plano/demos/use_cases/llm_routing/arch_config.yaml

46 lines
921 B
YAML
Raw Normal View History

version: v0.1.0
listeners:
egress_traffic:
address: 0.0.0.0
port: 12000
message_format: openai
timeout: 30s
llm_providers:
2025-05-28 19:08:00 -07:00
2025-07-11 16:42:16 -07:00
- access_key: $OPENAI_API_KEY
model: openai/gpt-4o-mini
- access_key: $OPENAI_API_KEY
model: openai/gpt-4.1
2025-07-11 16:42:16 -07:00
- access_key: $OPENAI_API_KEY
model: openai/gpt-4o
2025-05-28 19:08:00 -07:00
default: true
2025-07-11 16:42:16 -07:00
- access_key: $MISTRAL_API_KEY
model: mistral/ministral-3b-latest
- access_key: $ANTHROPIC_API_KEY
model: claude/claude-3-7-sonnet-latest
- access_key: $ANTHROPIC_API_KEY
model: claude/claude-sonnet-4-0
- access_key: $DEEPSEEK_API_KEY
model: deepseek/deepseek-reasoner
- access_key: $GROQ_API_KEY
model: groq/llama-3.1-8b-instant
- access_key: $GEMINI_API_KEY
model: gemini/gemini-1.5-pro-latest
- model: custom/test-model
base_url: http://host.docker.internal:11223
provider_interface: openai
2025-06-11 15:15:00 -07:00
tracing:
random_sampling: 100