fix(deploy): rename gitignored config.yaml to proxy_config.yaml

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
This commit is contained in:
Moe Khalil 2026-09-07 16:42:13 +00:00
parent ab4d4af3db
commit fa418a8b14
3 changed files with 42 additions and 3 deletions

View file

@ -8,6 +8,6 @@ COPY enterprise/ /app/enterprise/
COPY litellm-proxy-extras/litellm_proxy_extras/ ${SITE_PACKAGES}/litellm_proxy_extras/
COPY litellm-proxy-extras/ /app/litellm-proxy-extras/
COPY schema.prisma /app/schema.prisma
COPY deploy/autorouter-sandbox/config.yaml /app/config.yaml
COPY deploy/autorouter-sandbox/proxy_config.yaml /app/config.yaml
CMD ["--config", "/app/config.yaml", "--port", "4000"]

View file

@ -2,7 +2,7 @@
A LiteLLM deployment built from this branch that you can push to and redeploy in about a minute, for iterating on the auto router without touching the real sandbox or prod
Every model that is not defined in `config.yaml` is forwarded to the upstream gateway (`UPSTREAM_LITELLM_BASE_URL`, i.e. gateway.litellm-sandbox.ai) through `litellm_proxy/*`, so this instance inherits all upstream models and provider credentials without copying any keys. Auto routers defined here (`moe-router`) pick between those upstream models and the router code running is whatever is on this branch
Every model that is not defined in `proxy_config.yaml` is forwarded to the upstream gateway (`UPSTREAM_LITELLM_BASE_URL`, i.e. gateway.litellm-sandbox.ai) through `litellm_proxy/*`, so this instance inherits all upstream models and provider credentials without copying any keys. Auto routers defined here (`moe-router`) pick between those upstream models and the router code running is whatever is on this branch
## Image
@ -18,7 +18,7 @@ curl localhost:4000/v1/chat/completions -H "Authorization: Bearer $LITELLM_MASTE
-d '{"model":"moe-router","messages":[{"role":"user","content":"hi"}]}' -i | grep x-litellm-model-name
```
Without Docker: `uv run --no-sync litellm --config deploy/autorouter-sandbox/config.yaml --port 4000`
Without Docker: `uv run --no-sync litellm --config deploy/autorouter-sandbox/proxy_config.yaml --port 4000`
## Deploy

View file

@ -0,0 +1,39 @@
model_list:
- model_name: "*"
litellm_params:
model: litellm_proxy/*
api_base: os.environ/UPSTREAM_LITELLM_BASE_URL
api_key: os.environ/UPSTREAM_LITELLM_API_KEY
- model_name: moe-router
litellm_params:
model: auto_router/complexity_router
complexity_router_default_model: anthropic/claude-sonnet-5
complexity_router_config:
classifier_type: llm
classifier_llm_config:
model: openai/gpt-4o-mini
timeout_ms: 3000
classification_rubric: agentic
classifier_context_window_size: 35
classifier_context_per_turn_chars: 500
deployment_affinity: true
session_affinity: false
escalation_keywords:
- LITELLM ESCALATE
tiers:
SIMPLE:
- anthropic/claude-haiku-4-5
MEDIUM:
- anthropic/claude-sonnet-5
COMPLEX:
- anthropic/claude-opus-5
REASONING:
- anthropic/claude-fable-5
general_settings:
master_key: os.environ/LITELLM_MASTER_KEY
litellm_settings:
drop_params: true
telemetry: false