diff --git a/Dockerfile b/Dockerfile index df60248..67cfa22 100644 --- a/Dockerfile +++ b/Dockerfile @@ -1,4 +1,4 @@ -FROM ghcr.io/berriai/litellm:main-v1.43.16-stable +FROM ghcr.io/berriai/litellm:main-v1.50.4-stable WORKDIR /app COPY config.yaml . diff --git a/README.md b/README.md index c7f330e..07dc48f 100644 --- a/README.md +++ b/README.md @@ -10,15 +10,3 @@ - `AZURE_API_KEY_5` - `AZURE_API_KEY_6` - `ANTHROPIC_API_KEY` -- `ANYSCALE_API_KEY` -- `TOGETHER_API_KEY` -- `FIREWORKS_API_KEY` -- `MASTER_KEY` -- `REDIS_HOST` -- `REDIS_PASSWORD` -- `REDIS_PORT` -- `REDIS_SSL` -- `SLACK_WEBHOOK_URL` -- `SLACK_DAILY_REPORT_FREQUENCY` -- `OTEL_ENDPOINT` -- `OTEL_HEADERS` diff --git a/config.yaml b/config.yaml index 1a7e582..b6721ed 100644 --- a/config.yaml +++ b/config.yaml @@ -1,4 +1,3 @@ ---- model_list: - model_name: gpt-4o litellm_params: @@ -6,12 +5,6 @@ model_list: api_key: os.environ/OPENAI_API_KEY rpm: 500 tpm: 30000 - - model_name: text-embedding-3-small - litellm_params: - model: openai/text-embedding-3-small - api_key: os.environ/OPENAI_API_KEY - rpm: 3000 - tpm: 1000000 - model_name: gpt-4o litellm_params: model: azure/fastrepl-2-gpt-4o-2024-05-13 @@ -54,62 +47,14 @@ model_list: api_key: os.environ/AZURE_API_KEY_8 rpm: 2700 tpm: 450000 - - model_name: text-embedding-3-small - litellm_params: - model: azure/fastrepl-1-text-embedding-3-small - api_base: https://fastrepl-1.openai.azure.com - api_key: os.environ/AZURE_API_KEY_1 - tpm: 350000 - - model_name: text-embedding-3-small - litellm_params: - model: azure/fastrepl-2-text-embedding-3-small - api_base: https://fastrepl-2.openai.azure.com - api_key: os.environ/AZURE_API_KEY_2 - tpm: 350000 - - model_name: text-embedding-3-small - litellm_params: - model: azure/fastrepl-3-text-embedding-3-small - api_base: https://fastrepl-3.openai.azure.com - api_key: os.environ/AZURE_API_KEY_3 - tpm: 350000 - - model_name: claude-3-haiku - litellm_params: - model: anthropic/claude-3-haiku-20240307 - api_key: os.environ/ANTHROPIC_API_KEY - rpm: 1000 - tpm: 100000 - model_name: claude-3-sonnet litellm_params: - model: anthropic/claude-3-sonnet-20240229 + model: anthropic/claude-3-5-sonnet-20241022 api_key: os.environ/ANTHROPIC_API_KEY rpm: 1000 tpm: 80000 - - model_name: claude-3-opus - litellm_params: - model: anthropic/claude-3-sonnet-20240229 - api_key: os.environ/ANTHROPIC_API_KEY - rpm: 1000 - tpm: 40000 - - model_name: llama-3.1-8b - litellm_params: - model: fireworks_ai/accounts/fireworks/models/llama-v3p1-8b-instruct - api_key: os.environ/FIREWORKS_API_KEY - - model_name: llama-3.1-70b - litellm_params: - model: fireworks_ai/accounts/fireworks/models/llama-v3p1-70b-instruct - api_key: os.environ/FIREWORKS_API_KEY - # - model_name: llama-3.1-8b - # litellm_params: - # model: together_ai/meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo - # api_key: os.environ/TOGETHER_API_KEY - # - model_name: llama-3.1-70b - # litellm_params: - # model: together_ai/meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo - # api_key: os.environ/TOGETHER_API_KEY litellm_settings: drop_params: true - callbacks: - - otel router_settings: routing_strategy: simple-shuffle timeout: 60 @@ -119,5 +64,5 @@ router_settings: general_settings: master_key: os.environ/MASTER_KEY environment_variables: - AZURE_API_VERSION: 2024-05-01-preview + AZURE_API_VERSION: 2024-10-01-preview REDIS_SSL: 1