.env.example

# =============================================================================
# Pantheon — Environment Configuration
# =============================================================================
# Copy this file to .env and fill in your API keys.
# At minimum, you need ONE primary LLM provider + ONE embedding provider.
#
# Quick start: set both OPENCODE_GO_API_KEY and OPENROUTER_API_KEY below.
# =============================================================================

# ── Primary LLM Provider ─────────────────────────────────────────────────────
# Used for chat, entity extraction, classification, and general AI tasks.
# Pick ONE:

# 🟢 OpenCode Go — $10/mo flat, multiple open models
# Get key: https://opencode.ai/auth
# OPENCODE_GO_API_KEY=sk-put_your_key_here

# 🟡 OpenRouter — pay-as-you-go, 200+ models including free ones
# Get key: https://openrouter.ai/keys
# OPENROUTER_API_KEY=sk-or-v1-put_your_key_here

# 🟣 Ollama (local) — free, offline, runs on your machine
# Install: https://ollama.com/download
# OLLAMA_API_KEY=ollama  # Any value works — Ollama ignores it
# OLLAMA_BASE_URL=http://localhost:11434  # Optional, default

# 🔵 Anthropic / OpenAI / Google Gemini / DeepSeek / Groq / etc.
# See full reference: https://github.com/Duskript/Pantheon/wiki/Providers

# ── Embedding Provider ───────────────────────────────────────────────────────
# Required for semantic search, knowledge retrieval, and Athenaeum ingestion.
# If your primary provider is OpenCode Go, you MUST set one of these.
# If your primary provider is OpenRouter, it can serve double duty.

# 🟡 OpenRouter (recommended — free model, no extra cost)
#   Model: nvidia/llama-nemotron-embed-vl-1b-v2:free (2048-dim)
#   Get key: https://openrouter.ai/keys
# No extra env var needed if OPENROUTER_API_KEY is set above.
# Override with ATHENAEUM_EMBED_API_KEY if using a different key.

# 🟣 Ollama (local) — free, offline
#   Requires Ollama installed + an embedding model pulled:
#     ollama pull nomic-embed-text    # 768-dim, good quality
#     ollama pull bge-m3              # 1024-dim, best quality
# ATHENAEUM_EMBED_PROVIDER=ollama
# ATHENAEUM_EMBED_API_KEY=ollama
# ATHENAEUM_EMBED_MODEL=nomic-embed-text
# ATHENAEUM_EMBED_URL=http://localhost:11434/api/embeddings

# 🔵 Jina AI — $0.01/1M tokens (first 1M free)
#   Model: jina-embeddings-v3 (1024-dim)
#   Get key: https://jina.ai/embeddings
# ATHENAEUM_EMBED_PROVIDER=jina
# ATHENAEUM_EMBED_API_KEY=jina_put_your_key_here
# ATHENAEUM_EMBED_MODEL=jina-embeddings-v3
# ATHENAEUM_EMBED_URL=https://api.jina.ai/v1/embeddings

# 🟢 Voyage AI — $0.01/1M tokens
#   Model: voyage-lite-02-instruct (1024-dim)
#   Get key: https://dash.voyageai.com/
# ATHENAEUM_EMBED_PROVIDER=voyage
# ATHENAEUM_EMBED_API_KEY=pa-put_your_key_here
# ATHENAEUM_EMBED_MODEL=voyage-lite-02-instruct
# ATHENAEUM_EMBED_URL=https://api.voyageai.com/v1/embeddings

# ── System Configuration (optional overrides) ────────────────────────────────
# These are auto-configured — only set if you need to override.

# ChromaDB (vector store) — defaults to local persistent
# CHROMA_HOST=localhost
# CHROMA_PORT=8000

# Ollama base URL (used when primary provider is Ollama)
# OLLAMA_BASE_URL=http://localhost:11434

# ── Advanced Model Overrides ─────────────────────────────────────────────────
# Only set these if you want to pin specific models for specific roles.
# By default, Pantheon selects the best model for each task.

# ATHENAEUM_EXTRACT_MODEL=minimax-m2.5
# ATHENAEUM_EXTRACT_MODEL_FALLBACK=deepseek-v4-flash
# ATHENAEUM_EMBED_MODEL=nvidia/llama-nemotron-embed-vl-1b-v2:free