open-agent/.env.example at main · haseeb-heaven/open-agent · 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
# Local providers need NO key: Ollama (localhost:11434) and LM Studio
# (localhost:1234) are auto-detected. Cloud providers each take one key.
# Add keys interactively with:  --byok  (or /byok inside a session)

# Core providers
OPENAI_API_KEY=
ANTHROPIC_API_KEY=
GEMINI_API_KEY=
GROQ_API_KEY=
DEEPSEEK_API_KEY=

# OpenAI-compatible providers
NVIDIA_API_KEY=
TOGETHER_API_KEY=
OPENROUTER_API_KEY=
Z_AI_API_KEY=

# HuggingFace Inference (HF_TOKEN preferred; HUGGINGFACE_API_KEY also accepted)
HF_TOKEN=

# Cerebras (ultra-fast LPU inference, cloud.cerebras.ai — free public endpoints, rate-limited)
CEREBRAS_API_KEY=

# -----------------------------------------------------------------------------
# Web search backends (google_web_search tool)
# OpenAgent auto-picks a backend from available keys. Recommended:
#   - Gemini models  → GEMINI_API_KEY (Google Search grounding)
#   - Open-source / free models → BRAVE_API_KEY (or TAVILY / SERPER / EXA)
#   - Local models / no keys → DuckDuckGo (no key)
# Force a backend: WEB_SEARCH_PROVIDER=brave|tavily|serper|exa|gemini|duckduckgo
# Interactive setup: /websearch  (wizard) · /websearch open <id>  · /websearch list
# -----------------------------------------------------------------------------
BRAVE_API_KEY=
TAVILY_API_KEY=
SERPER_API_KEY=
EXA_API_KEY=
# GEMINI_API_KEY already listed above — also enables Google Search grounding
WEB_SEARCH_PROVIDER=