Repository navigation
Releases: TEA-ching/continue
Releases · TEA-ching/continue
Release list
typo keyPoollive !
KeytoolLive with dashboard
v1.3.39c-vscode
v1.3.39b-vscode
v1.3.39a-vscode
v1.3.39-vscode
KeypoolLive — Shared Encrypted API Key Vault
This fork adds KeypoolLive, a system that loads AI provider API keys from an encrypted vault file hosted at a public URL. Keys are shared across a team, rotated round-robin, and never stored in plain text in version control.
How it works
- An AES-256-CBC encrypted vault (
ai.json.enc, produced withopenssl enc -aes-256-cbc -pbkdf2) is hosted at a URL you control. - Continue fetches and decrypts the vault at startup using a secret you provide.
- All models in the vault are injected into the Continue model selector automatically — no manual
config.yamlentries per model needed. - Optionally, all requests are routed through a Cloudflare AI Gateway for unified logging, rate limiting, and caching.
Configuration
Add a keypoollive block to ~/.continue/config.yaml:
keypoollive:
vaultUrl: https://your-host/ai.json.enc # URL of the encrypted vault
secret: <decryption-secret> # AES decryption passphrase
useGateway: true # Route via Cloudflare AI Gateway
gatewaySecret: cfut_xxxxxxxxxxxxxxxxxxxx # cf-aig-authorization token
gatewayId: <your-cloudflare-gateway-id> # Cloudflare account/gateway IDAll fields can be overridden with environment variables (env takes precedence over YAML):
| YAML field | Environment variable |
|---|---|
vaultUrl |
KEYPOOL_LIVE_VAULT_URL |
secret |
KEYPOOL_LIVE_SECRET |
useGateway |
KEYPOOL_LIVE_USE_GATEWAY (true/false) |
gatewaySecret |
KEYPOOL_LIVE_GATEWAY_SECRET |
gatewayId |
KEYPOOL_LIVE_GATEWAY_ID |
Sample ai.json vault format
{
"version": 1,
"providers": {
"anthropic": {
"protocol": "anthropic",
"gatewayEndpoint": "https://gateway.ai.cloudflare.com/v1/cloudflare_account_id/default/compat",
"gatewayModelPrefix": "anthropic",
"endpoint": "https://api.anthropic.com/v1",
"keys": [
{
"key": "sk-ant-api03-EXAMPLE_KEY_1234567890",
"owner": "john.doe@example.com",
"type": "free"
},
{
"key": "sk-ant-api03-EXAMPLE_KEY_0987654321",
"owner": "jane.smith@example.com",
"type": "paid"
}
],
"models": [
{
"id": "claude-haiku-4-5-20251001",
"usage": "chat",
"contextWindow": 200000,
"maxOutputTokens": 64000,
"tpmLimit": null,
"priority": 200,
"tags": ["fast", "cheap", "code"]
},
{
"id": "claude-sonnet-4-6",
"usage": "chat",
"contextWindow": 1000000,
"maxOutputTokens": 64000,
"tpmLimit": null,
"priority": 210,
"tags": ["code", "large", "reasoning"]
},
{
"id": "claude-opus-4-6",
"usage": "chat",
"contextWindow": 1000000,
"maxOutputTokens": 128000,
"tpmLimit": null,
"priority": 220,
"tags": ["powerful", "large", "reasoning"]
}
]
},
"gemini": {
"protocol": "gemini",
"gatewayEndpoint": "https://gateway.ai.cloudflare.com/v1/cloudflare_account_id/default/compat",
"gatewayModelPrefix": "google-ai-studio",
"endpoint": "https://generativelanguage.googleapis.com/v1beta",
"keys": [
{
"key": "AIzaSyEXAMPLE_KEY_1234567890",
"owner": "demo.user@example.com|projects/1234567890",
"type": "free"
},
{
"key": "AIzaSyEXAMPLE_KEY_0987654321",
"owner": "test.user@example.com|projects/0987654321",
"type": "free"
}
],
"models": [
{
"id": "gemini-embedding-2-preview",
"contextWindow": 8192,
"maxOutputTokens": 0,
"tpmLimit": null,
"priority": 1,
"usage": "embedding",
"defaultDimensions": 256,
"tags": ["embedding"]
},
{
"id": "gemini-3-flash-preview",
"contextWindow": 1048576,
"maxOutputTokens": 65536,
"tpmLimit": null,
"priority": 104,
"usage": "chat",
"tags": ["flash", "powerful", "large"]
}
]
},
"groq": {
"protocol": "openai",
"gatewayEndpoint": "https://gateway.ai.cloudflare.com/v1/cloudflare_account_id/default/compat",
"gatewayModelPrefix": "groq",
"endpoint": "https://api.groq.com/openai/v1",
"keys": [
{
"key": "gsk_EXAMPLE_KEY_1234567890",
"owner": "github|demo_user",
"type": "free"
},
{
"key": "gsk_EXAMPLE_KEY_0987654321",
"owner": "demo.user@example.com",
"type": "free"
}
],
"models": [
{
"id": "llama-3.3-70b-versatile",
"usage": "chat",
"contextWindow": 131072,
"maxOutputTokens": 32768,
"tpmLimit": 12000,
"priority": 1,
"tags": ["fast", "code", "large"]
},
{
"id": "llama-3.1-8b-instant",
"usage": "chat",
"contextWindow": 131072,
"maxOutputTokens": 8192,
"tpmLimit": 6000,
"priority": 20,
"tags": ["fast", "cheap"]
}
]
},
"mistral": {
"protocol": "openai",
"gatewayEndpoint": "https://gateway.ai.cloudflare.com/v1/cloudflare_account_id/default/compat",
"gatewayModelPrefix": "mistral",
"endpoint": "https://api.mistral.ai/v1",
"keys": [
{
"key": "EXAMPLE_KEY_1234567890",
"owner": "demo.user@example.com|+1234567890",
"type": "free"
},
{
"key": "EXAMPLE_KEY_0987654321",
"owner": "test.user@example.com|+0987654321",
"type": "free"
}
],
"models": [
{
"id": "codestral-latest",
"usage": "chat",
"contextWindow": 256000,
"maxOutputTokens": 256000,
"tpmLimit": 500000,
"priority": 60,
"tags": ["codestral", "small", "code"]
},
{
"id": "mistral-medium-latest",
"usage": "chat",
"contextWindow": 128000,
"maxOutputTokens": 128000,
"tpmLimit": 500000,
"priority": 50,
"tags": ["mistral", "medium", "fast"]
}
]
},
"openrouter": {
"protocol": "openai",
"gatewayEndpoint": "https://gateway.ai.cloudflare.com/v1/cloudflare_account_id/default/compat",
"gatewayModelPrefix": "openrouter",
"endpoint": "https://openrouter.ai/api/v1",
"keys": [
{
"key": "sk-or-v1-EXAMPLE_KEY_1234567890",
"owner": "demo.user@example.com",
"type": "free"
},
{
"key": "sk-or-v1-EXAMPLE_KEY_0987654321",
"owner": "test.user@example.com",
"type": "free"
}
],
"models": [
{
"id": "qwen/qwen3.5-flash-02-23",
"usage": "chat",
"contextWindow": 131072,
"maxOutputTokens": 65536,
"tpmLimit": null,
"priority": 150,
"tags": ["openrouter", "qwen", "chat"]
},
{
"id": "qwen/qwen3-coder:free",
"usage": "chat",
"contextWindow": 262000,
"maxOutputTokens": 262000,
"tpmLimit": null,
"priority": 80,
"tags": ["openrouter", "qwen", "chat"]
}
]
},
"sambanova": {
"protocol": "openai",
"gatewayEndpoint": "https://gateway.ai.cloudflare.com/v1/cloudflare_account_id/default/compat",
"gatewayModelPrefix": "custom-sambanova",
"endpoint": "https://api.sambanova.com/v1",
"keys": [
{
"key": "EXAMPLE_KEY_1234567890",
"owner": "demo.user@example.com",
"type": "free"
},
{
"key": "EXAMPLE_KEY_0987654321",
"owner": "test.user@example.com",
"type": "free"
}
],
"models": [
{
"id": "DeepSeek-V3.1",
"usage": "chat",
"contextWindow": 131072,
"maxOutputTokens": 65536,
"tpmLimit": null,
"priority": 100,
"tags": ["sambanova", "chat"]
},
{
"id": "Meta-Llama-3.3-70B-Instruct",
"usage": "chat",
"contextWindow": 131072,
"maxOutputTokens": 65536,
"tpmLimit": null,
"priority": 100,
"tags": ["sambanova", "chat"]
}
]
}
}
}Direct mode vs Gateway mode
Direct mode (useGateway: false or gateway credentials absent): each model connects directly to the provider's API endpoint using a provider API key selected round-robin from the vault.
Gateway mode (useGateway: true + gatewaySecret + gatewayId): all requests go through the Cloudflare AI Gateway endpoint. The gateway is OpenAI-compatible, so all providers are exposed as a single OpenAI endpoint with the model prefixed by the provider name (e.g. groq/llama-3.3-70b-instruct). Two authorization headers are sent:
cf-aig-authorization: Bearer <gatewaySecret>— Cloudflare gateway tokenAuthorization: Bearer <provider-api-key>— provider key selected from the vault
Vault format
The vault is a JSON file (ai.json) encrypted with:
openssl enc -aes-256-cbc -a -pbkdf2 -iter 100000 -salt \
-in ai.json -out ai.json.enc -pass pass:"${SECRET}"...