Merge GCP Vertex AI implementation from cg-dot/oai-reverse-proxy (khanon/oai-reverse-proxy!72)

This commit is contained in:
khanon
2024-08-05 14:27:51 +00:00
parent 29ed07492e
commit 0c936e97fe
25 changed files with 1133 additions and 13 deletions
+12 -3
View File
@@ -40,15 +40,21 @@ NODE_ENV=production
# Which model types users are allowed to access.
# The following model families are recognized:
# turbo | gpt4 | gpt4-32k | gpt4-turbo | gpt4o | dall-e | claude | claude-opus | gemini-flash | gemini-pro | gemini-ultra | mistral-tiny | mistral-small | mistral-medium | mistral-large | aws-claude | aws-claude-opus | azure-turbo | azure-gpt4 | azure-gpt4-32k | azure-gpt4-turbo | azure-gpt4o | azure-dall-e
# turbo | gpt4 | gpt4-32k | gpt4-turbo | gpt4o | dall-e | claude | claude-opus
# | gemini-flash | gemini-pro | gemini-ultra | mistral-tiny | mistral-small
# | mistral-medium | mistral-large | aws-claude | aws-claude-opus | gcp-claude
# | gcp-claude-opus | azure-turbo | azure-gpt4 | azure-gpt4-32k
# | azure-gpt4-turbo | azure-gpt4o | azure-dall-e
# By default, all models are allowed except for 'dall-e' / 'azure-dall-e'.
# To allow DALL-E image generation, uncomment the line below and add 'dall-e' or
# 'azure-dall-e' to the list of allowed model families.
# ALLOWED_MODEL_FAMILIES=turbo,gpt4,gpt4-32k,gpt4-turbo,gpt4o,claude,claude-opus,gemini-flash,gemini-pro,gemini-ultra,mistral-tiny,mistral-small,mistral-medium,mistral-large,aws-claude,aws-claude-opus,azure-turbo,azure-gpt4,azure-gpt4-32k,azure-gpt4-turbo,azure-gpt4o
# ALLOWED_MODEL_FAMILIES=turbo,gpt4,gpt4-32k,gpt4-turbo,gpt4o,claude,claude-opus,gemini-flash,gemini-pro,gemini-ultra,mistral-tiny,mistral-small,mistral-medium,mistral-large,aws-claude,aws-claude-opus,gcp-claude,gcp-claude-opus,azure-turbo,azure-gpt4,azure-gpt4-32k,azure-gpt4-turbo,azure-gpt4o
# Which services can be used to process prompts containing images via multimodal
# models. The following services are recognized:
# openai | anthropic | aws | azure | google-ai | mistral-ai
# openai | anthropic | aws | gcp | azure | google-ai | mistral-ai
# Do not enable this feature unless all users are trusted, as you will be liable
# for any user-submitted images containing illegal content.
# By default, no image services are allowed and image prompts are rejected.
@@ -118,6 +124,7 @@ NODE_ENV=production
# TOKEN_QUOTA_CLAUDE=0
# TOKEN_QUOTA_GEMINI_PRO=0
# TOKEN_QUOTA_AWS_CLAUDE=0
# TOKEN_QUOTA_GCP_CLAUDE=0
# "Tokens" for image-generation models are counted at a rate of 100000 tokens
# per US$1.00 generated, which is similar to the cost of GPT-4 Turbo.
# DALL-E 3 costs around US$0.10 per image (10000 tokens).
@@ -142,6 +149,7 @@ NODE_ENV=production
# You can add multiple API keys by separating them with a comma.
# For AWS credentials, separate the access key ID, secret key, and region with a colon.
# For GCP credentials, separate the project ID, client email, region, and private key with a colon.
OPENAI_KEY=sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
ANTHROPIC_KEY=sk-ant-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
GOOGLE_AI_KEY=AIzaxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
@@ -149,6 +157,7 @@ GOOGLE_AI_KEY=AIzaxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
AWS_CREDENTIALS=myaccesskeyid:mysecretkey:us-east-1,anotheraccesskeyid:anothersecretkey:us-west-2
# See `docs/azure-configuration.md` for more information, there may be additional steps required to set up Azure.
AZURE_CREDENTIALS=azure-resource-name:deployment-id:api-key,another-azure-resource-name:another-deployment-id:another-api-key
GCP_CREDENTIALS=project-id:client-email:region:private-key
# With proxy_key gatekeeper, the password users must provide to access the API.
# PROXY_KEY=your-secret-key