TEI rerank API (in-container listen port is 80).
TEI_RERANKER_LOCALHOST_PORT	tei-reranker	63049	Host port for the host-installed TEI rerank source variant.
TEI_RERANKER_MODEL_ID	tei-reranker	mixedbread-ai/mxbai-rerank-base-v1	Hugging Face model id TEI serves (--model-id).
TEI_RERANKER_REVISION	tei-reranker	800f24c113213a187e65bde9db00c15a2bb12738	Model revision TEI loads (--revision). The default pins a commit of the default model.
TEI_RERANKER_MAX_CLIENT_BATCH_SIZE	tei-reranker	32	Maximum inputs per client request TEI accepts (--max-client-batch-size).
TEI_RERANKER_MEMORY_LIMIT	tei-reranker	4g	Container memory limit for tei-reranker (deploy.resources.limits.memory).
TEI_RERANKER_CPU_LIMIT	tei-reranker	2.0	Container CPU limit for tei-reranker (deploy.resources.limits.cpus).
TEI_RERANKER_HF_CACHE_DIR	tei-reranker	/data	In-container Hugging Face cache (HF_HOME). The default /data is the tei-reranker-data volume, so models persist across restarts.
TEI_RERANKER_SCALE	tei-reranker		Set by ./start.sh from TEI_RERANKER_SOURCE; do not edit. 1 runs tei-reranker for container-cpu and container-gpu, 0 otherwise.
TEI_RERANKER_ENDPOINT	tei-reranker		Resolved per TEI_RERANKER_SOURCE. Exposes TEI's rerank endpoint to compatible consumers.
TEI_RERANKER_IMAGE_RESOLVED	tei-reranker		Image picked per source variant (CPU vs GPU).
TEI_RERANKER_CPU_IMAGE	tei-reranker	ghcr.io/huggingface/text-embeddings-inference:cpu-1.9@sha256:ad950d30878eceb72aaf32024d26fa2b1d04a75304fa0b4776b49aa1941fea07	Used when TEI_RERANKER_SOURCE=container-cpu on amd64 hosts. Compiled with ORT backend (requires ONNX model weights). Digest-pinned: the cpu-1.9 family tag is republished per patch.
TEI_RERANKER_CPU_ARM64_IMAGE	tei-reranker	ghcr.io/huggingface/text-embeddings-inference:cpu-arm64-latest@sha256:35c50d7494de22deecdb783b8f5b7e1d05765709bd90071b03469b9440d28656	Used when TEI_RERANKER_SOURCE=container-cpu on arm64 hosts (Apple Silicon). Compiled with candle backend (loads safetensors natively, no ONNX required). Unversioned tag — HF doesn't publish versioned arm64 tags as of 2026-06-06.
TEI_RERANKER_GPU_IMAGE	tei-reranker	ghcr.io/huggingface/text-embeddings-inference:1.9@sha256:536efce2a0dc0acd0336d683ef1b81fcf900f76c8fb850e25a6d48a54a97df91	Used when TEI_RERANKER_SOURCE=container-gpu. CUDA host required. Digest-pinned: the 1.9 minor tag is republished per patch.
TEMPO_SOURCE	tempo	disabled	Deployment source for Grafana Tempo trace storage: container or disabled (default).
TEMPO_SCALE	tempo		Set by ./start.sh from TEMPO_SOURCE; do not edit. 1 runs tempo, 0 does not.
TEMPO_ENDPOINT	tempo		Internal Tempo query/API endpoint consumed by Grafana.
TEMPO_RETENTION_PERIOD	tempo	336h	Trace retention (Tempo block_retention, applied by its backend scheduler and worker). Tempo's own default, 14 days; lower it to bound the tempo-data volume.
TEMPO_IMAGE	tempo	grafana/tempo:3.0.0	Container image for tempo.
TIKA_SOURCE	tika	disabled	Deployment source for Apache Tika: container, tika-localhost or disabled (default).
TIKA_PORT	tika	63057	Host port for the Tika server (in-container 9998), bound to HOST_BIND_IP (loopback by default).
TIKA_LOCALHOST_PORT	tika	9998	Host port for the tika-localhost source variant.
TIKA_JAVA_TOOL_OPTIONS	tika	-Xmx768m	JVM memory cap for the Tika server container.
TIKA_MAX_FILE_SIZE	tika	52428800	Positive Backend maximum bytes accepted for Docling/Tika extraction; invalid values fail Backend startup.
TIKA_TIMEOUT_SECONDS	tika	30	Finite Backend timeout in seconds for Tika fallback extraction calls. Must be greater than 0 and no greater than 3600; invalid values fail Backend startup.
TIKA_SCALE	tika		Set by ./start.sh from TIKA_SOURCE; do not edit. 1 runs tika for container, 0 otherwise.
TIKA_ENDPOINT	tika		Resolved per TIKA_SOURCE: http://tika:9998, the tika-localhost URL, or empty when disabled. Read by backend, celery and n8n.
TIKA_IMAGE	tika	apache/tika:3.3.1.0	Container image for tika.
TRINO_SOURCE	trino	disabled	Deployment source for the Trino coordinator: container or disabled (default). container requires MinIO and the Iceberg REST catalog.
TRINO_PORT	trino	63029	Host port for the Trino coordinator UI/API (in-container 8080).
TRINO_SCALE	trino		Set by ./start.sh from TRINO_SOURCE; do not edit. 1 runs trino, 0 does not.
TRINO_IMAGE	trino	trinodb/trino:482	Pinned to the current Trino release tag validated against the Iceberg REST + MinIO catalog contract.
TRUEFORGE_SOURCE	trueforge	disabled	Deployment source for TrueForge (server, schedule controller and init): container or disabled (default).
TRUEFORGE_PORT	trueforge	63079	Host port for the TrueForge server UI + API (in-container 8790).
TRUEFORGE_ENDPOINT	trueforge		Resolved per TRUEFORGE_SOURCE — http://trueforge:8790 for container, empty for disabled.
TRUEFORGE_API_KEY	trueforge		Auto-generated by the bootstrapper on first run. Service credential for the controller's internal schedule-dispatch calls to the server.
TRUEFORGE_DB_NAME	trueforge	trueforge	Name of TrueForge's dedicated database, created by supabase-db-init.
TRUEFORGE_DB_NAME_URI	trueforge	trueforge	Bootstrapper-synchronized percent-encoded TRUEFORGE_DB_NAME for DSN path components.
TRUEFORGE_DB_USER	trueforge	trueforge	Login name of the owner of TrueForge's dedicated database, created by supabase-db-init.
TRUEFORGE_DB_PASSWORD	trueforge	atlas-db-password	Auto-generated password for the owner of TrueForge's dedicated database.
TRUEFORGE_DB_USER_URI	trueforge	trueforge	Bootstrapper-synchronized percent-encoded TRUEFORGE_DB_USER for DSN userinfo.
TRUEFORGE_DB_PASSWORD_URI	trueforge	atlas-db-password	Bootstrapper-synchronized percent-encoded TRUEFORGE_DB_PASSWORD for DSN userinfo.
TRUEFORGE_SCALE	trueforge		Set by ./start.sh from TRUEFORGE_SOURCE; do not edit. 1 runs the trueforge server, 0 does not.
TRUEFORGE_CONTROLLER_SCALE	trueforge		Set by ./start.sh from TRUEFORGE_SOURCE; do not edit. 1 runs trueforge-controller, 0 does not.
TRUEFORGE_INIT_SCALE	trueforge		Set by ./start.sh from TRUEFORGE_SOURCE; do not edit. 1 runs trueforge-init, 0 does not.
TRUEFORGE_IMAGE	trueforge	tfy.jfrog.io/tfy-images/trueforge:0.2.0-8b1d98e	Container image for trueforge.
TTS_PROVIDER_SOURCE	tts-provider	speaches-container-cpu	Text-to-speech engine: speaches-container-cpu (default), speaches-container-gpu, chatterbox-container-gpu, chatterbox-localhost or disabled.
TTS_PROVIDER_PORT	tts-provider	63058	Wizard display slot. Real engine port lives on its own container manifest.
TTS_PROVIDER_SCALE	tts-provider		Set by ./start.sh; do not edit. 1 when a container engine serves text-to-speech, else 0. The wizard uses it to colour the TTS row.
TTS_ENDPOINT	tts-provider		Resolved per TTS_PROVIDER_SOURCE. Read by n8n and jupyterhub; hermes-init receives it as TTS_INTERNAL_URL, and Open WebUI gets derived OPEN_WEB_UI_TTS_* values. The backend receives it but does not use it.
VERBA_SOURCE	verba	disabled	Deployment source for the Verba RAG UI: container or disabled (default).
VERBA_PORT	verba	63098	Host port for the Verba UI/API (in-container 8000).
VERBA_ENDPOINT	verba		In-network URL for Atlas services and notebooks.
VERBA_SCALE	verba		Set by ./start.sh from VERBA_SOURCE; do not edit. 1 runs verba, 0 does not.
VERBA_WEAVIATE_URL	verba		Resolved Weaviate URL passed to upstream WEAVIATE_URL_VERBA, which Verba reads only for its Weaviate (cloud) deployment; the default Docker deployment always dials weaviate:8080. For WEAVIATE_SOURCE=localhost pick Custom on Verba's connect screen.
VERBA_OPENAI_MODEL	verba		Optional Verba OpenAI generator model name. Empty uses LITELLM_DEFAULT_MODEL (Verba itself would otherwise start with an empty model name).
VERBA_OPENAI_EMBED_MODEL	verba		Optional Verba OpenAI-compatible embedding model name. Empty uses LITELLM_EMBEDDING_MODEL.
VERBA_DEFAULT_DEPLOYMENT	verba	Docker	Verba deployment selector. Atlas uses Docker to target the external Weaviate container.
VERBA_IMAGE	verba	semitechnologies/verba@sha256:0947d289ebff2c9814941c8d4282ee994dc79598e76162ae82e6efda4682b0b7	Upstream publishes semitechnologies/verba:latest but no v2.1.3 Docker tag; Atlas pins the observed multi-arch index digest.
VLLM_METAL_SOURCE	vllm-metal	disabled	Options: managed-localhost (bootstrapper-supervised host process on Apple silicon), disabled.
VLLM_METAL_MODEL	vllm-metal	Qwen/Qwen2.5-7B-Instruct	Hugging Face model id served by the managed vLLM Metal process and registered with LiteLLM under the same alias.
VLLM_METAL_LOCALHOST_PORT	vllm-metal	8000	Host port the managed vLLM Metal OpenAI server listens on. Not a BASE_PORT slot — vLLM Metal has no Kong route or exposed stack port.
VLLM_METAL_PLUGIN_VERSION	vllm-metal	0.3.0.dev20260713103604	Atlas-verified vllm-metal GitHub release wheel installed with checksum verification; unverified overrides are rejected.
VLLM_METAL_CORE_VERSION	vllm-metal	0.24.0	Atlas-verified compatible vllm core source release; it must match the supported plugin release.
VLLM_METAL_PYTHON	vllm-metal	python3.12	Python interpreter used to build the managed host venv (vllm-metal requires 3.12).
VLLM_METAL_STATE_DIR	vllm-metal	~/.atlas/vllm-metal	Host directory holding the managed venv plus the pid / log / status files.
VLLM_METAL_MODELS_PATH	vllm-metal		Optional Hugging Face cache dir (HF_HOME) for the managed process. Blank = default HF cache.
VLLM_METAL_MIN_MEMORY_GB	vllm-metal	16	Unified-memory warning floor (GB) checked before lifecycle work; memory below it warns and unreadable memory skips the check. The result does not block install or start and does not certify model fit or prevent OOM.
VLLM_METAL_SCALE	vllm-metal		Always 0 — vLLM Metal never runs as a container. Reserved for interface parity with container services.
VLLM_METAL_ENDPOINT	vllm-metal		Resolved docker-internal URL of the managed host process (http://host.docker.internal:${VLLM_METAL_LOCALHOST_PORT}). Consumed by litellm-init to register the model. Blank when disabled.
WEAVIATE_SOURCE	weaviate	container	Deployment source for Weaviate: container (default), localhost or disabled.
WEAVIATE_PORT	weaviate	63030	Host port for the Weaviate REST API (in-container 8080), bound to HOST_BIND_IP (loopback by default).
WEAVIATE_GRPC_PORT	weaviate	63031	Host port for the Weaviate gRPC API (in-container 50051), bound to HOST_BIND_IP (loopback by default).
WEAVIATE_INIT_SOURCE	weaviate	container	Auto-managed by weaviate. Container/disabled.
MULTI2VEC_CLIP_SOURCE	weaviate	container-cpu	Options: container-cpu, container-gpu, disabled. container-gpu sets ENABLE_CUDA=1, but Compose reserves no GPU device yet. The CLIP container then fails to start; use container-cpu.
WEAVIATE_ENABLE_MODULES	weaviate	text2vec-openai,text2vec-ollama,multi2vec-clip,generative-openai,generative-ollama,backup-filesystem	Weaviate module list (ENABLE_MODULES). ./start.sh adds or removes multi2vec-clip to match MULTI2VEC_CLIP_SOURCE and keeps any other modules you list.
CLIP_INFERENCE_API	weaviate	http://multi2vec-clip:8080	URL Weaviate uses for the multi2vec-clip module. ./start.sh empties it when MULTI2VEC_CLIP_SOURCE=disabled.
MULTI2VEC_CLIP_SIGLIP2_IMAGE	weaviate	semitechnologies/multi2vec-clip:google-siglip2-so400m-patch16-512-1.5.1	Opt-in SigLIP 2 image reference for MULTI2VEC_CLIP_IMAGE. Produces 1152-d vectors; revectorize or create new Weaviate collections before switching.
WEAVIATE_URL	weaviate		Resolved per WEAVIATE_SOURCE. Consumed by n8n, backend, jupyterhub.
WEAVIATE_LITELLM_API_KEY	weaviate		Mirror of LITELLM_MASTER_KEY for Weaviate's vectorizer auth.
WEAVIATE_SCALE	weaviate		Set by ./start.sh from WEAVIATE_SOURCE; do not edit. 1 runs weaviate for container, 0 otherwise.
WEAVIATE_INIT_SCALE	weaviate		Set by ./start.sh from WEAVIATE_SOURCE; do not edit. 1 runs weaviate-init with a container Weaviate, 0 otherwise.
CLIP_SCALE	weaviate		Computed from MULTI2VEC_CLIP_SOURCE.
CLIP_ENABLE_CUDA	weaviate		Set by ./start.sh from MULTI2VEC_CLIP_SOURCE; do not edit. Passed to multi2vec-clip as ENABLE_CUDA. container-gpu sets 1, but compose requests no GPU device.
CLIP_DEPLOY_RESOURCES	weaviate	~	Set by ./start.sh from the multi2vec-clip deploy slice; ~ means none. No compose file reads it.
WEAVIATE_LOCALHOST_PORT	weaviate	8080	Host port for the weaviate-localhost source variant. URL is derived at compose-render time as http://host.docker.internal:8080.
WEAVIATE_MEMORY_LIMIT	weaviate	2g	Container memory limit for the weaviate container (deploy.resources.limits.memory). The 2g default is conservative for a 32 GB host. It fits the Go data plane and in-memory HNSW index for small-to-medium collections.
WEAVIATE_CPU_LIMIT	weaviate	1.5	Container CPU limit for the weaviate container (deploy.resources.limits.cpus). 1.5 cores covers routine indexing and query serving on a shared host.
WEAVIATE_IMAGE	weaviate	cr.weaviate.io/semitechnologies/weaviate:1.38.17	Container image for weaviate.
WEAVIATE_INIT_IMAGE	weaviate	alpine:3.24.2	Container image for weaviate-init.
MULTI2VEC_CLIP_IMAGE	weaviate	semitechnologies/multi2vec-clip:sentence-transformers-clip-ViT-B-32-1.5.1	Container image for multi2vec-clip.
ZEPPELIN_SPARK_CORES_MAX	zeppelin	1	Cores the long-lived Zeppelin Spark interpreter may hold (spark.cores.max, re-seeded each start). Uncapped, it takes every free worker core and leaves Airflow SparkSubmit drivers waiting. Raise it only with spare cores beyond Spark Connect and Airflow.
ZEPPELIN_DB_USER	zeppelin	atlas_zeppelin	Read-only PostgreSQL login exposed to Zeppelin's JDBC interpreter.
ZEPPELIN_DB_PASSWORD	zeppelin	atlas-db-password	Auto-generated password for Zeppelin's read-only database role.
ZEPPELIN_SOURCE	zeppelin	disabled	Deployment source for Zeppelin: container or disabled (default). container requires SPARK_SOURCE=container.
ZEPPELIN_PORT	zeppelin	63099	Host port for the Zeppelin Web UI (in-container 8080).
ZEPPELIN_SCALE	zeppelin		Set by ./start.sh from ZEPPELIN_SOURCE; do not edit. 1 runs zeppelin, 0 does not.
ZEPPELIN_INIT_SCALE	zeppelin		Set by ./start.sh from ZEPPELIN_SOURCE; do not edit. 1 runs zeppelin-init, 0 does not.
ZEPPELIN_MEMORY_LIMIT	zeppelin	2g	Container memory limit for the zeppelin container (deploy.resources.limits.memory). Zeppelin's JVM-based interpreter host typically uses under 1 g for a single-user session; 2 g leaves comfortable headroom on a 32 GB host.
ZEPPELIN_CPU_LIMIT	zeppelin	1.5	Container CPU limit for the zeppelin container (deploy.resources.limits.cpus). 1.5 cores covers notebook rendering and interpreter dispatch. Compute-heavy work runs on the remote Spark cluster, not in this process.
ZEPPELIN_IMAGE	zeppelin	apache/zeppelin:0.12.1	Container image for zeppelin.
ZEPPELIN_INIT_IMAGE	zeppelin	python:3.12.13-alpine	Container image for zeppelin-init.