UDA-Bench/.env.example at main · unimib-datAI/UDA-Bench · GitHub

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
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
# =========================================================================
# UDA-Bench — root environment file
#
# Copy to `.env` in this same directory and fill in the values you need:
#
#     cp .env.example .env
#
# This single file is read by every system: DocETL and Evaporate load it from
# the repository root explicitly, QUEST finds it by walking up the tree.
# Do NOT create a second `.env` under `systems/quest/`: it would shadow this
# one for QUEST only, silently splitting the configuration.
# =========================================================================


# -------------------------------------------------------------------------
# LLM provider — configure at least one
# -------------------------------------------------------------------------

# Azure OpenAI. Required by QUEST, also usable by Evaporate.
AZURE_OPENAI_API_KEY=
AZURE_OPENAI_ENDPOINT=https://<resource>.openai.azure.com
AZURE_OPENAI_DEPLOYMENT=
OPENAI_API_VERSION=2024-12-01-preview

# Azure through LiteLLM. Required by DocETL when it runs on Azure.
# AZURE_API_KEY=
# AZURE_API_BASE=https://<resource>.cognitiveservices.azure.com/
# AZURE_API_VERSION=2024-12-01-preview

# Gemini / OpenAI, alternatives to Azure for DocETL and Evaporate.
GEMINI_API_KEY=
# OPENAI_API_KEY=


# -------------------------------------------------------------------------
# Model selection (optional — defaults apply when unset)
# -------------------------------------------------------------------------

# DocETL default: gemini/gemini-2.5-flash
# DOCETL_DEFAULT_MODEL=azure/<deployment-name>

# Evaporate resolves, in order: EVAPORATE_MODEL, AZURE_OPENAI_DEPLOYMENT,
# AZURE_OPENAI_MODEL, then gemini-2.5-flash.
# EVAPORATE_MODEL=


# -------------------------------------------------------------------------
# QUEST — pgvector database (see systems/quest/docker-compose.yml)
# -------------------------------------------------------------------------

HOST=localhost
DATABASE=quest

# DB_USER is the DB role and is read before USER. On macOS/Linux the shell
# already exports USER (your login name) and it would win over this file,
# so leaving only USER would create the database under the wrong name.
# USER is kept for backward compatibility with existing setups.
DB_USER=quest
USER=quest

PASSWORD=quest_password
DB_PORT_EXTERNAL=5433
DB_PORT_INTERNAL=5432

# Device used to compute embeddings during indexing: "cpu" (default),
# "cuda" on NVIDIA, or "mps" to use the GPU on Apple Silicon.
# With "mps" also set PYTORCH_ENABLE_MPS_FALLBACK=1, so any operation not
# implemented in the Metal backend falls back to CPU instead of raising.
# QUEST_EMBEDDING_DEVICE=mps
# PYTORCH_ENABLE_MPS_FALLBACK=1


# -------------------------------------------------------------------------
# DQL — external API service, must be running separately
# -------------------------------------------------------------------------

# Falls back to http://127.0.0.1:9000/api/answer when unset.
# DQL_API_URL=http://127.0.0.1:8000/api/v2/chat

# Seconds; raise it for long-running queries (e.g. 1800 = 30 min).
# DQL_REQUEST_TIMEOUT=1800


# -------------------------------------------------------------------------
# Python interpreters (optional)
#
# The orchestrator looks for `.venv-<model>/bin/python` in the repository
# root. Set these only if your virtualenvs live elsewhere.
# -------------------------------------------------------------------------

# DOCETL_PYTHON=
# EVAPORATE_PYTHON=
# QUEST_PYTHON=
# LOTUS_PYTHON=
# DQL_PYTHON=


# -------------------------------------------------------------------------
# Optional, only for legacy/ZenDB paths
# -------------------------------------------------------------------------

# DEEPSEEK_API_KEY=
# DEEPSEEK_BASE_URL=https://aihubmix.com/v1
# TOGETHER_API_KEY=


QUEST_EMBEDDING_DEVICE=mps
PYTORCH_ENABLE_MPS_FALLBACK=1