Environment Variables Reference

Complete reference for all Sibyl environment variables.

Configuration Loading

Sibyl uses Pydantic Settings to load configuration from the process environment:

Environment variables (highest priority)
Explicit deployment env files loaded by the launcher (docker compose --env-file, systemd EnvironmentFile, Kubernetes secrets, etc.)
Default values

Local development does not read repo .env files. Use shell exports, the web onboarding UI, or an explicit deployment env file.

All variables use the SIBYL_ prefix. Some common variables (API keys) also support unprefixed versions as fallbacks.

Server Configuration

Variable	Default	Description
`SIBYL_ENVIRONMENT`	`development`	Runtime environment: development/staging/production
`SIBYL_SERVER_NAME`	`sibyl`	MCP server name
`SIBYL_SERVER_HOST`	`localhost`	Server bind host
`SIBYL_SERVER_PORT`	`3334`	Server bind port
`SIBYL_LOG_LEVEL`	`INFO`	Logging level: DEBUG/INFO/WARNING/ERROR

Storage Mode

Variable	Default	Description
`SIBYL_STORE`	`surreal`	Active persistence runtime
`SIBYL_AUTH_STORE`	`surreal`	Auth persistence. Only `surreal` is supported
`SIBYL_COORDINATION_BACKEND`	`auto`	Jobs, locks, pub/sub: `auto`, `local`, or `redis`

auto resolves to local in-process coordination for the default Surreal runtime. Use redis for multi-pod deployments. See storage-modes.md for the full mode matrix.

SurrealDB

SurrealDB is the default and only runtime store. These settings apply to every Sibyl process.

Variable	Default	Description
`SIBYL_SURREAL_URL`	(empty)	Connection URL (`ws://`, `http://`, `surrealkv://`, `memory://`)
`SIBYL_SURREAL_DATA_DIR`	(empty)	Local SurrealKV path used when `SIBYL_SURREAL_URL` is unset
`SIBYL_SURREAL_USERNAME`	(empty)	Root username for remote runtimes
`SIBYL_SURREAL_PASSWORD`	(empty)	Root password for remote runtimes
`SIBYL_SURREAL_TOKEN`	(empty)	Bearer token for remote runtimes (alternative to username/password)
`SIBYL_SURREAL_NAMESPACE_PREFIX`	`org_`	Namespace prefix for per-org isolation (`org_<uuid_hex>`)
`SIBYL_SURREAL_DATABASE`	`graph`	Database name inside each org namespace
`SIBYL_SURREAL_SLOW_QUERY_MS`	`500`	Log SurrealDB queries at warning level when elapsed time exceeds this

Connection Pooling

Variable	Default	Description
`SIBYL_SURREAL_POOL_SIZE`	`8`	Concurrent connections per dedicated client (1-256)
`SIBYL_SURREAL_AUTH_POOL_SIZE`	(pool size)	Override for the auth client pool
`SIBYL_SURREAL_CONTENT_POOL_SIZE`	(pool size)	Override for the content client pool
`SIBYL_SURREAL_GRAPH_POOL_SIZE`	(pool size)	Override for org graph client pools
`SIBYL_SURREAL_GRAPH_CLIENT_CACHE_SIZE`	`64`	Org-scoped graph clients kept open per process (LRU)
`SIBYL_ALLOW_EMBEDDED_SINGLE_WRITER`	`false`	Allow embedded (`surrealkv://`) storage in production

SIBYL_SURREAL_URL and SIBYL_SURREAL_DATA_DIR are mutually exclusive; set only one. When neither is set, Sibyl falls back to in-memory mode. In-memory mode (memory://) is rejected in production.

Embedded (surrealkv://) and in-memory URLs clamp every pool to a single connection because the storage is single-writer. Running embedded storage in production additionally requires the explicit SIBYL_ALLOW_EMBEDDED_SINGLE_WRITER=1 opt-in, and only when one daemon owns the database; otherwise startup fails validation.

URL Configuration

Variable	Default	Description
`SIBYL_PUBLIC_URL`	`http://localhost:3337`	Public base URL for OAuth callbacks, redirects
`SIBYL_SERVER_URL`	(derived from public_url)	API base URL override
`SIBYL_FRONTEND_URL`	(derived from public_url)	Frontend base URL override

When using Kong or similar ingress, SIBYL_PUBLIC_URL is typically set to the external domain (e.g., https://sibyl.example.com), and both API and frontend are served from the same origin.

Authentication

Variable	Default	Description
`SIBYL_JWT_SECRET`	(dev auto)	JWT signing secret, required in production
`SIBYL_JWT_ALGORITHM`	`HS256`	JWT signing algorithm
`SIBYL_ACCESS_TOKEN_EXPIRE_MINUTES`	`60`	Access token TTL in minutes
`SIBYL_REFRESH_TOKEN_EXPIRE_DAYS`	`30`	Local-auth refresh token TTL in days
`SIBYL_DISABLE_AUTH`	`false`	Disable auth enforcement (dev only)
`SIBYL_MCP_AUTH_MODE`	`auto`	MCP auth: auto/on/off
`SIBYL_SETTINGS_KEY`	(auto)	Fernet key for encrypting DB-stored secrets
`SIBYL_LOCAL_AUTH_ENABLED`	`false`	Enable local username/password login after setup
`SIBYL_PUBLIC_SIGNUPS_ENABLED`	`false`	Allow public self-serve account creation after setup
`SIBYL_OIDC`	`{}`	JSON object for optional OIDC providers and session UX
`SIBYL_BREAK_GLASS_ENABLED`	`false`	Enable bounded emergency local login for SSO outages
`SIBYL_BREAK_GLASS_ALLOWED_IPS`	`[]`	JSON array of CIDRs allowed to use break-glass login
`SIBYL_BREAK_GLASS_EXPIRES_AT`	(empty)	UTC expiry for break-glass, no more than four hours out

SIBYL_LOCAL_AUTH_ENABLED defaults to false and is auto-enabled only when SIBYL_ENVIRONMENT=development and the variable is unset. Production deployments that want local username/password login must set it explicitly. The Helm chart (auth.localAuthEnabled: true) and the Ansible self-host stack enable it by default; docker-compose.prod.yml leaves it false. The local-first single-user flow is otherwise unchanged: the first setup signup creates the owner/admin user, and account creation after setup is invite-based unless SIBYL_PUBLIC_SIGNUPS_ENABLED=true.

OIDC, silent refresh, extra OAuth providers, public signups, disabled local auth, and break-glass are all opt-in. Enterprise SSO deployments should configure a corporate OIDC provider first, verify an owner can sign in through it, and only then set SIBYL_LOCAL_AUTH_ENABLED=false.

SIBYL_OIDC is a JSON object with these fields:

json

{
  "providers": [
    {
      "name": "entra",
      "issuer": "https://login.microsoftonline.com/<tenant-id>/v2.0",
      "client_id": "<app-client-id>",
      "client_secret_env": "SIBYL_OIDC_ENTRA_CLIENT_SECRET",
      "organization_slug": "acme",
      "scopes": ["openid", "profile", "email"]
    }
  ],
  "role_claim": "roles",
  "redirect_uri_base": "",
  "session_minutes": 60,
  "silent_refresh_enabled": false,
  "extra_providers_enabled": false
}

Every provider requires organization_slug. It binds that provider to one exact non-personal organization; OIDC login never chooses an organization from the user's other memberships.

Non-corporate providers such as GitHub or Google require "extra_providers_enabled": true; leave it false for enterprise SSO.

Fallback Variables

These unprefixed variables are checked if SIBYL_* versions are empty:

JWT_SECRET -> SIBYL_JWT_SECRET

Security Warning

bash

# NEVER set disable_auth in production!
# This validation is enforced:
if environment == "production" and disable_auth:
    raise ValueError("disable_auth=True is forbidden in production")

GitHub OAuth

Variable	Default	Description
`SIBYL_GITHUB_CLIENT_ID`	(empty)	GitHub OAuth application ID
`SIBYL_GITHUB_CLIENT_SECRET`	(empty)	GitHub OAuth application secret

Fallbacks:

GITHUB_CLIENT_ID -> SIBYL_GITHUB_CLIENT_ID
GITHUB_CLIENT_SECRET -> SIBYL_GITHUB_CLIENT_SECRET

Variable	Default	Description
`SIBYL_COOKIE_DOMAIN`	(none)	Cookie domain override
`SIBYL_COOKIE_SECURE`	(auto)	Force Secure cookies (auto-detects from URL)

Password Hashing

Variable	Default	Description
`SIBYL_PASSWORD_PEPPER`	(empty)	Optional pepper for password hashing
`SIBYL_PASSWORD_ITERATIONS`	`310000`	PBKDF2-HMAC-SHA256 iterations

Rate Limiting

Variable	Default	Description
`SIBYL_RATE_LIMIT_ENABLED`	`true`	Enable rate limiting
`SIBYL_RATE_LIMIT_DEFAULT`	`100/minute`	Default rate limit
`SIBYL_RATE_LIMIT_STORAGE`	`memory://`	Storage backend (memory:// or redis://)

PostgreSQL (migration/rehearsal only)

PostgreSQL is not part of the Sibyl runtime and is not read by the Settings model. It is consumed only by the standalone migrate CLI when explicitly restoring a retained postgres.sql payload against an operator-managed PostgreSQL database. Structured auth and content archive export reads SurrealDB. Remove any stale SIBYL_AUTH_STORE=postgres value before starting the API.

The variables below are not SIBYL_ Settings fields. They are read directly by the migration tooling for a rehearsal database connection and are ignored by the running API and worker.

Variable	Default	Description
`SIBYL_POSTGRES_HOST`	`localhost`	External rehearsal database host
`SIBYL_POSTGRES_PORT`	`5433`	External rehearsal database port
`SIBYL_POSTGRES_USER`	`sibyl`	External rehearsal database username
`SIBYL_POSTGRES_PASSWORD`	`sibyl_dev`	External rehearsal database password
`SIBYL_POSTGRES_DB`	`sibyl`	External rehearsal database name
`SIBYL_POSTGRES_POOL_SIZE`	`10`	External rehearsal database connection pool
`SIBYL_POSTGRES_MAX_OVERFLOW`	`20`	External rehearsal database overflow limit

Configure them only when running a migration that explicitly restores a retained postgres.sql payload. They have no effect on the default Surreal runtime.

Redis/Valkey Coordination

Redis/Valkey is optional. The default Surreal runtime uses local in-process coordination.

Variable	Default	Description
`SIBYL_REDIS_HOST`	`127.0.0.1`	Redis/Valkey host
`SIBYL_REDIS_PORT`	`6381`	Redis/Valkey port
`SIBYL_REDIS_PASSWORD`	-	Redis/Valkey password
`SIBYL_REDIS_JOBS_DB`	`1`	Redis DB for job queue

LLM Configuration

Variable	Default	Description
`SIBYL_LLM_PROVIDER`	`anthropic`	LLM provider: openai or anthropic
`SIBYL_LLM_MODEL`	`claude-haiku-4-5`	LLM model for entity extraction

Embeddings

Document chunk embeddings and graph node/relationship embeddings are configured separately. The graph embedding dimensions also size the native Surreal vector indexes.

Variable	Default	Description
`SIBYL_EMBEDDING_PROVIDER`	`openai`	Document chunk embedding provider: openai or gemini
`SIBYL_EMBEDDING_MODEL`	`text-embedding-3-small`	Document chunk embedding model
`SIBYL_EMBEDDING_DIMENSIONS`	`1536`	Document chunk embedding vector dimensions
`SIBYL_GRAPH_EMBEDDING_PROVIDER`	`openai`	Graph embedding provider: openai, gemini, or local
`SIBYL_GRAPH_EMBEDDING_MODEL`	`text-embedding-3-small`	Graph node/relationship embedding model
`SIBYL_GRAPH_EMBEDDING_DIMENSIONS`	`1024`	Graph embedding dimensions (sizes vector indexes)

The local graph embedding provider runs sentence-transformers models in-process with no API key. When SIBYL_GRAPH_EMBEDDING_PROVIDER=local and no model is set, the model defaults to sentence-transformers/all-MiniLM-L6-v2 and the dimensions are derived from the model. Document chunk embeddings (SIBYL_EMBEDDING_PROVIDER) support openai and gemini only.

Retrieval Tuning

Vector index and reranking knobs. Changing HNSW parameters affects newly built indexes.

Variable	Default	Description
`SIBYL_GRAPH_HNSW_EFC`	`150`	Surreal HNSW graph index EF-construction value
`SIBYL_GRAPH_HNSW_M`	`12`	Surreal HNSW index max connections per element
`SIBYL_GRAPH_KNN_EF`	`40`	Surreal KNN query effort for graph vector retrieval
`SIBYL_RERANK_ENABLED`	`false`	Cross-encoder reranking after RRF fusion
`SIBYL_RERANK_MODEL`	`cross-encoder/ms-marco-MiniLM-L-6-v2`	Cross-encoder model for reranking
`SIBYL_RERANK_TOP_K`	`20`	Top candidates to rerank; the rest pass through (1-100)

Reranking requires the optional reranking extra (sentence-transformers). When the extra is not installed, the path degrades cleanly to the fused order instead of raising.

API Keys

Variable	Default	Description
`SIBYL_OPENAI_API_KEY`	(empty)	OpenAI API key (required for embeddings)
`SIBYL_ANTHROPIC_API_KEY`	(empty)	Anthropic API key
`SIBYL_GEMINI_API_KEY`	(empty)	Gemini API key (for Google embeddings)

Lookup Priority

API keys are resolved in this order:

Database - Keys stored via web UI (Settings, AI Services)
Environment variables - SIBYL_OPENAI_API_KEY, SIBYL_ANTHROPIC_API_KEY, SIBYL_GEMINI_API_KEY
Unprefixed fallbacks - OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY / GOOGLE_API_KEY

This allows zero-config deployments where API keys are entered through the onboarding wizard and stored encrypted in the database (using SIBYL_SETTINGS_KEY).

Unprefixed Fallbacks

OPENAI_API_KEY -> SIBYL_OPENAI_API_KEY
ANTHROPIC_API_KEY -> SIBYL_ANTHROPIC_API_KEY
GEMINI_API_KEY or GOOGLE_API_KEY -> SIBYL_GEMINI_API_KEY

Native Memory Configuration

Variable	Default	Description
`SIBYL_NATIVE_WRITE`	`enabled`	Set `disabled` to skip persisting reflection candidates
`SIBYL_AUTO_EXTRACT_ENTITIES`	`false`	Queue LLM entity extraction for prose-bearing memories
`SIBYL_OPERATIONAL_NOTE_DISTILLATION_MAX_TOKENS`	`2048`	Max output tokens per note distillation call (256-8192)
`SIBYL_RAW_CAPTURE_CHANGEFEED_POLL_ENABLED`	`true`	Poll the raw-captures changefeed for incremental promotion
`SIBYL_RAW_CAPTURE_LIVE_QUERY_ENABLED`	`false`	Use SurrealDB live queries for realtime promotion hints
`SIBYL_RAW_CAPTURE_LIVE_QUERY_RETRY_SECONDS`	`5.0`	Delay before reconnecting the raw-capture live query

Runtime Telemetry

Variable	Default	Description
`SIBYL_METRICS_SCRAPE_TOKEN`	(empty)	Bearer/header token for non-local `/metrics` scraping

Email

Sibyl sends through SMTP when SIBYL_SMTP_HOST is set, otherwise through Resend when SIBYL_RESEND_API_KEY is set. The JSONL outbox writes regardless of live delivery provider. When setting SMTP passwords in a Compose .env file, escape literal $ characters as $$.

Variable	Default	Description
`SIBYL_RESEND_API_KEY`	(empty)	Resend API key for transactional email
`SIBYL_SMTP_HOST`	(empty)	SMTP host for transactional email
`SIBYL_SMTP_PORT`	`587`	SMTP port
`SIBYL_SMTP_USERNAME`	(empty)	SMTP authentication username
`SIBYL_SMTP_PASSWORD`	(empty)	SMTP authentication password
`SIBYL_SMTP_STARTTLS`	`true`	Upgrade SMTP connection with STARTTLS
`SIBYL_SMTP_SSL`	`false`	Use implicit TLS instead of STARTTLS
`SIBYL_SMTP_TIMEOUT_SECONDS`	`20`	SMTP connection timeout
`SIBYL_EMAIL_FROM`	`Sibyl <noreply@sibyl.dev>`	Default from address
`SIBYL_EMAIL_OUTBOX_PATH`	(empty)	Optional JSONL outbox path for local/staging capture

Content Ingestion

Variable	Default	Description
`SIBYL_CHUNK_MAX_TOKENS`	`1000`	Maximum tokens per chunk
`SIBYL_CHUNK_OVERLAP_TOKENS`	`100`	Token overlap between chunks
`SIBYL_SOURCE_IMPORT_DIR`	`./source-imports`	Directory of local source archives API imports may read

Backups

Scheduled archive backups run from the worker. See Monitoring for operational detail.

Variable	Default	Description
`SIBYL_BACKUP_ENABLED`	`true`	Enable scheduled automatic backups
`SIBYL_BACKUP_DIR`	`./backups`	Directory to store backup archives
`SIBYL_BACKUP_RETENTION_DAYS`	`30`	Days to retain backups before auto-cleanup
`SIBYL_BACKUP_SCHEDULE`	`0 2 * * *`	Cron schedule for automatic backups (2 AM daily)

Worker Configuration

Variable	Default	Description
`SIBYL_RUN_WORKER`	`false`	Embed a worker in the API process when Redis coordination is active
`SIBYL_WORKER_MAX_JOBS`	(auto)	Override maximum concurrent background jobs (1-1024)

When SIBYL_WORKER_MAX_JOBS is unset, the limit is derived from CPU count (2x cores, minimum 3) capped by the effective content-client pool size.

Example Environment Blocks

Local Development Shell

bash

export SIBYL_ENVIRONMENT=development

# Recommended local runtime
export SIBYL_STORE=surreal
export SIBYL_COORDINATION_BACKEND=local
export SIBYL_SURREAL_URL=ws://127.0.0.1:8000/rpc
export SIBYL_SURREAL_USERNAME=root
export SIBYL_SURREAL_PASSWORD=root

# LLM
export SIBYL_OPENAI_API_KEY=sk-...
export SIBYL_ANTHROPIC_API_KEY=sk-ant-...

# Logging
export SIBYL_LOG_LEVEL=DEBUG

Production Env File (Surreal, Default)

bash

SIBYL_ENVIRONMENT=production
SIBYL_JWT_SECRET=<generate with: openssl rand -hex 32>
SIBYL_SETTINGS_KEY=<generate with: openssl rand -base64 32 | tr '+/' '-_'>

# Public URL (Kong/ingress domain)
SIBYL_PUBLIC_URL=https://sibyl.example.com

# Storage (fully Surreal)
SIBYL_STORE=surreal
SIBYL_AUTH_STORE=surreal
SIBYL_SURREAL_URL=ws://prod-surrealdb.internal:8000/rpc
SIBYL_SURREAL_USERNAME=root
SIBYL_SURREAL_PASSWORD=<secure-password>

# LLM
SIBYL_OPENAI_API_KEY=sk-...
SIBYL_ANTHROPIC_API_KEY=sk-ant-...
SIBYL_LLM_PROVIDER=anthropic
SIBYL_LLM_MODEL=claude-sonnet-4

# Email
SIBYL_SMTP_HOST=smtp.gmail.com
SIBYL_SMTP_PORT=587
SIBYL_SMTP_USERNAME=sibyl@example.com
SIBYL_SMTP_PASSWORD=<google-app-password>
SIBYL_SMTP_STARTTLS=true
SIBYL_EMAIL_FROM=Sibyl <sibyl@example.com>

Migration Archive Rehearsal

Use PostgreSQL settings only when explicitly restoring or validating a retained postgres.sql payload. New production deployments should use the fully Surreal example above.

bash

SIBYL_ENVIRONMENT=production
SIBYL_JWT_SECRET=<generate with: openssl rand -hex 32>
SIBYL_SETTINGS_KEY=<generate with: openssl rand -base64 32 | tr '+/' '-_'>
SIBYL_PUBLIC_URL=https://sibyl.example.com

# Surreal target
SIBYL_SURREAL_URL=ws://prod-surrealdb.internal:8000/rpc
SIBYL_SURREAL_USERNAME=root
SIBYL_SURREAL_PASSWORD=<secure-password>

# Optional historical archive rehearsal database
SIBYL_POSTGRES_HOST=prod-postgres.internal
SIBYL_POSTGRES_PORT=5433
SIBYL_POSTGRES_PASSWORD=<secure-password>

# LLM
SIBYL_OPENAI_API_KEY=sk-...
SIBYL_ANTHROPIC_API_KEY=sk-ant-...

# Rate limiting with Redis
SIBYL_RATE_LIMIT_STORAGE=redis://prod-redis.internal:6379

Kubernetes ConfigMap

Non-secret environment variables in ConfigMap:

yaml

apiVersion: v1
kind: ConfigMap
metadata:
  name: sibyl-config
  namespace: sibyl
data:
  SIBYL_ENVIRONMENT: "production"
  SIBYL_SERVER_HOST: "0.0.0.0"
  SIBYL_SERVER_PORT: "3334"
  SIBYL_PUBLIC_URL: "https://sibyl.example.com"
  SIBYL_LLM_PROVIDER: "anthropic"
  SIBYL_LLM_MODEL: "claude-haiku-4-5"
  SIBYL_EMBEDDING_MODEL: "text-embedding-3-small"
  SIBYL_EMBEDDING_DIMENSIONS: "1536"

Kubernetes Secret

Sensitive values in Secret:

yaml

apiVersion: v1
kind: Secret
metadata:
  name: sibyl-secrets
  namespace: sibyl
type: Opaque
stringData:
  SIBYL_JWT_SECRET: "<jwt-secret>"
  SIBYL_SETTINGS_KEY: "<fernet-key>" # Generate with: python -c "from cryptography.fernet import Fernet; print(Fernet.generate_key().decode())"
  SIBYL_OPENAI_API_KEY: "sk-..." # Optional if using DB-stored keys
  SIBYL_ANTHROPIC_API_KEY: "sk-ant-..." # Optional if using DB-stored keys
  SIBYL_SURREAL_PASSWORD: "<surreal-password>" # For surreal mode
  # Migration/archive only:
  # SIBYL_POSTGRES_PASSWORD: "<db-password>"

Running Multiple Instances

You can run multiple Sibyl instances on the same machine (e.g., dev + test environments) by configuring different ports and container names.

Port Configuration

Note: SIBYL_WEB_PORT is a docker-compose-level variable used only for port mapping in docker-compose.yml. It is not consumed by Pydantic Settings or the Python application.

Variable	Default	Description
`SIBYL_SERVER_PORT`	`3334`	API/MCP server port
`SIBYL_WEB_PORT`	`3337`	Web frontend port
`SIBYL_SURREAL_PORT`	`8000`	SurrealDB port (default store)
`SIBYL_BACKEND_URL`	(auto)	Backend URL for web app

Quick Setup: Test Instance

Export offset ports for this shell:

bash

export COMPOSE_PROJECT_NAME=sibyl-test
export SIBYL_SERVER_PORT=3344
export SIBYL_WEB_PORT=3347
export SIBYL_SURREAL_PORT=8010
export SIBYL_SURREAL_URL=ws://127.0.0.1:8010/rpc

Start databases with isolated containers and volumes:

bash

docker compose --env-file /dev/null -p "$COMPOSE_PROJECT_NAME" up -d

Start API pointing to test databases:

bash

sibyld serve

Start web frontend:

bash

SIBYL_WEB_PORT=3347 SIBYL_BACKEND_URL=http://localhost:3344 pnpm -C apps/web dev

How It Works

COMPOSE_PROJECT_NAME isolates Docker containers and volumes
Each port variable controls the corresponding service
SIBYL_BACKEND_URL tells the web frontend where to proxy API requests

Tips

Use docker compose --env-file /dev/null -p sibyl-test ps to see test instance containers
Local Surreal data directories are namespaced by project when you override SURREAL_DATA_DIR
CLI contexts let you switch between instances: sibyl config context use test

Computed Properties

The Settings class exposes computed connection URLs and runtime-shape helpers:

python

settings.resolved_surreal_url  # ws://..., surrealkv://..., or memory://
settings.redis_url             # redis:// URL assembled from host/port/password
settings.fully_surreal         # always True; graph, content, and auth are SurrealDB
settings.resolved_coordination_backend  # resolves "auto" to "local" or "redis"

Sibyl is fully SurrealDB-backed, so fully_surreal always returns True. There are no PostgreSQL connection helpers or relational-shape flags on Settings (postgres_url, postgres_url_sync, uses_relational_auth, and requires_relational_support do not exist). PostgreSQL connection details live with the migration tooling described under PostgreSQL (migration/rehearsal only).

Environment Variables Reference ​

Configuration Loading ​

Server Configuration ​

Storage Mode ​

SurrealDB ​

Connection Pooling ​

URL Configuration ​

Authentication ​

Fallback Variables ​

Security Warning ​

GitHub OAuth ​

Cookie Configuration ​

Password Hashing ​

Rate Limiting ​

PostgreSQL (migration/rehearsal only) ​

Redis/Valkey Coordination ​

LLM Configuration ​

Embeddings ​

Retrieval Tuning ​

API Keys ​

Lookup Priority ​

Unprefixed Fallbacks ​

Native Memory Configuration ​

Runtime Telemetry ​

Email ​

Content Ingestion ​

Backups ​

Worker Configuration ​

Example Environment Blocks ​

Local Development Shell ​

Production Env File (Surreal, Default) ​

Migration Archive Rehearsal ​

Kubernetes ConfigMap ​

Kubernetes Secret ​

Running Multiple Instances ​

Port Configuration ​

Quick Setup: Test Instance ​

How It Works ​

Tips ​

Computed Properties ​

Environment Variables Reference

Configuration Loading

Server Configuration

Storage Mode

SurrealDB

Connection Pooling

URL Configuration

Authentication

Fallback Variables

Security Warning

GitHub OAuth

Cookie Configuration

Password Hashing

Rate Limiting

PostgreSQL (migration/rehearsal only)

Redis/Valkey Coordination

LLM Configuration

Embeddings

Retrieval Tuning

API Keys

Lookup Priority

Unprefixed Fallbacks

Native Memory Configuration

Runtime Telemetry

Email

Content Ingestion

Backups

Worker Configuration

Example Environment Blocks

Local Development Shell

Production Env File (Surreal, Default)

Migration Archive Rehearsal

Kubernetes ConfigMap

Kubernetes Secret

Running Multiple Instances

Port Configuration

Quick Setup: Test Instance

How It Works

Tips

Computed Properties