LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
# Flask
FLASK_APP=src/unstract/platform_service/run.py
FLASK_RUN_PORT=3001
# Redis
REDIS_HOST=unstract-redis
REDIS_PORT=6379
REDIS_USERNAME=default
REDIS_PASSWORD=
# Redis Sentinel mode - set to True for HA deployments (K8s only)
# When True: point REDIS_HOST to Sentinel K8s service DNS, REDIS_PORT to 26379
# REDIS_PASSWORD is reused for Sentinel auth.
REDIS_SENTINEL_MODE=False
REDIS_SENTINEL_MASTER_NAME=mymaster
# Managed / external Redis with TLS (UN-4123). Optional — unset keeps the plaintext
# connection above. Either set REDIS_SSL=true beside REDIS_HOST/REDIS_PORT, or give a
# full URL whose scheme carries TLS:
# REDIS_URL=rediss://:<password>@<host>:6380/0?ssl_cert_reqs=required
# A URL wins over the discrete vars. Auth is password-only as the built-in `default`
# user (what a managed AUTH string is), so leave the username empty for a managed
# endpoint. REDIS_SSL_CA_CERTS is only needed when the server's CA is not publicly
# trusted (Memorystore).
# REDIS_SSL=false
# REDIS_SSL_CERT_REQS=required
# REDIS_SSL_CA_CERTS=
# REDIS_URL=
# REDIS_HEALTH_CHECK_INTERVAL=30
# Backend DB
PG_BE_HOST=unstract-db
PG_BE_PORT=5432
PG_BE_USERNAME=unstract_dev
PG_BE_PASSWORD=unstract_pass
PG_BE_DATABASE=unstract_db
DB_SCHEMA="unstract"
# Encryption Key
# key must be 32 url-safe base64-encoded bytes.
ENCRYPTION_KEY="Sample-Key"
# Feature Flags
EVALUATION_SERVER_IP=unstract-flipt
EVALUATION_SERVER_PORT=9000
PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
# Flipt Service
FLIPT_SERVICE_AVAILABLE=False
# Cost calculation related ENVs
MODEL_PRICES_URL="https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
MODEL_PRICES_TTL_IN_DAYS=7
MODEL_PRICES_FILE_PATH="<bucket-name>/cost/model_prices.json"
#Remote storage config
FILE_STORAGE_CREDENTIALS='{"provider":"local"}'
REMOTE_MODEL_PRICES_FILE_PATH="unstract/cost/model_prices.json"
LOG_LEVEL=INFO