48 lines
1.4 KiB
Text
48 lines
1.4 KiB
Text
# haiku.rag configuration for Docker deployment
|
|
# See https://ggozad.github.io/haiku.rag/configuration/ for details
|
|
|
|
environment: production
|
|
|
|
storage:
|
|
data_dir: /data
|
|
|
|
ingester:
|
|
# Queue lives next to the LanceDB so both persist in the data volume.
|
|
queue:
|
|
path: /data/ingester.db
|
|
workers:
|
|
# Match the docling-serve fleet size below — each instance processes
|
|
# one task at a time, so more workers would just queue submissions
|
|
# inside docling-serve.
|
|
worker_count: 2
|
|
api:
|
|
# Bind to all interfaces inside the container so docker port-mapping works.
|
|
host: 0.0.0.0
|
|
# Non-loopback bind: set a bearer token. Source from an env var so the
|
|
# token doesn't sit in this file checked into ops repos.
|
|
auth_token: ${INGESTER_TOKEN}
|
|
sources:
|
|
- type: fs
|
|
id: docs
|
|
root: /docs
|
|
delete_orphans: true
|
|
|
|
# Remote document processing with docling-serve
|
|
processing:
|
|
converter: docling-serve
|
|
chunker: docling-serve
|
|
|
|
providers:
|
|
docling_serve:
|
|
# Two replicas; ingester round-robins jobs across them, each call's
|
|
# submit / poll / result pinned to one instance.
|
|
base_url:
|
|
- http://docling-serve-1:5001
|
|
- http://docling-serve-2:5001
|
|
api_key: ""
|
|
ollama:
|
|
base_url: http://host.docker.internal:11434
|
|
|
|
|
|
# For other providers (OpenAI, Anthropic, VoyageAI, etc.),
|
|
# see: https://ggozad.github.io/haiku.rag/configuration/
|