# ─────────────────────────────────────────────
# OpenSwarm — environment configuration
# Copy this file to .env and fill in your keys.
# Run `uv run python run.py` to launch the onboarding
# wizard, which fills this file automatically.
# ─────────────────────────────────────────────


# ── Model routing ─────────────────────────────

# OpenSwarm can use multiple model/auth backends at the same time.
# Subscription-backed options use your local CLI login:
#   subscription/codex   -> `codex login`
#   subscription/claude  -> `claude auth login`
#
# API-backed options use API keys below:
#   gpt-5.2
#   litellm/anthropic/claude-sonnet-4-6
#   litellm/gemini/gemini-3-flash
DEFAULT_MODEL=subscription/codex

# Optional per-agent overrides. Leave blank to use DEFAULT_MODEL.
ORCHESTRATOR_MODEL=
GENERAL_AGENT_MODEL=
DEEP_RESEARCH_MODEL=
DATA_ANALYST_MODEL=
DOCS_AGENT_MODEL=
SLIDES_AGENT_MODEL=
IMAGE_AGENT_MODEL=
VIDEO_AGENT_MODEL=

# Optional model-call timeout in seconds. Blank/0/none disables OpenSwarm hard
# model timeouts; use the TUI cancel flow to stop long-running requests.
OPENSWARM_MODEL_TIMEOUT_SECONDS=

# Optional request-local failover for clear quota/rate/usage limit errors. This
# does not rewrite DEFAULT_MODEL or per-agent model env vars; it only retries the
# current model call with an available fallback.
OPENSWARM_MODEL_FAILOVER=auto
OPENSWARM_MODEL_FAILOVER_ORDER=subscription/codex,subscription/claude,gpt-5.2,litellm/anthropic/claude-sonnet-4-6,litellm/gemini/gemini-3-flash
OPENSWARM_MODEL_FAILOVER_MAX_ATTEMPTS=2

# Optional Security Research Swarm per-agent overrides. Leave blank to use DEFAULT_MODEL.
SECURITY_RESEARCH_ORCHESTRATOR_MODEL=
SECURITY_RESEARCH_LEAD_MODEL=
THREAT_INTELLIGENCE_ANALYST_MODEL=
VULNERABILITY_RESEARCHER_MODEL=
OSINT_ENRICHMENT_SPECIALIST_MODEL=
SECURITY_LAB_ANALYST_MODEL=
TECHNICAL_BLOG_WRITER_MODEL=
SECURITY_RESEARCH_SLIDES_AGENT_MODEL=
SECURITY_VISUAL_DESIGNER_MODEL=

# Optional OpenSwarm skills directory. Defaults to ./openswarm_skills.
OPENSWARM_SKILLS_DIR=


# ── Web research routing ─────────────────────

# WebResearchSearch can use Codex and Claude Code subscription web search,
# then fall back to platform/API search when configured.
WEB_SEARCH_MODE=auto
WEB_SEARCH_PROVIDER_ORDER=codex,claude
WEB_SEARCH_DEEP_MIX=true
WEB_SEARCH_PLATFORM_FALLBACK=true
WEB_SEARCH_TIMEOUT_SECONDS=60
WEB_SEARCH_OPENAI_MODEL=gpt-5.2


# ── Optional model/API provider keys ──────────

# OpenAI API — direct OpenAI model/API access, hosted web search fallback,
# OpenAI Images, and Sora.
OPENAI_API_KEY=

# Anthropic API — direct Claude API access through LiteLLM.
ANTHROPIC_API_KEY=

# Google Gemini API — Gemini models/images and Veo video.
GOOGLE_API_KEY=


# ── Optional: Composio ───────────────────────

# Composio unlocks 10,000+ external integrations (Gmail, Slack, Google Calendar,
# HubSpot, GitHub, and more) for the General Agent.
# Get your key and user ID at https://composio.dev
COMPOSIO_API_KEY=
COMPOSIO_USER_ID=


# ── Optional: Search ─────────────────────────

# Search API key — enables SearchAPI fallback, Scholar search, and product
# search across agents.
# Get your key at https://www.searchapi.io
SEARCH_API_KEY=


# ── Optional: Images / Video ─────────────────────────

# Google AI / Gemini — used for image generation, editing, and composition
# (Gemini 2.5 Flash Image, Gemini 3 Pro Image) across Image Agent, Slides Agent,
# and Video Agent; also required for Veo video generation (Video Agent).
# Configure GOOGLE_API_KEY once in the Provider section above.
# Get your key at https://aistudio.google.com/app/apikey

# Pexels — stock photo search used by Slides Agent (ImageSearch tool).
# Get your key at https://www.pexels.com/api
PEXELS_API_KEY=

# Pixabay — stock photo search used by Slides Agent (ImageSearch tool).
# Get your key at https://pixabay.com/api/docs
PIXABAY_API_KEY=

# Unsplash — stock photo search used by Slides Agent (ImageSearch tool).
# Get your key at https://unsplash.com/developers
UNSPLASH_ACCESS_KEY=

# fal.ai — used for Seedance 1.5 Pro video generation, video editing (Video Agent),
# and background removal (Image Agent).
# Get your key at https://fal.ai/dashboard/keys
FAL_KEY=