Forge.diy/.env.example at main · DeveloperKush/Forge.diy · 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
# ---- FORGE engine credentials ----
# Paste whatever keys you have. FORGE auto-detects which are present and builds
# its model fallback chain from ONLY those providers — missing ones are skipped,
# and when one hits its rate/quota limit the chain falls through to the next.
# More keys = more total free throughput. All below are free; only SambaNova
# uses a (no-card) starter credit rather than a perpetual quota.
#
# MULTIPLE KEYS PER PROVIDER: add numbered slots to rotate across accounts when
# one key's limit runs out. The base var is tried first, then _2, _3, ... e.g.:
#   GROQ_API_KEY=key_from_account_a
#   GROQ_API_KEY_2=key_from_account_b
#   GROQ_API_KEY_3=key_from_account_c
# Works for every provider below (OPENROUTER_API_KEY_2, NVIDIA_NIM_API_KEY_2, …).

# Anthropic — PAID (Claude). Optional; when set it auto-promotes to TOP priority
# for both reasoning and build, no code change needed. https://console.anthropic.com
ANTHROPIC_API_KEY=

# OpenRouter — free Qwen3-Coder, gpt-oss, Llama. https://openrouter.ai/keys
OPENROUTER_API_KEY=

# Groq — fast; gpt-oss-120b, qwen3-32b. No card. https://console.groq.com/keys
GROQ_API_KEY=

# NVIDIA NIM — hosts qwen3-coder-480b (best free coder). No card. key starts nvapi-
# https://build.nvidia.com  (profile -> API keys)
NVIDIA_NIM_API_KEY=

# Cerebras — 1M tokens/day, very fast, 8K context cap. No card.
# https://cloud.cerebras.ai
CEREBRAS_API_KEY=

# SambaNova — 20M tokens/day, DeepSeek-V3. No card ($5 starter credit).
# https://cloud.sambanova.ai
SAMBANOVA_API_KEY=

# Mistral La Plateforme — mistral-large. No card (phone verify).
# https://console.mistral.ai/api-keys
MISTRAL_API_KEY=

# Codestral — Mistral's free coding model, SEPARATE key + endpoint from above.
# https://console.mistral.ai/codestral
CODESTRAL_API_KEY=

# Z.ai (Zhipu/GLM) — free glm-4.7-flash. No card. https://z.ai
ZAI_API_KEY=

# Chutes — decentralized free DeepSeek/Qwen (variable availability). key starts cpk_
# https://chutes.ai
CHUTES_API_KEY=

# Google AI (Gemini API). https://aistudio.google.com/apikey
GEMINI_API_KEY=

# Google Jules (async coding agent). Used by the Jules engine adapter for heavy tasks.
JULES_API_KEY=

# HEADLESS CLIENT PATHS (Optional, override system PATH detection)
AGY_PATH=
OPENCODE_PATH=


# ---- Route overrides (optional) ----
# Comma-separated LiteLLM model IDs, tried in order. Leave blank to use built-in defaults.
# Verify current free IDs on OpenRouter before pinning.
FORGE_REASONING_MODELS=
FORGE_BULK_MODELS=

# ---- Behaviour ----
# Workspace directory where artifacts + state.db live.
FORGE_WORKSPACE=workspace
# Per-call retry attempts inside the fallback chain.
FORGE_MAX_RETRIES=3