Configuration Reference

All settings are controlled via environment variables (or a .env file).

Storage

Variable	Default	Description
BACKEND	local	Storage backend: local (SQLite + LanceDB) or postgres (PostgreSQL + pgvector)
DATA_DIR	~/.team-memory	Local data directory (SQLite, LanceDB, private store)
DATABASE_URL	—	PostgreSQL connection string (required when BACKEND=postgres)
LOG_LEVEL	INFO	Logging level

Embedding provider

Variable	Default	Description
EMBEDDING_PROVIDER	ollama	Provider: ollama, gemini, vertex, bedrock, azure
EMBEDDING_MODEL	(per provider)	Embedding model name. Defaults: ollama=nomic-embed-text, gemini=text-embedding-004, vertex=text-embedding-005, bedrock=amazon.titan-embed-text-v2:0, azure=text-embedding-3-small

All embeddings must produce 768-dimensional vectors.

Distillation provider

Variable	Default	Description
DISTILLER_PROVIDER	ollama	Provider: ollama, gemini
LLM_MODEL	(per provider)	Distillation model name. Defaults: ollama=gemma3:4b, gemini=gemini-2.0-flash

Privacy tradeoff

With DISTILLER_PROVIDER=gemini, raw text is sent to Google's API for distillation. With ollama, raw text stays on your device.

Provider credentials

Ollama (default, no credentials)

Variable	Default	Description
OLLAMA_HOST	http://localhost:11434	Ollama API endpoint

Ollama also reads its own environment variables for GPU control (OLLAMA_NUM_GPU, CUDA_VISIBLE_DEVICES, etc.). See the GPU Setup guide.

Gemini

Variable	Default	Description
GEMINI_API_KEY	—	Required when EMBEDDING_PROVIDER=gemini or DISTILLER_PROVIDER=gemini

Free tier available at AI Studio. See Gemini Backend.

Vertex AI

Variable	Default	Description
GCP_PROJECT	—	GCP project ID (required when EMBEDDING_PROVIDER=vertex)
GCP_LOCATION	us-central1	GCP region
CLOUD_SQL_CONNECTION	—	Cloud SQL instance connection name

Bedrock

Variable	Default	Description
AWS_REGION	us-east-1	AWS region

Uses the default AWS credential chain (environment variables, ~/.aws/credentials, or instance role).

Azure OpenAI

Variable	Default	Description
AZURE_OPENAI_ENDPOINT	—	Endpoint URL (required when EMBEDDING_PROVIDER=azure)
AZURE_OPENAI_API_KEY	—	API key (required when EMBEDDING_PROVIDER=azure)

Auto-observe pipeline

Variable	Default	Description
DISTILL_INGEST_PORT	21746	Port for the local HTTP /observe endpoint

Privacy & processing

Variable	Default	Description
DISTILL_ENABLED	true	Enable LLM distillation (disable for testing)
DEFAULT_AUTHOR	unknown	Default author attribution
AUTH_ENABLED	false	Enable git-based identity + PostgreSQL RLS

Search tuning

Variable	Default	Description
RRF_K	60	Reciprocal Rank Fusion constant
MAX_MEMORY_SIZE	8000	Maximum memory content size in characters
FTS_LANGUAGE	simple	Full-text search language configuration

Reranking (optional)

Variable	Default	Description
RERANK_ENABLED	false	Enable cross-encoder reranking
JINA_API_KEY	—	Jina Reranker API key
RERANK_MODEL	jina-reranker-v2-base-multilingual	Reranker model

Port interfaces

Port	Adapters
StoragePort	local: SQLite + FTS5 + LanceDB
EmbeddingPort	ollama gemini vertex bedrock azure — selected by EMBEDDING_PROVIDER
DistillerPort	ollama gemini — selected by DISTILLER_PROVIDER
ScannerPort	secrets + PII detection (always active)
RerankerPort	Jina Reranker API (opt-in via RERANK_ENABLED)