Sebastian Krüger
f3f32c163f
- Replace COMFYUI_BACKEND_HOST and SUPERVISOR_BACKEND_HOST with GPU_TAILSCALE_IP - Update LiteLLM config to use os.environ/GPU_TAILSCALE_IP for vLLM models - Add GPU_TAILSCALE_IP env var to LiteLLM service - Configure qwen-2.5-7b and llama-3.1-8b to route through orchestrator 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
90 lines
3.1 KiB
YAML
90 lines
3.1 KiB
YAML
model_list:
|
|
- model_name: claude-sonnet-4
|
|
litellm_params:
|
|
model: anthropic/claude-sonnet-4-20250514
|
|
api_key: os.environ/ANTHROPIC_API_KEY
|
|
|
|
- model_name: claude-sonnet-4.5
|
|
litellm_params:
|
|
model: anthropic/claude-sonnet-4-5-20250929
|
|
api_key: os.environ/ANTHROPIC_API_KEY
|
|
|
|
- model_name: claude-3-5-sonnet
|
|
litellm_params:
|
|
model: anthropic/claude-3-5-sonnet-20241022
|
|
api_key: os.environ/ANTHROPIC_API_KEY
|
|
|
|
- model_name: claude-3-opus
|
|
litellm_params:
|
|
model: anthropic/claude-3-opus-20240229
|
|
api_key: os.environ/ANTHROPIC_API_KEY
|
|
|
|
- model_name: claude-3-haiku
|
|
litellm_params:
|
|
model: anthropic/claude-3-haiku-20240307
|
|
api_key: os.environ/ANTHROPIC_API_KEY
|
|
|
|
# ===========================================================================
|
|
# SELF-HOSTED MODELS VIA ORCHESTRATOR (GPU Server via Tailscale VPN)
|
|
# ===========================================================================
|
|
# All requests route through orchestrator (port 9000) which manages model loading
|
|
|
|
# Text Generation
|
|
- model_name: qwen-2.5-7b
|
|
litellm_params:
|
|
model: hosted_vllm/openai/qwen-2.5-7b # hosted_vllm/openai/ for vLLM via orchestrator
|
|
api_base: http://os.environ/GPU_TAILSCALE_IP:9000/v1 # RunPod GPU via Tailscale
|
|
api_key: dummy
|
|
rpm: 1000
|
|
tpm: 100000
|
|
timeout: 600 # 10 minutes for generation
|
|
stream_timeout: 600
|
|
supports_system_messages: false # vLLM handles system messages differently
|
|
stream: true # Enable streaming by default
|
|
|
|
- model_name: llama-3.1-8b
|
|
litellm_params:
|
|
model: hosted_vllm/openai/llama-3.1-8b # hosted_vllm/openai/ for vLLM via orchestrator
|
|
api_base: http://os.environ/GPU_TAILSCALE_IP:9000/v1 # RunPod GPU via Tailscale
|
|
api_key: dummy
|
|
rpm: 1000
|
|
tpm: 100000
|
|
timeout: 600 # 10 minutes for generation
|
|
stream_timeout: 600
|
|
supports_system_messages: true # Llama supports system messages
|
|
stream: true # Enable streaming by default
|
|
|
|
litellm_settings:
|
|
drop_params: false # DISABLED: Was breaking streaming
|
|
set_verbose: true # Enable verbose logging for debugging streaming issues
|
|
# Enable caching now that streaming is fixed
|
|
cache: true
|
|
cache_params:
|
|
type: redis
|
|
host: core_redis
|
|
port: 6379
|
|
ttl: 3600 # Cache for 1 hour
|
|
# Force strip specific parameters globally
|
|
allowed_fails: 0
|
|
# Modify params before sending to provider
|
|
modify_params: false # DISABLED: Was breaking streaming
|
|
# Enable success and failure logging but minimize overhead
|
|
success_callback: [] # Disable all success callbacks to reduce DB writes
|
|
failure_callback: [] # Disable all failure callbacks
|
|
|
|
router_settings:
|
|
allowed_fails: 0
|
|
|
|
# Drop unsupported parameters
|
|
default_litellm_params:
|
|
drop_params: false # DISABLED: Was breaking streaming
|
|
|
|
general_settings:
|
|
disable_responses_id_security: true
|
|
# Disable spend tracking to reduce database overhead
|
|
disable_spend_logs: true
|
|
# Disable tag tracking
|
|
disable_tag_tracking: true
|
|
# Disable daily spend updates
|
|
disable_daily_spend_logs: true
|