Repository navigation
Expand file tree
/
Copy pathenv.example
More file actions
79 lines (65 loc) · 3.15 KB
/
Copy pathenv.example
File metadata and controls
79 lines (65 loc) · 3.15 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
# Copy this file to .env and adjust for your machine.
# docker compose reads .env automatically from the same directory as
# docker-compose.local.yaml. Values here override the defaults baked
# into the compose file; anything left unset falls back to the
# default shown in docker-compose.local.yaml itself.
# ---------------------------------------------------------------------------
# Image tag
# ---------------------------------------------------------------------------
TAG=latest
# ---------------------------------------------------------------------------
# Storage locations
#
# MARIGOLD_CACHE_DIR holds the pre-built model weight cache (read-only
# into the worker). Populate it first with `make cache/local`.
#
# MARIGOLD_MODEL_TMP_DIR is scratch space the worker uses for model
# offload during load (4-bit quantisation, MoE, format conversion).
# Point this at a separate disk if your model cache disk is small or
# slow -- this was previously hardcoded to an absolute host path
# (/data/marigold/model-tmp); set it here instead if you need the
# same separation.
#
# MARIGOLD_MODELS_CONFIG selects which generated model registry file
# the worker and api services read. assets/models-3060.json and
# assets/models-2060.json are consumer-GPU presets; point this at a
# different generated file for a different model set.
# ---------------------------------------------------------------------------
MARIGOLD_CACHE_DIR=./cache/models
MARIGOLD_MODEL_TMP_DIR=./cache/tmp
# ---------------------------------------------------------------------------
# Model selection
#
# Comma-separated list of model hash IDs the worker will serve, taken
# from the generated models.json file referenced above. The default
# below matches the models-3060.json preset. There is currently no
# tool to look up a hash from a model name directly -- cross-reference
# the generated JSON file by name until that exists.
# ---------------------------------------------------------------------------
# MARIGOLD_MODELS=
# ---------------------------------------------------------------------------
# Ports (host side; container-side ports are fixed in the compose file)
# ---------------------------------------------------------------------------
POSTGRES_PORT=5432
API_PORT=8000
WEBUI_PORT=3000
# ---------------------------------------------------------------------------
# Postgres credentials (local development defaults -- change for any
# shared or multi-user environment)
# ---------------------------------------------------------------------------
POSTGRES_DB=marigold
POSTGRES_USER=marigold
POSTGRES_PASSWORD=marigold
# ---------------------------------------------------------------------------
# Worker behaviour
# ---------------------------------------------------------------------------
IDLE_TIMEOUT=600
LOG_LEVEL=INFO
# ---------------------------------------------------------------------------
# open-webui
#
# WEBUI_DUMMY_KEY is not a real credential -- the local API accepts
# any request (see README.md, "Authentication"). open-webui's client
# requires a non-empty value to be set; it is never checked.
# ---------------------------------------------------------------------------
WEBUI_DUMMY_KEY=marigold