Skip to main content
Glama
Tunedness

Contextator

Official
by Tunedness

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
HOSTNoThe host interface to bind the server to.0.0.0.0
PORTNoThe port to bind the server to.3444
DATA_DIRNoWritable directory holding the materialised sources (git checkouts, uploads, Notion pulls). /data inside the container..data
SECRET_KEYNoAt least 32 characters (e.g. openssl rand -hex 32). Encrypts git/Notion tokens at rest (AES-256-GCM). Needed only once such a source exists; changing it invalidates stored tokens.
SETUP_CODENoOne-time code required by /setup to create the first account. If unset, the server generates one and prints it at every start until an account exists. Case, dashes and punctuation are ignored when checked.
ADMIN_TOKENNoMachine access token for /api/*, acting with root permissions. Treat like a root password; do not use in a browser.
DATABASE_URLNoLocal development database connection string. The container ignores it and talks to its embedded PostgreSQL via PG* variables set by the entrypoint.postgres://contextator:contextator@localhost:5432/contextator
IGNORE_GLOBSNoComma-separated glob patterns to ignore, e.g. '**/CHANGELOG.md,drafts/**'. Applies to every source.
RESET_VECTORSNoSet to '1' to drop all existing vectors and re-type the vector column (used when changing embedding dimension).0
DOCS_HOST_PATHNoHost folder mounted read-only at /docs (docker-compose only)../docs
OPENAI_API_KEYNoOpenAI API key, used when EMBEDDING_PROVIDER is 'openai'.
ALLOWED_ORIGINSNoExtra browser origins allowed on /mcp/* (non-browser clients are always allowed).
EMBEDDING_DTYPENoData type for the local embedding model. 'fp32' or 'q8' (q8 downloads a ~4× smaller quantized model).fp32
EMBEDDING_MODELNoEmbedding model to use. Any transformers.js feature-extraction model. English-only & faster: Xenova/all-MiniLM-L6-v2 (also 384-d).Xenova/paraphrase-multilingual-MiniLM-L12-v2
MODEL_CACHE_DIRNoModel download directory; /app/.cache/models inside the container..cache/models
PUBLIC_BASE_URLNoPublic base URL, e.g. https://docs.example.com, used for URLs shown in the dashboard.
CHUNK_MAX_TOKENSNoMaximum tokens per document chunk.400
ALLOWED_DOC_ROOTSNoComma-separated list of directories that project document roots must reside in. Path-escape protection. On Windows dev, e.g. C:/path/to/docs./docs
EMBEDDING_OFFLINENoSet to '1' to prevent the embedding model from downloading at startup (for air-gapped hosts).0
POSTGRES_PASSWORDNoPassword of the embedded PostgreSQL (loopback only), applied when the cluster is first created.contextator
AUTH_COOKIE_SECURENoCookie Secure flag: 'auto' sets Secure when the request arrives over HTTPS. Force with '1', use '0' for plain-HTTP LAN installs.auto
EMBEDDING_PROVIDERNoEmbedding provider: 'local' or 'openai'.local
ARCHIVE_MAX_ENTRIESNoZip-bomb guard: maximum number of entries in an archive (applied while extracting).20000
PASSWORD_MIN_LENGTHNoMinimum password length for all passwords, including temporary ones. No composition rules.12
SESSION_IDLE_TTL_MSNoIdle Streamable HTTP sessions are closed after this many milliseconds.1800000
AUTH_SESSION_IDLE_MSNoDashboard session idle timeout in milliseconds. A session unused for this long must sign in again.43200000
CHUNK_OVERLAP_TOKENSNoToken overlap between consecutive chunks.50
EMBEDDING_DIMENSIONSNoDimension of the embedding vectors. Must match the model. 1536 for text-embedding-3-small.384
AUTH_LOGIN_WINDOW_MINNoThe IP window (in minutes) and the first lockout step (doubles, capped at an hour).15
AUTH_SESSION_TTL_DAYSNoHard maximum lifetime of a dashboard session in days, however actively used.30
UPLOAD_MAX_FILE_BYTESNoMaximum size per uploaded file.52428800
OPENAI_EMBEDDING_MODELNoOpenAI embedding model to use when EMBEDDING_PROVIDER is 'openai'.text-embedding-3-small
ARCHIVE_MAX_TOTAL_BYTESNoZip-bomb guard: maximum total uncompressed size of an archive (applied while extracting).1073741824
AUTH_LOGIN_MAX_ATTEMPTSNoMaximum failed sign-in attempts per account and per IP before lockout or 429.10
CONTEXTATOR_MODELS_PATHNoAbsolute host directory used for embedding models instead of the volume (docker-compose only).
CONTEXTATOR_PGDATA_PATHNoAbsolute host directory used for PostgreSQL data instead of the volume (docker-compose only).
UPLOAD_MAX_ARCHIVE_BYTESNoMaximum size per uploaded archive.268435456
CONTEXTATOR_MODELS_VOLUMENoDocker volume name for embedding models (docker-compose only).contextator-models
CONTEXTATOR_PGDATA_VOLUMENoDocker volume name for PostgreSQL data (docker-compose only).contextator-pgdata
UPLOAD_MAX_FILES_PER_REQUESTNoMaximum number of files per upload request.500

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

This server publishes no instructions, or was last inspected before Glama recorded them.

Capabilities

Server capabilities have not been inspected yet.

Tools

Functions exposed to the LLM to take actions

NameDescription

No tools

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources