Environment Variables¶

Startup configuration is read from WSE_-prefixed variables through src/world_simulation_engine/misc/config.py. By default the backend loads .env; set WSE_ENV_FILE if you want to use a different file.

WSE_ENV_FILE=/path/to/local.env uvicorn world_simulation_engine.app:app --host 127.0.0.1 --port 9797

For Docker Compose, the provided docker-compose.yml passes .env into the application container:

services:
  app:
    env_file:
      - .env

Required variables¶

Variable	Default	Description
`WSE_NEO4J_PASSWORD`	Required	Password used to connect to Neo4j. The Docker Compose dependency example uses `neo4j/password`; change it before any non-local use.

Common variables¶

Variable	Default	Description
`WSE_APP_HOST`	`127.0.0.1`	Host used by the local start script. When running behind Docker/nginx, the container entrypoint and proxy decide the externally exposed host.
`WSE_APP_PORT`	`9797`	Port used by the local start script. The full Docker Compose stack exposes the web app on host port `9798`.
`WSE_LOGGING_LEVEL`	`INFO`	Python logging level. Supported values are `CRITICAL`, `ERROR`, `WARNING`, `INFO`, `DEBUG`, and `NOTSET`.
`WSE_NEO4J_URI`	`bolt://localhost:7687`	Bolt URI for Neo4j. Use the host and port visible from the backend process or container.
`WSE_NEO4J_USERNAME`	`neo4j`	Username used to connect to Neo4j.
`WSE_DATA_FOLDER`	`data/storage`	Content-addressed storage folder for uploaded media, generated images, prompt overrides, and workflow files. In Docker, this is normally mounted under `/app/data/storage`.

TTS service limits¶

These variables do not choose a TTS provider. They only control backend behavior once TTS is configured in the interface.

Variable	Default	Description
`WSE_TTS_MAX_CONCURRENCY`	`3`	Maximum number of concurrent outbound TTS generation requests. Lower this if the TTS backend becomes unstable under load.
`WSE_TTS_MEDIA_RETENTION_TURNS`	`50`	Number of most-recent turns whose generated voice media are retained per simulation before older generated files are pruned.

SillyTavern import limits¶

These variables tune the SillyTavern character card import pipeline. They do not need to be set for normal simulation use.

Variable	Default	Description
`WSE_SILLYTAVERN_IMPORT_MAX_CONCURRENCY`	`4`	Maximum concurrent LLM calls within one import pipeline stage (e.g. classifying many lorebook entries at once).
`WSE_SILLYTAVERN_IMAGE_DOWNLOAD_MAX_CONCURRENCY`	`4`	Maximum concurrent outbound HEAD/GET requests when probing or downloading image links found in a card.
`WSE_SILLYTAVERN_IMAGE_DOWNLOAD_MAX_BYTES`	`10485760`	Maximum response body size (bytes) accepted for a candidate image download.
`WSE_SILLYTAVERN_IMAGE_DOWNLOAD_CONNECT_TIMEOUT`	`5.0`	Connect timeout, in seconds, for outbound image link requests.
`WSE_SILLYTAVERN_IMAGE_DOWNLOAD_READ_TIMEOUT`	`10.0`	Read timeout, in seconds, for outbound image link GET requests.
`WSE_SILLYTAVERN_IMAGE_DOWNLOAD_HEAD_TIMEOUT`	`5.0`	Total timeout, in seconds, for the cheaper HEAD probe used before offering a non-whitelisted image link for review.
`WSE_SILLYTAVERN_IMAGE_DOWNLOAD_MAX_REDIRECTS`	`3`	Maximum redirect hops followed when downloading a candidate image link; each hop is re-validated against the SSRF filter.

Example `.env`¶

WSE_LOGGING_LEVEL=INFO
WSE_NEO4J_URI=bolt://localhost:7687
WSE_NEO4J_USERNAME=neo4j
WSE_NEO4J_PASSWORD=password
WSE_DATA_FOLDER=data/storage
WSE_TTS_MAX_CONCURRENCY=3
WSE_TTS_MEDIA_RETENTION_TURNS=50

Local-only security model

The application is designed for local use. It does not include authentication or multi-user isolation, and prompts or world content may be sent to configured external model providers.

What is not configured here¶

Do not put LLM, embedding, image, TTS, STT, prompt, or per-component model assignments in .env. Those are runtime service settings managed through the web interface and stored in Neo4j.