Vapora/config/llm-router.toml

# Multi-IA Router Configuration
# Phase 0: Configuration for LLM provider selection

[routing]
# Default provider if no specific routing rules match
default_provider = "claude"

# Enable cost tracking
cost_tracking_enabled = true

# Enable fallback on provider failure
fallback_enabled = true

[providers.claude]
enabled = true
# ANTHROPIC_API_KEY environment variable required
api_key = "${ANTHROPIC_API_KEY}"
model = "claude-sonnet-4-5-20250929"
max_tokens = 8192
temperature = 0.7

# Cost per 1M tokens (input/output)
cost_per_1m_input = 3.00
cost_per_1m_output = 15.00

[providers.openai]
enabled = true
# OPENAI_API_KEY environment variable required
api_key = "${OPENAI_API_KEY}"
model = "gpt-4o"
max_tokens = 4096
temperature = 0.7

# Cost per 1M tokens (input/output)
cost_per_1m_input = 2.50
cost_per_1m_output = 10.00

[providers.gemini]
enabled = true
# GOOGLE_API_KEY environment variable required
api_key = "${GOOGLE_API_KEY}"
model = "gemini-2.0-flash"
max_tokens = 8192
temperature = 0.7

# Cost per 1M tokens (input/output)
cost_per_1m_input = 0.30
cost_per_1m_output = 1.20

[providers.ollama]
enabled = true
# Local Ollama instance, no API key needed
url = "${OLLAMA_URL:-http://localhost:11434}"
model = "llama3.2"
max_tokens = 4096
temperature = 0.7

# No cost for local models
cost_per_1m_input = 0.00
cost_per_1m_output = 0.00

# Routing rules: assign providers based on task characteristics
[[routing_rules]]
name = "architecture_design"
condition = { task_type = "architecture" }
provider = "claude"
model_override = "claude-opus-4-20250514"

[[routing_rules]]
name = "code_generation"
condition = { task_type = "development" }
provider = "claude"

[[routing_rules]]
name = "documentation"
condition = { task_type = "documentation" }
provider = "openai"

[[routing_rules]]
name = "monitoring"
condition = { task_type = "monitoring" }
provider = "gemini"

[[routing_rules]]
name = "local_testing"
condition = { environment = "development" }
provider = "ollama"