[Feat] Add CyberArk Secrets Manager Integration (#16278)

* KeyManagementSystem add cyberark

* add CyberArkSecretManager

* add CyberArkSecretManager

* add CyberArkSecretManager

* docs add CyberArkSecretManager

* docs

* refactor to use get_secret_from_manager

* Potential fix for code scanning alert no. 3645: Clear-text logging of sensitive information

Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>

* Potential fix for code scanning alert no. 3650: Clear-text logging of sensitive information

Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>

* Potential fix for code scanning alert no. 3649: Clear-text logging of sensitive information

Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>

* Potential fix for code scanning alert no. 3646: Clear-text logging of sensitive information

Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>

---------

Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
This commit is contained in:
Ishaan Jaff
2025-11-05 14:00:45 -08:00
committed by GitHub
co-authored by Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
parent f19356db60
commit 9a372bfad6
19 changed files with 1208 additions and 446 deletions
+11 -355
View File
@@ -1,8 +1,4 @@
import Tabs from '@theme/Tabs';
import TabItem from '@theme/TabItem';
import Image from '@theme/IdealImage';
# Secret Manager
# Secret Managers Overview
:::info
@@ -14,359 +10,19 @@ import Image from '@theme/IdealImage';
:::
LiteLLM supports **reading secrets (eg. `OPENAI_API_KEY`)** and **writing secrets (eg. Virtual Keys)** from Azure Key Vault, Google Secret Manager, Hashicorp Vault, and AWS Secret Manager.
LiteLLM supports **reading secrets (eg. `OPENAI_API_KEY`)** and **writing secrets (eg. Virtual Keys)** from Azure Key Vault, Google Secret Manager, Hashicorp Vault, CyberArk Conjur, and AWS Secret Manager.
## Supported Secret Managers
- AWS Key Management Service
- AWS Secret Manager
- [Azure Key Vault](#azure-key-vault)
- [Google Secret Manager](#google-secret-manager)
- Google Key Management Service
- [Hashicorp Vault](#hashicorp-vault)
## AWS Secret Manager
Store your proxy keys in AWS Secret Manager.
| Feature | Support | Description |
|---------|----------|-------------|
| Reading Secrets | ✅ | Read secrets e.g `OPENAI_API_KEY` |
| Writing Secrets | ✅ | Store secrets e.g `Virtual Keys` |
#### Proxy Usage
1. Save AWS Credentials in your environment
```bash
os.environ["AWS_ACCESS_KEY_ID"] = "" # Access key
os.environ["AWS_SECRET_ACCESS_KEY"] = "" # Secret access key
os.environ["AWS_REGION_NAME"] = "" # us-east-1, us-east-2, us-west-1, us-west-2
```
2. Enable AWS Secret Manager in config.
<Tabs>
<TabItem value="read_only" label="Read Keys from AWS Secret Manager">
```yaml
general_settings:
master_key: os.environ/litellm_master_key
key_management_system: "aws_secret_manager" # 👈 KEY CHANGE
key_management_settings:
hosted_keys: ["litellm_master_key"] # 👈 Specify which env keys you stored on AWS
```
</TabItem>
<TabItem value="write_only" label="Write Virtual Keys to AWS Secret Manager">
This will only store virtual keys in AWS Secret Manager. No keys will be read from AWS Secret Manager.
```yaml
general_settings:
key_management_system: "aws_secret_manager" # 👈 KEY CHANGE
key_management_settings:
store_virtual_keys: true # OPTIONAL. Defaults to False, when True will store virtual keys in secret manager
prefix_for_stored_virtual_keys: "litellm/" # OPTIONAL. If set, this prefix will be used for stored virtual keys in the secret manager
access_mode: "write_only" # Literal["read_only", "write_only", "read_and_write"]
description: "litellm virtual key" # OPTIONAL, if set will set this as the description for all virtual keys
tags: # OPTIONAL, if set will set this as the tags for all virtual keys
Environment: "Prod"
Owner: "AI Platform team"
```
</TabItem>
<TabItem value="read_and_write" label="Read + Write Keys with AWS Secret Manager">
```yaml
general_settings:
master_key: os.environ/litellm_master_key
key_management_system: "aws_secret_manager" # 👈 KEY CHANGE
key_management_settings:
store_virtual_keys: true # OPTIONAL. Defaults to False, when True will store virtual keys in secret manager
prefix_for_stored_virtual_keys: "litellm/" # OPTIONAL. If set, this prefix will be used for stored virtual keys in the secret manager
access_mode: "read_and_write" # Literal["read_only", "write_only", "read_and_write"]
hosted_keys: ["litellm_master_key"] # OPTIONAL. Specify which env keys you stored on AWS
```
</TabItem>
</Tabs>
3. Run proxy
```bash
litellm --config /path/to/config.yaml
```
#### Using K/V pairs in 1 AWS Secret
You can read multiple keys from a single AWS Secret using the `primary_secret_name` parameter:
```yaml
general_settings:
key_management_system: "aws_secret_manager"
key_management_settings:
hosted_keys: [
"OPENAI_API_KEY_MODEL_1",
"OPENAI_API_KEY_MODEL_2",
]
primary_secret_name: "litellm_secrets" # 👈 Read multiple keys from one JSON secret
```
The `primary_secret_name` allows you to read multiple keys from a single AWS Secret as a JSON object. For example, the "litellm_secrets" would contain:
```json
{
"OPENAI_API_KEY_MODEL_1": "sk-key1...",
"OPENAI_API_KEY_MODEL_2": "sk-key2..."
}
```
This reduces the number of AWS Secrets you need to manage.
## Hashicorp Vault
| Feature | Support | Description |
|---------|----------|-------------|
| Reading Secrets | ✅ | Read secrets e.g `OPENAI_API_KEY` |
| Writing Secrets | ✅ | Store secrets e.g `Virtual Keys` |
Read secrets from [Hashicorp Vault](https://developer.hashicorp.com/vault/docs/secrets/kv/kv-v2)
**Step 1.** Add Hashicorp Vault details in your environment
LiteLLM supports two methods of authentication:
1. TLS cert authentication - `HCP_VAULT_CLIENT_CERT` and `HCP_VAULT_CLIENT_KEY`
2. Token authentication - `HCP_VAULT_TOKEN`
```bash
HCP_VAULT_ADDR="https://test-cluster-public-vault-0f98180c.e98296b2.z1.hashicorp.cloud:8200"
HCP_VAULT_NAMESPACE="admin"
# Authentication via TLS cert
HCP_VAULT_CLIENT_CERT="path/to/client.pem"
HCP_VAULT_CLIENT_KEY="path/to/client.key"
# OR - Authentication via token
HCP_VAULT_TOKEN="hvs.CAESIG52gL6ljBSdmq*****"
# OPTIONAL
HCP_VAULT_REFRESH_INTERVAL="86400" # defaults to 86400, frequency of cache refresh for Hashicorp Vault
```
**Step 2.** Add to proxy config.yaml
```yaml
general_settings:
key_management_system: "hashicorp_vault"
# [OPTIONAL SETTINGS]
key_management_settings:
store_virtual_keys: true # OPTIONAL. Defaults to False, when True will store virtual keys in secret manager
prefix_for_stored_virtual_keys: "litellm/" # OPTIONAL. If set, this prefix will be used for stored virtual keys in the secret manager
access_mode: "read_and_write" # Literal["read_only", "write_only", "read_and_write"]
```
**Step 3.** Start + test proxy
```
$ litellm --config /path/to/config.yaml
```
[Quick Test Proxy](./proxy/user_keys)
#### How it works
**Reading Secrets**
LiteLLM reads secrets from Hashicorp Vault's KV v2 engine using the following URL format:
```
{VAULT_ADDR}/v1/{NAMESPACE}/secret/data/{SECRET_NAME}
```
For example, if you have:
- `HCP_VAULT_ADDR="https://vault.example.com:8200"`
- `HCP_VAULT_NAMESPACE="admin"`
- Secret name: `AZURE_API_KEY`
LiteLLM will look up:
```
https://vault.example.com:8200/v1/admin/secret/data/AZURE_API_KEY
```
#### Expected Secret Format
LiteLLM expects all secrets to be stored as a JSON object with a `key` field containing the secret value.
For example, for `AZURE_API_KEY`, the secret should be stored as:
```json
{
"key": "sk-1234"
}
```
<Image img={require('../img/hcorp.png')} />
**Writing Secrets**
When a Virtual Key is Created / Deleted on LiteLLM, LiteLLM will automatically create / delete the secret in Hashicorp Vault.
- Create Virtual Key on LiteLLM either through the LiteLLM Admin UI or API
<Image img={require('../img/hcorp_create_virtual_key.png')} />
- Check Hashicorp Vault for secret
LiteLLM stores secret under the `prefix_for_stored_virtual_keys` path (default: `litellm/`)
<Image img={require('../img/hcorp_virtual_key.png')} />
## Azure Key Vault
#### Usage with LiteLLM Proxy Server
1. Install Proxy dependencies
```bash
pip install 'litellm[proxy]' 'litellm[extra_proxy]'
```
2. Save Azure details in your environment
```bash
export["AZURE_CLIENT_ID"]="your-azure-app-client-id"
export["AZURE_CLIENT_SECRET"]="your-azure-app-client-secret"
export["AZURE_TENANT_ID"]="your-azure-tenant-id"
export["AZURE_KEY_VAULT_URI"]="your-azure-key-vault-uri"
```
3. Add to proxy config.yaml
```yaml
model_list:
- model_name: "my-azure-models" # model alias
litellm_params:
model: "azure/<your-deployment-name>"
api_key: "os.environ/AZURE-API-KEY" # reads from key vault - get_secret("AZURE_API_KEY")
api_base: "os.environ/AZURE-API-BASE" # reads from key vault - get_secret("AZURE_API_BASE")
general_settings:
key_management_system: "azure_key_vault"
```
You can now test this by starting your proxy:
```bash
litellm --config /path/to/config.yaml
```
[Quick Test Proxy](./proxy/quick_start#using-litellm-proxy---curl-request-openai-package-langchain-langchain-js)
## Google Secret Manager
Support for [Google Secret Manager](https://cloud.google.com/security/products/secret-manager)
1. Save Google Secret Manager details in your environment
```shell
GOOGLE_SECRET_MANAGER_PROJECT_ID="your-project-id-on-gcp" # example: adroit-crow-413218
```
Optional Params
```shell
export GOOGLE_SECRET_MANAGER_REFRESH_INTERVAL = "" # (int) defaults to 86400
export GOOGLE_SECRET_MANAGER_ALWAYS_READ_SECRET_MANAGER = "" # (str) set to "true" if you want to always read from google secret manager without using in memory caching. NOT RECOMMENDED in PROD
```
2. Add to proxy config.yaml
```yaml
model_list:
- model_name: fake-openai-endpoint
litellm_params:
model: openai/fake
api_base: https://exampleopenaiendpoint-production.up.railway.app/
api_key: os.environ/OPENAI_API_KEY # this will be read from Google Secret Manager
general_settings:
key_management_system: "google_secret_manager"
```
You can now test this by starting your proxy:
```bash
litellm --config /path/to/config.yaml
```
[Quick Test Proxy](./proxy/quick_start#using-litellm-proxy---curl-request-openai-package-langchain-langchain-js)
## Google Key Management Service
Use encrypted keys from Google KMS on the proxy
Step 1. Add keys to env
```
export GOOGLE_APPLICATION_CREDENTIALS="/path/to/credentials.json"
export GOOGLE_KMS_RESOURCE_NAME="projects/*/locations/*/keyRings/*/cryptoKeys/*"
export PROXY_DATABASE_URL_ENCRYPTED=b'\n$\x00D\xac\xb4/\x8e\xc...'
```
Step 2: Update Config
```yaml
general_settings:
key_management_system: "google_kms"
database_url: "os.environ/PROXY_DATABASE_URL_ENCRYPTED"
master_key: sk-1234
```
Step 3: Start + test proxy
```
$ litellm --config /path/to/config.yaml
```
And in another terminal
```
$ litellm --test
```
[Quick Test Proxy](./proxy/user_keys)
<!--
## .env Files
If no secret manager client is specified, Litellm automatically uses the `.env` file to manage sensitive data. -->
## AWS Key Management V1
:::tip
[BETA] AWS Key Management v2 is on the enterprise tier. Go [here for docs](./proxy/enterprise.md#beta-aws-key-manager---key-decryption)
:::
Use AWS KMS to storing a hashed copy of your Proxy Master Key in the environment.
```bash
export LITELLM_MASTER_KEY="djZ9xjVaZ..." # 👈 ENCRYPTED KEY
export AWS_REGION_NAME="us-west-2"
```
```yaml
general_settings:
key_management_system: "aws_kms"
key_management_settings:
hosted_keys: ["LITELLM_MASTER_KEY"] # 👈 WHICH KEYS ARE STORED ON KMS
```
[**See Decryption Code**](https://github.com/BerriAI/litellm/blob/a2da2a8f168d45648b61279d4795d647d94f90c9/litellm/utils.py#L10182)
## **All Secret Manager Settings**
- [AWS Key Management Service](./secret_managers/aws_kms)
- [AWS Secret Manager](./secret_managers/aws_secret_manager)
- [Azure Key Vault](./secret_managers/azure_key_vault)
- [CyberArk Conjur](./secret_managers/cyberark)
- [Google Secret Manager](./secret_managers/google_secret_manager)
- [Google Key Management Service](./secret_managers/google_kms)
- [Hashicorp Vault](./secret_managers/hashicorp_vault)
## All Secret Manager Settings
All settings related to secret management
@@ -0,0 +1,34 @@
# AWS Key Management V1
:::info
**This is an Enterprise Feature**
[Enterprise Pricing](https://www.litellm.ai/#pricing)
[Contact us here to get a free trial](https://calendly.com/d/4mp-gd3-k5k/litellm-1-1-onboarding-chat)
:::
:::tip
[BETA] AWS Key Management v2 is on the enterprise tier. Go [here for docs](../proxy/enterprise.md#beta-aws-key-manager---key-decryption)
:::
Use AWS KMS to storing a hashed copy of your Proxy Master Key in the environment.
```bash
export LITELLM_MASTER_KEY="djZ9xjVaZ..." # 👈 ENCRYPTED KEY
export AWS_REGION_NAME="us-west-2"
```
```yaml
general_settings:
key_management_system: "aws_kms"
key_management_settings:
hosted_keys: ["LITELLM_MASTER_KEY"] # 👈 WHICH KEYS ARE STORED ON KMS
```
[**See Decryption Code**](https://github.com/BerriAI/litellm/blob/a2da2a8f168d45648b61279d4795d647d94f90c9/litellm/utils.py#L10182)
@@ -0,0 +1,112 @@
import Tabs from '@theme/Tabs';
import TabItem from '@theme/TabItem';
# AWS Secret Manager
:::info
**This is an Enterprise Feature**
[Enterprise Pricing](https://www.litellm.ai/#pricing)
[Contact us here to get a free trial](https://calendly.com/d/4mp-gd3-k5k/litellm-1-1-onboarding-chat)
:::
Store your proxy keys in AWS Secret Manager.
| Feature | Support | Description |
|---------|----------|-------------|
| Reading Secrets | ✅ | Read secrets e.g `OPENAI_API_KEY` |
| Writing Secrets | ✅ | Store secrets e.g `Virtual Keys` |
## Proxy Usage
1. Save AWS Credentials in your environment
```bash
os.environ["AWS_ACCESS_KEY_ID"] = "" # Access key
os.environ["AWS_SECRET_ACCESS_KEY"] = "" # Secret access key
os.environ["AWS_REGION_NAME"] = "" # us-east-1, us-east-2, us-west-1, us-west-2
```
2. Enable AWS Secret Manager in config.
<Tabs>
<TabItem value="read_only" label="Read Keys from AWS Secret Manager">
```yaml
general_settings:
master_key: os.environ/litellm_master_key
key_management_system: "aws_secret_manager" # 👈 KEY CHANGE
key_management_settings:
hosted_keys: ["litellm_master_key"] # 👈 Specify which env keys you stored on AWS
```
</TabItem>
<TabItem value="write_only" label="Write Virtual Keys to AWS Secret Manager">
This will only store virtual keys in AWS Secret Manager. No keys will be read from AWS Secret Manager.
```yaml
general_settings:
key_management_system: "aws_secret_manager" # 👈 KEY CHANGE
key_management_settings:
store_virtual_keys: true # OPTIONAL. Defaults to False, when True will store virtual keys in secret manager
prefix_for_stored_virtual_keys: "litellm/" # OPTIONAL. If set, this prefix will be used for stored virtual keys in the secret manager
access_mode: "write_only" # Literal["read_only", "write_only", "read_and_write"]
description: "litellm virtual key" # OPTIONAL, if set will set this as the description for all virtual keys
tags: # OPTIONAL, if set will set this as the tags for all virtual keys
Environment: "Prod"
Owner: "AI Platform team"
```
</TabItem>
<TabItem value="read_and_write" label="Read + Write Keys with AWS Secret Manager">
```yaml
general_settings:
master_key: os.environ/litellm_master_key
key_management_system: "aws_secret_manager" # 👈 KEY CHANGE
key_management_settings:
store_virtual_keys: true # OPTIONAL. Defaults to False, when True will store virtual keys in secret manager
prefix_for_stored_virtual_keys: "litellm/" # OPTIONAL. If set, this prefix will be used for stored virtual keys in the secret manager
access_mode: "read_and_write" # Literal["read_only", "write_only", "read_and_write"]
hosted_keys: ["litellm_master_key"] # OPTIONAL. Specify which env keys you stored on AWS
```
</TabItem>
</Tabs>
3. Run proxy
```bash
litellm --config /path/to/config.yaml
```
## Using K/V pairs in 1 AWS Secret
You can read multiple keys from a single AWS Secret using the `primary_secret_name` parameter:
```yaml
general_settings:
key_management_system: "aws_secret_manager"
key_management_settings:
hosted_keys: [
"OPENAI_API_KEY_MODEL_1",
"OPENAI_API_KEY_MODEL_2",
]
primary_secret_name: "litellm_secrets" # 👈 Read multiple keys from one JSON secret
```
The `primary_secret_name` allows you to read multiple keys from a single AWS Secret as a JSON object. For example, the "litellm_secrets" would contain:
```json
{
"OPENAI_API_KEY_MODEL_1": "sk-key1...",
"OPENAI_API_KEY_MODEL_2": "sk-key2..."
}
```
This reduces the number of AWS Secrets you need to manage.
@@ -0,0 +1,47 @@
# Azure Key Vault
:::info
**This is an Enterprise Feature**
[Enterprise Pricing](https://www.litellm.ai/#pricing)
[Contact us here to get a free trial](https://calendly.com/d/4mp-gd3-k5k/litellm-1-1-onboarding-chat)
:::
## Usage with LiteLLM Proxy Server
1. Install Proxy dependencies
```bash
pip install 'litellm[proxy]' 'litellm[extra_proxy]'
```
2. Save Azure details in your environment
```bash
export["AZURE_CLIENT_ID"]="your-azure-app-client-id"
export["AZURE_CLIENT_SECRET"]="your-azure-app-client-secret"
export["AZURE_TENANT_ID"]="your-azure-tenant-id"
export["AZURE_KEY_VAULT_URI"]="your-azure-key-vault-uri"
```
3. Add to proxy config.yaml
```yaml
model_list:
- model_name: "my-azure-models" # model alias
litellm_params:
model: "azure/<your-deployment-name>"
api_key: "os.environ/AZURE-API-KEY" # reads from key vault - get_secret("AZURE_API_KEY")
api_base: "os.environ/AZURE-API-BASE" # reads from key vault - get_secret("AZURE_API_BASE")
general_settings:
key_management_system: "azure_key_vault"
```
You can now test this by starting your proxy:
```bash
litellm --config /path/to/config.yaml
```
[Quick Test Proxy](../proxy/quick_start#using-litellm-proxy---curl-request-openai-package-langchain-langchain-js)
@@ -0,0 +1,163 @@
# CyberArk Conjur
import Image from '@theme/IdealImage';
:::info
**This is an Enterprise Feature**
[Enterprise Pricing](https://www.litellm.ai/#pricing)
[Contact us here to get a free trial](https://calendly.com/d/4mp-gd3-k5k/litellm-1-1-onboarding-chat)
:::
| Feature | Support | Description |
|---------|----------|-------------|
| Reading Secrets | ✅ | Read secrets e.g `OPENAI_API_KEY` |
| Writing Secrets | ✅ | Store secrets e.g `Virtual Keys` |
| Deleting Secrets | ❌ | Secrets must be removed via policy updates |
Read and write secrets from [CyberArk Conjur](https://www.cyberark.com/products/secrets-management/) (self-hosted secrets manager)
**Step 1.** Add CyberArk Conjur details in your environment
LiteLLM supports two methods of authentication:
1. API key authentication - `CYBERARK_API_KEY` (recommended)
2. Certificate authentication - `CYBERARK_CLIENT_CERT` and `CYBERARK_CLIENT_KEY`
```bash
CYBERARK_API_BASE="http://your-conjur-instance:8080"
CYBERARK_ACCOUNT="default"
CYBERARK_USERNAME="admin"
# Authentication via API key (recommended)
CYBERARK_API_KEY="your-api-key-here"
# OR - Authentication via certificate
CYBERARK_CLIENT_CERT="path/to/client.pem"
CYBERARK_CLIENT_KEY="path/to/client.key"
# OPTIONAL
CYBERARK_REFRESH_INTERVAL="300" # defaults to 300 seconds (5 minutes), frequency of token refresh
```
**Step 2.** Add to proxy config.yaml
```yaml
general_settings:
key_management_system: "cyberark"
# [OPTIONAL SETTINGS]
key_management_settings:
store_virtual_keys: true # OPTIONAL. Defaults to False, when True will store virtual keys in secret manager
prefix_for_stored_virtual_keys: "litellm/" # OPTIONAL. If set, this prefix will be used for stored virtual keys in the secret manager
access_mode: "read_and_write" # Literal["read_only", "write_only", "read_and_write"]
```
**Step 3.** Start + test proxy
```bash
$ litellm --config /path/to/config.yaml
```
[Quick Test Proxy](../proxy/user_keys)
## Writing Virtual Keys to CyberArk
When you create a virtual key in the LiteLLM UI, it automatically gets stored in CyberArk Conjur.
**Step 1:** Create a virtual key in the LiteLLM Admin UI
In this example, we create a key named `litellm-cyber-ark-secret-key`:
<Image img={require('../../static/img/cyberark1.png')} alt="Creating virtual key in LiteLLM UI" />
**Step 2:** Verify the secret exists in CyberArk
You can verify the virtual key was stored in CyberArk by querying the secrets API:
```bash
TOKEN=$(curl -s -X POST http://0.0.0.0:8080/authn/default/admin/authenticate \
-d "your-api-key" | base64 | tr -d '\n')
curl -H "Authorization: Token token=\"$TOKEN\"" \
"http://0.0.0.0:8080/resources/default/variable" | jq .
```
The response shows `litellm-cyber-ark-secret-key` exists in CyberArk:
<Image img={require('../../static/img/cyberark2.png')} alt="Virtual key stored in CyberArk API" />
The virtual key is stored with the full path: `default:variable:litellm/litellm-cyber-ark-secret-key`
## How it works
**Authentication**
CyberArk Conjur uses a two-step authentication process:
1. LiteLLM authenticates with your API key to get a session token
2. The session token (base64-encoded) is used for subsequent API requests
3. Tokens expire after ~8 minutes, so LiteLLM caches and refreshes them automatically
**Reading Secrets**
LiteLLM reads secrets from CyberArk Conjur using the following URL format:
```
{CYBERARK_API_BASE}/secrets/{ACCOUNT}/variable/{SECRET_NAME}
```
For example, if you have:
- `CYBERARK_API_BASE="http://conjur.example.com:8080"`
- `CYBERARK_ACCOUNT="default"`
- Secret name: `AZURE_API_KEY`
LiteLLM will look up:
```
http://conjur.example.com:8080/secrets/default/variable/AZURE_API_KEY
```
**Writing Secrets**
When a Virtual Key is created on LiteLLM, the following happens automatically:
1. LiteLLM creates a policy entry to define the variable in Conjur (if it doesn't exist)
2. LiteLLM sets the secret value via the Conjur API
LiteLLM stores secrets under the `prefix_for_stored_virtual_keys` path (default: `litellm/`)
For example, a virtual key would be stored as: `litellm/virtual-key-name`
**Working curl examples**
Authenticate and get a token:
```bash
TOKEN=$(curl -s -X POST http://conjur.example.com:8080/authn/default/admin/authenticate \
-d "your-api-key" | base64 | tr -d '\n')
```
Read a secret:
```bash
curl -H "Authorization: Token token=\"$TOKEN\"" \
"http://conjur.example.com:8080/secrets/default/variable/test-secret"
```
Write a secret:
```bash
curl -X POST \
-H "Authorization: Token token=\"$TOKEN\"" \
--data "my-secret-value" \
"http://conjur.example.com:8080/secrets/default/variable/test-secret"
```
**Important Notes**
- Variables must be defined in a Conjur policy before setting their values
- LiteLLM automatically creates policy entries when writing new secrets
- Secret names with slashes (e.g., `litellm/key`) are automatically URL-encoded
- CyberArk Conjur does not support direct secret deletion via API (must use policy updates)
- Session tokens are cached for 5 minutes by default to minimize API calls
@@ -0,0 +1,43 @@
# Google Key Management Service
:::info
**This is an Enterprise Feature**
[Enterprise Pricing](https://www.litellm.ai/#pricing)
[Contact us here to get a free trial](https://calendly.com/d/4mp-gd3-k5k/litellm-1-1-onboarding-chat)
:::
Use encrypted keys from Google KMS on the proxy
Step 1. Add keys to env
```
export GOOGLE_APPLICATION_CREDENTIALS="/path/to/credentials.json"
export GOOGLE_KMS_RESOURCE_NAME="projects/*/locations/*/keyRings/*/cryptoKeys/*"
export PROXY_DATABASE_URL_ENCRYPTED=b'\n$\x00D\xac\xb4/\x8e\xc...'
```
Step 2: Update Config
```yaml
general_settings:
key_management_system: "google_kms"
database_url: "os.environ/PROXY_DATABASE_URL_ENCRYPTED"
master_key: sk-1234
```
Step 3: Start + test proxy
```
$ litellm --config /path/to/config.yaml
```
And in another terminal
```
$ litellm --test
```
[Quick Test Proxy](../proxy/user_keys)
@@ -0,0 +1,47 @@
# Google Secret Manager
:::info
**This is an Enterprise Feature**
[Enterprise Pricing](https://www.litellm.ai/#pricing)
[Contact us here to get a free trial](https://calendly.com/d/4mp-gd3-k5k/litellm-1-1-onboarding-chat)
:::
Support for [Google Secret Manager](https://cloud.google.com/security/products/secret-manager)
1. Save Google Secret Manager details in your environment
```shell
GOOGLE_SECRET_MANAGER_PROJECT_ID="your-project-id-on-gcp" # example: adroit-crow-413218
```
Optional Params
```shell
export GOOGLE_SECRET_MANAGER_REFRESH_INTERVAL = "" # (int) defaults to 86400
export GOOGLE_SECRET_MANAGER_ALWAYS_READ_SECRET_MANAGER = "" # (str) set to "true" if you want to always read from google secret manager without using in memory caching. NOT RECOMMENDED in PROD
```
2. Add to proxy config.yaml
```yaml
model_list:
- model_name: fake-openai-endpoint
litellm_params:
model: openai/fake
api_base: https://exampleopenaiendpoint-production.up.railway.app/
api_key: os.environ/OPENAI_API_KEY # this will be read from Google Secret Manager
general_settings:
key_management_system: "google_secret_manager"
```
You can now test this by starting your proxy:
```bash
litellm --config /path/to/config.yaml
```
[Quick Test Proxy](../proxy/quick_start#using-litellm-proxy---curl-request-openai-package-langchain-langchain-js)
@@ -0,0 +1,115 @@
import Image from '@theme/IdealImage';
# Hashicorp Vault
:::info
**This is an Enterprise Feature**
[Enterprise Pricing](https://www.litellm.ai/#pricing)
[Contact us here to get a free trial](https://calendly.com/d/4mp-gd3-k5k/litellm-1-1-onboarding-chat)
:::
| Feature | Support | Description |
|---------|----------|-------------|
| Reading Secrets | ✅ | Read secrets e.g `OPENAI_API_KEY` |
| Writing Secrets | ✅ | Store secrets e.g `Virtual Keys` |
Read secrets from [Hashicorp Vault](https://developer.hashicorp.com/vault/docs/secrets/kv/kv-v2)
**Step 1.** Add Hashicorp Vault details in your environment
LiteLLM supports two methods of authentication:
1. TLS cert authentication - `HCP_VAULT_CLIENT_CERT` and `HCP_VAULT_CLIENT_KEY`
2. Token authentication - `HCP_VAULT_TOKEN`
```bash
HCP_VAULT_ADDR="https://test-cluster-public-vault-0f98180c.e98296b2.z1.hashicorp.cloud:8200"
HCP_VAULT_NAMESPACE="admin"
# Authentication via TLS cert
HCP_VAULT_CLIENT_CERT="path/to/client.pem"
HCP_VAULT_CLIENT_KEY="path/to/client.key"
# OR - Authentication via token
HCP_VAULT_TOKEN="hvs.CAESIG52gL6ljBSdmq*****"
# OPTIONAL
HCP_VAULT_REFRESH_INTERVAL="86400" # defaults to 86400, frequency of cache refresh for Hashicorp Vault
```
**Step 2.** Add to proxy config.yaml
```yaml
general_settings:
key_management_system: "hashicorp_vault"
# [OPTIONAL SETTINGS]
key_management_settings:
store_virtual_keys: true # OPTIONAL. Defaults to False, when True will store virtual keys in secret manager
prefix_for_stored_virtual_keys: "litellm/" # OPTIONAL. If set, this prefix will be used for stored virtual keys in the secret manager
access_mode: "read_and_write" # Literal["read_only", "write_only", "read_and_write"]
```
**Step 3.** Start + test proxy
```
$ litellm --config /path/to/config.yaml
```
[Quick Test Proxy](../proxy/user_keys)
## How it works
**Reading Secrets**
LiteLLM reads secrets from Hashicorp Vault's KV v2 engine using the following URL format:
```
{VAULT_ADDR}/v1/{NAMESPACE}/secret/data/{SECRET_NAME}
```
For example, if you have:
- `HCP_VAULT_ADDR="https://vault.example.com:8200"`
- `HCP_VAULT_NAMESPACE="admin"`
- Secret name: `AZURE_API_KEY`
LiteLLM will look up:
```
https://vault.example.com:8200/v1/admin/secret/data/AZURE_API_KEY
```
### Expected Secret Format
LiteLLM expects all secrets to be stored as a JSON object with a `key` field containing the secret value.
For example, for `AZURE_API_KEY`, the secret should be stored as:
```json
{
"key": "sk-1234"
}
```
<Image img={require('../../img/hcorp.png')} />
**Writing Secrets**
When a Virtual Key is Created / Deleted on LiteLLM, LiteLLM will automatically create / delete the secret in Hashicorp Vault.
- Create Virtual Key on LiteLLM either through the LiteLLM Admin UI or API
<Image img={require('../../img/hcorp_create_virtual_key.png')} />
- Check Hashicorp Vault for secret
LiteLLM stores secret under the `prefix_for_stored_virtual_keys` path (default: `litellm/`)
<Image img={require('../../img/hcorp_virtual_key.png')} />
@@ -0,0 +1,47 @@
# Secret Managers Overview
:::info
**This is an Enterprise Feature**
[Enterprise Pricing](https://www.litellm.ai/#pricing)
[Contact us here to get a free trial](https://calendly.com/d/4mp-gd3-k5k/litellm-1-1-onboarding-chat)
:::
LiteLLM supports **reading secrets (eg. `OPENAI_API_KEY`)** and **writing secrets (eg. Virtual Keys)** from Azure Key Vault, Google Secret Manager, Hashicorp Vault, CyberArk Conjur, and AWS Secret Manager.
## Supported Secret Managers
- [AWS Key Management Service](./aws_kms)
- [AWS Secret Manager](./aws_secret_manager)
- [Azure Key Vault](./azure_key_vault)
- [CyberArk Conjur](./cyberark)
- [Google Secret Manager](./google_secret_manager)
- [Google Key Management Service](./google_kms)
- [Hashicorp Vault](./hashicorp_vault)
## All Secret Manager Settings
All settings related to secret management
```yaml
general_settings:
key_management_system: "aws_secret_manager" # REQUIRED
key_management_settings:
# Storing Virtual Keys Settings
store_virtual_keys: true # OPTIONAL. Defaults to False, when True will store virtual keys in secret manager
prefix_for_stored_virtual_keys: "litellm/" # OPTIONAL.I f set, this prefix will be used for stored virtual keys in the secret manager
# Access Mode Settings
access_mode: "write_only" # OPTIONAL. Literal["read_only", "write_only", "read_and_write"]. Defaults to "read_only"
# Hosted Keys Settings
hosted_keys: ["litellm_master_key"] # OPTIONAL. Specify which env keys you stored on AWS
# K/V pairs in 1 AWS Secret Settings
primary_secret_name: "litellm_secrets" # OPTIONAL. Read multiple keys from one JSON secret on AWS Secret Manager
```
Binary file not shown.

After

Width:  |  Height:  |  Size: 605 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 980 KiB

+8 -1
View File
@@ -260,7 +260,14 @@ const sidebars = {
type: "category",
label: "Secret Managers",
items: [
"secret",
"secret_managers/overview",
"secret_managers/aws_secret_manager",
"secret_managers/aws_kms",
"secret_managers/azure_key_vault",
"secret_managers/cyberark",
"secret_managers/google_secret_manager",
"secret_managers/google_kms",
"secret_managers/hashicorp_vault",
"oidc"
]
},
+7
View File
@@ -1,3 +1,10 @@
general_settings:
key_management_system: "cyberark"
key_management_settings:
store_virtual_keys: true
prefix_for_stored_virtual_keys: "litellm/"
access_mode: "read_and_write"
model_list:
- model_name: aws/anthropic/bedrock-claude-3-5-sonnet-v1
litellm_params:
+6
View File
@@ -2658,6 +2658,12 @@ class ProxyConfig:
)
HashicorpSecretManager()
elif key_management_system == KeyManagementSystem.CYBERARK.value:
from litellm.secret_managers.cyberark_secret_manager import (
CyberArkSecretManager,
)
CyberArkSecretManager()
else:
raise ValueError("Invalid Key Management System selected")
@@ -0,0 +1,354 @@
import base64
import os
from typing import Any, Dict, Optional, Union
from urllib.parse import quote
import httpx
import litellm
from litellm._logging import verbose_logger
from litellm.caching import InMemoryCache
from litellm.llms.custom_httpx.http_handler import (
_get_httpx_client,
get_async_httpx_client,
httpxSpecialProvider,
)
from litellm.proxy._types import KeyManagementSystem
from .base_secret_manager import BaseSecretManager
class CyberArkSecretManager(BaseSecretManager):
def __init__(self):
from litellm.proxy.proxy_server import CommonProxyErrors, premium_user
# CyberArk Conjur-specific config
self.conjur_addr = os.getenv("CYBERARK_API_BASE", "http://127.0.0.1:8080")
self.conjur_account = os.getenv("CYBERARK_ACCOUNT", "default")
self.conjur_username = os.getenv("CYBERARK_USERNAME", "admin")
self.conjur_api_key = os.getenv("CYBERARK_API_KEY", "")
# Optional config for certificate-based auth
self.tls_cert_path = os.getenv("CYBERARK_CLIENT_CERT", "")
self.tls_key_path = os.getenv("CYBERARK_CLIENT_KEY", "")
# Validate environment
if not self.conjur_api_key and not (
self.tls_cert_path and self.tls_key_path
):
raise ValueError(
"Missing CyberArk credentials. Please set CYBERARK_API_KEY or both CYBERARK_CLIENT_CERT and CYBERARK_CLIENT_KEY in your environment."
)
litellm.secret_manager_client = self
litellm._key_management_system = KeyManagementSystem.CYBERARK
# Tokens expire after ~8 minutes, so we cache for 5 minutes to be safe
_refresh_interval = int(os.environ.get("CYBERARK_REFRESH_INTERVAL", "300"))
self.cache = InMemoryCache(default_ttl=_refresh_interval)
if premium_user is not True:
raise ValueError(
f"CyberArk secret manager is only available for premium users. {CommonProxyErrors.not_premium_user.value}"
)
def _authenticate(self) -> str:
"""
Authenticate with CyberArk Conjur and get a session token.
The token is a JSON object that must be base64-encoded for use in subsequent requests.
Returns:
str: Base64-encoded session token
"""
# Check if we have a cached token
cached_token = self.cache.get_cache("cyberark_auth_token")
if cached_token is not None:
return cached_token
verbose_logger.debug("Authenticating with CyberArk Conjur...")
auth_url = f"{self.conjur_addr}/authn/{self.conjur_account}/{self.conjur_username}/authenticate"
try:
if self.tls_cert_path and self.tls_key_path:
# Certificate-based authentication
client = httpx.Client(cert=(self.tls_cert_path, self.tls_key_path))
resp = client.post(auth_url, content=self.conjur_api_key)
else:
# API key authentication
client = _get_httpx_client()
resp = client.post(auth_url, content=self.conjur_api_key)
resp.raise_for_status()
# The response is a JSON token that needs to be base64-encoded
token_json = resp.text
token_b64 = base64.b64encode(token_json.encode()).decode()
verbose_logger.debug("Successfully authenticated with CyberArk Conjur.")
# Cache the token for the refresh interval
self.cache.set_cache(key="cyberark_auth_token", value=token_b64)
return token_b64
except Exception as e:
raise RuntimeError(f"Could not authenticate to CyberArk Conjur: {e}")
def _get_request_headers(self) -> dict:
"""
Get headers for CyberArk API requests including authentication.
Returns:
dict: Headers with authentication token
"""
token = self._authenticate()
return {"Authorization": f'Token token="{token}"'}
def _ensure_variable_exists(self, secret_name: str) -> None:
"""
Ensure a variable exists in CyberArk Conjur by creating a policy entry if needed.
Args:
secret_name: Name of the variable to ensure exists
"""
# In production, we'd check if the variable exists first
# For now, we'll attempt to create it and ignore if it already exists
policy_url = f"{self.conjur_addr}/policies/{self.conjur_account}/policy/root"
policy_yaml = f"- !variable {secret_name}\n"
try:
client = _get_httpx_client()
resp = client.post(
policy_url,
headers={
**self._get_request_headers(),
"Content-Type": "application/x-yaml",
},
content=policy_yaml,
)
resp.raise_for_status()
verbose_logger.debug("Created policy entry for a variable.")
except httpx.HTTPStatusError as e:
# Variable might already exist, which is fine
if e.response.status_code in [409, 422]:
verbose_logger.debug(
"A variable already exists or policy conflict (expected)"
)
else:
verbose_logger.warning(
f"Could not ensure variable exists: {e.response.status_code} - {e.response.text}"
)
except Exception as e:
verbose_logger.warning(f"Error ensuring variable exists: {e}")
def get_url(self, secret_name: str) -> str:
"""
Build the URL for accessing a secret in CyberArk Conjur.
Args:
secret_name: Name of the secret (will be URL-encoded)
Returns:
str: Full URL for the secret
"""
# URL-encode the secret name to handle slashes and special characters
encoded_name = quote(secret_name, safe="")
return (
f"{self.conjur_addr}/secrets/{self.conjur_account}/variable/{encoded_name}"
)
async def async_read_secret(
self,
secret_name: str,
optional_params: Optional[dict] = None,
timeout: Optional[Union[float, httpx.Timeout]] = None,
) -> Optional[str]:
"""
Reads a secret from CyberArk Conjur using an async HTTPX client.
Args:
secret_name: Name/path of the secret to read
optional_params: Additional parameters (not used for Conjur)
timeout: Request timeout
Returns:
Optional[str]: The secret value if found, None otherwise
"""
# Check cache first
if self.cache.get_cache(secret_name) is not None:
return self.cache.get_cache(secret_name)
async_client = get_async_httpx_client(
llm_provider=httpxSpecialProvider.SecretManager,
)
try:
url = self.get_url(secret_name)
response = await async_client.get(url, headers=self._get_request_headers())
response.raise_for_status()
# CyberArk Conjur returns the raw secret value as text
secret_value = response.text
self.cache.set_cache(secret_name, secret_value)
return secret_value
except httpx.HTTPStatusError as e:
if e.response.status_code == 404:
verbose_logger.debug(
f"Secret {secret_name} not found in CyberArk Conjur"
)
else:
verbose_logger.exception(
f"Error reading secret from CyberArk Conjur: {e}"
)
return None
except Exception as e:
verbose_logger.exception(f"Error reading secret from CyberArk Conjur: {e}")
return None
def sync_read_secret(
self,
secret_name: str,
optional_params: Optional[dict] = None,
timeout: Optional[Union[float, httpx.Timeout]] = None,
) -> Optional[str]:
"""
Reads a secret from CyberArk Conjur using a sync HTTPX client.
Args:
secret_name: Name/path of the secret to read
optional_params: Additional parameters (not used for Conjur)
timeout: Request timeout
Returns:
Optional[str]: The secret value if found, None otherwise
"""
# Check cache first
if self.cache.get_cache(secret_name) is not None:
return self.cache.get_cache(secret_name)
sync_client = _get_httpx_client()
try:
url = self.get_url(secret_name)
response = sync_client.get(url, headers=self._get_request_headers())
response.raise_for_status()
# CyberArk Conjur returns the raw secret value as text
secret_value = response.text
self.cache.set_cache(secret_name, secret_value)
return secret_value
except httpx.HTTPStatusError as e:
if e.response.status_code == 404:
verbose_logger.debug(
f"Secret {secret_name} not found in CyberArk Conjur"
)
else:
verbose_logger.exception(
f"Error reading secret from CyberArk Conjur: {e}"
)
return None
except Exception as e:
verbose_logger.exception(f"Error reading secret from CyberArk Conjur: {e}")
return None
async def async_write_secret(
self,
secret_name: str,
secret_value: str,
description: Optional[str] = None,
optional_params: Optional[dict] = None,
timeout: Optional[Union[float, httpx.Timeout]] = None,
tags: Optional[Union[dict, list]] = None,
) -> Dict[str, Any]:
"""
Writes a secret to CyberArk Conjur using an async HTTPX client.
Args:
secret_name: Name/path of the secret to write
secret_value: Value to store
description: Optional description (not used by Conjur)
optional_params: Additional parameters
timeout: Request timeout
tags: Optional tags (not used by Conjur)
Returns:
dict: Response containing status and details of the operation
"""
async_client = get_async_httpx_client(
llm_provider=httpxSpecialProvider.SecretManager,
params={"timeout": timeout},
)
try:
# Ensure the variable exists in the policy first
self._ensure_variable_exists(secret_name)
# Now set the secret value
url = self.get_url(secret_name)
response = await async_client.post(
url=url, headers=self._get_request_headers(), content=secret_value
)
response.raise_for_status()
# Update cache
self.cache.set_cache(secret_name, secret_value)
return {
"status": "success",
"message": f"Secret {secret_name} written successfully",
}
except Exception as e:
verbose_logger.exception(f"Error writing secret to CyberArk Conjur: {e}")
return {"status": "error", "message": str(e)}
async def async_rotate_secret(
self,
current_secret_name: str,
new_secret_name: str,
new_secret_value: str,
optional_params: Optional[Dict] = None,
timeout: Optional[Union[float, httpx.Timeout]] = None,
) -> Dict:
"""
CyberArk Conjur does not have built-in secret rotation.
Raises:
NotImplementedError: Always raised
"""
raise NotImplementedError("CyberArk Conjur does not support secret rotation")
async def async_delete_secret(
self,
secret_name: str,
recovery_window_in_days: Optional[int] = 7,
optional_params: Optional[dict] = None,
timeout: Optional[Union[float, httpx.Timeout]] = None,
) -> dict:
"""
CyberArk Conjur does not support direct secret deletion via API.
Secrets can only be removed through policy updates.
Args:
secret_name: Name of the secret
recovery_window_in_days: Not used
optional_params: Additional parameters
timeout: Request timeout
Returns:
dict: Response indicating operation not supported
"""
verbose_logger.warning(
"CyberArk Conjur does not support direct secret deletion. "
"Secrets must be removed through policy updates."
)
# Clear from cache
self.cache.delete_cache(secret_name)
return {
"status": "not_supported",
"message": "CyberArk Conjur does not support direct secret deletion. Use policy updates to remove variables.",
}
+8 -89
View File
@@ -1,6 +1,4 @@
import ast
import base64
import binascii
import os
import traceback
from typing import Any, Optional, Union
@@ -14,6 +12,7 @@ from litellm.llms.custom_httpx.http_handler import HTTPHandler
from litellm.secret_managers.get_azure_ad_token_provider import (
get_azure_ad_token_provider,
)
from litellm.secret_managers.secret_manager_handler import get_secret_from_manager
from litellm.types.secret_managers.main import KeyManagementSystem
oidc_cache = DualCache()
@@ -22,13 +21,6 @@ oidc_cache = DualCache()
######### Secret Manager ############################
# checks if user has passed in a secret manager client
# if passed in then checks the secret there
def _is_base64(s):
try:
return base64.b64encode(base64.b64decode(s)).decode() == s
except binascii.Error:
return False
def str_to_bool(value: Optional[str]) -> Optional[bool]:
"""
Converts a string to a boolean if it's a recognized boolean string.
@@ -222,86 +214,13 @@ def get_secret( # noqa: PLR0915
): # allow user to specify which keys to check in hosted key manager
key_manager = "local"
if (
key_manager == KeyManagementSystem.AZURE_KEY_VAULT.value
or type(client).__module__ + "." + type(client).__name__
== "azure.keyvault.secrets._client.SecretClient"
): # support Azure Secret Client - from azure.keyvault.secrets import SecretClient
secret = client.get_secret(secret_name).value
elif (
key_manager == KeyManagementSystem.GOOGLE_KMS.value
or client.__class__.__name__ == "KeyManagementServiceClient"
):
encrypted_secret: Any = os.getenv(secret_name)
if encrypted_secret is None:
raise ValueError("Google KMS requires the encrypted secret to be in the environment!")
b64_flag = _is_base64(encrypted_secret)
if b64_flag is True: # if passed in as encoded b64 string
encrypted_secret = base64.b64decode(encrypted_secret)
ciphertext = encrypted_secret
else:
raise ValueError(
"Google KMS requires the encrypted secret to be encoded in base64"
) # fix for this vulnerability https://huntr.com/bounties/ae623c2f-b64b-4245-9ed4-f13a0a5824ce
response = client.decrypt(
request={
"name": litellm._google_kms_resource_name,
"ciphertext": ciphertext,
}
)
secret = response.plaintext.decode("utf-8") # assumes the original value was encoded with utf-8
elif key_manager == KeyManagementSystem.AWS_KMS.value:
"""
Only check the tokens which start with 'aws_kms/'. This prevents latency impact caused by checking all keys.
"""
encrypted_value = os.getenv(secret_name, None)
if encrypted_value is None:
raise Exception("AWS KMS - Encrypted Value of Key={} is None".format(secret_name))
# Decode the base64 encoded ciphertext
ciphertext_blob = base64.b64decode(encrypted_value)
# Set up the parameters for the decrypt call
params = {"CiphertextBlob": ciphertext_blob}
# Perform the decryption
response = client.decrypt(**params)
# Extract and decode the plaintext
plaintext = response["Plaintext"]
secret = plaintext.decode("utf-8")
if isinstance(secret, str):
secret = secret.strip()
elif key_manager == KeyManagementSystem.AWS_SECRET_MANAGER.value:
from litellm.secret_managers.aws_secret_manager_v2 import (
AWSSecretsManagerV2,
)
if isinstance(client, AWSSecretsManagerV2):
secret = client.sync_read_secret(
secret_name=secret_name,
primary_secret_name=key_management_settings.primary_secret_name,
)
print_verbose(f"get_secret_value_response: {secret}")
elif key_manager == KeyManagementSystem.GOOGLE_SECRET_MANAGER.value:
try:
secret = client.get_secret_from_google_secret_manager(secret_name)
print_verbose(f"secret from google secret manager: {secret}")
if secret is None:
raise ValueError(f"No secret found in Google Secret Manager for {secret_name}")
except Exception as e:
print_verbose(f"An error occurred - {str(e)}")
raise e
elif key_manager == KeyManagementSystem.HASHICORP_VAULT.value:
try:
secret = client.sync_read_secret(secret_name=secret_name)
if secret is None:
raise ValueError(f"No secret found in Hashicorp Secret Manager for {secret_name}")
except Exception as e:
print_verbose(f"An error occurred - {str(e)}")
raise e
elif key_manager == "local":
secret = os.getenv(secret_name)
else: # assume the default is infisicial client
secret = client.get_secret(secret_name).secret_value
# Delegate to the secret manager handler
secret = get_secret_from_manager(
client=client,
key_manager=key_manager,
secret_name=secret_name,
key_management_settings=key_management_settings,
)
except Exception as e: # check if it's in os.environ
verbose_logger.error(
f"Defaulting to os.environ value for key={secret_name}. An exception occurred - {str(e)}.\n\n{traceback.format_exc()}"
@@ -0,0 +1,150 @@
"""
Secret Manager Handler
Handles retrieving secrets from different secret management systems.
"""
import base64
import os
from typing import Any, Optional
import litellm
from litellm._logging import print_verbose
from litellm.types.secret_managers.main import KeyManagementSystem
def _is_base64(s):
"""Check if a string is valid base64."""
import binascii
try:
return base64.b64encode(base64.b64decode(s)).decode() == s
except binascii.Error:
return False
def get_secret_from_manager(
client: Any,
key_manager: str,
secret_name: str,
key_management_settings: Optional[Any] = None,
) -> Optional[str]:
"""
Get a secret from the configured secret manager.
Args:
client: The secret manager client instance
key_manager: The type of key manager (e.g., "azure_key_vault", "google_kms", etc.)
secret_name: The name/path of the secret to retrieve
key_management_settings: Optional settings for the key management system
Returns:
The secret value as a string, or None if not found
Raises:
ValueError: If the secret cannot be retrieved or required parameters are missing
Exception: For other errors during secret retrieval
"""
secret = None
if (
key_manager == KeyManagementSystem.AZURE_KEY_VAULT.value
or type(client).__module__ + "." + type(client).__name__
== "azure.keyvault.secrets._client.SecretClient"
): # support Azure Secret Client - from azure.keyvault.secrets import SecretClient
secret = client.get_secret(secret_name).value
elif (
key_manager == KeyManagementSystem.GOOGLE_KMS.value
or client.__class__.__name__ == "KeyManagementServiceClient"
):
encrypted_secret: Any = os.getenv(secret_name)
if encrypted_secret is None:
raise ValueError("Google KMS requires the encrypted secret to be in the environment!")
b64_flag = _is_base64(encrypted_secret)
if b64_flag is True: # if passed in as encoded b64 string
encrypted_secret = base64.b64decode(encrypted_secret)
ciphertext = encrypted_secret
else:
raise ValueError(
"Google KMS requires the encrypted secret to be encoded in base64"
) # fix for this vulnerability https://huntr.com/bounties/ae623c2f-b64b-4245-9ed4-f13a0a5824ce
response = client.decrypt(
request={
"name": litellm._google_kms_resource_name,
"ciphertext": ciphertext,
}
)
secret = response.plaintext.decode("utf-8") # assumes the original value was encoded with utf-8
elif key_manager == KeyManagementSystem.AWS_KMS.value:
"""
Only check the tokens which start with 'aws_kms/'. This prevents latency impact caused by checking all keys.
"""
encrypted_value = os.getenv(secret_name, None)
if encrypted_value is None:
raise Exception("AWS KMS - Encrypted Value of Key={} is None".format(secret_name))
# Decode the base64 encoded ciphertext
ciphertext_blob = base64.b64decode(encrypted_value)
# Set up the parameters for the decrypt call
params = {"CiphertextBlob": ciphertext_blob}
# Perform the decryption
response = client.decrypt(**params)
# Extract and decode the plaintext
plaintext = response["Plaintext"]
secret = plaintext.decode("utf-8")
if isinstance(secret, str):
secret = secret.strip()
elif key_manager == KeyManagementSystem.AWS_SECRET_MANAGER.value:
from litellm.secret_managers.aws_secret_manager_v2 import (
AWSSecretsManagerV2,
)
if isinstance(client, AWSSecretsManagerV2):
primary_secret_name = None
if key_management_settings is not None:
primary_secret_name = key_management_settings.primary_secret_name
secret = client.sync_read_secret(
secret_name=secret_name,
primary_secret_name=primary_secret_name,
)
print_verbose(f"get_secret_value_response: {secret}")
elif key_manager == KeyManagementSystem.GOOGLE_SECRET_MANAGER.value:
try:
secret = client.get_secret_from_google_secret_manager(secret_name)
print_verbose(f"secret from google secret manager: {secret}")
if secret is None:
raise ValueError(f"No secret found in Google Secret Manager for {secret_name}")
except Exception as e:
print_verbose(f"An error occurred - {str(e)}")
raise e
elif key_manager == KeyManagementSystem.HASHICORP_VAULT.value:
try:
secret = client.sync_read_secret(secret_name=secret_name)
if secret is None:
raise ValueError(f"No secret found in Hashicorp Secret Manager for {secret_name}")
except Exception as e:
print_verbose(f"An error occurred - {str(e)}")
raise e
elif key_manager == KeyManagementSystem.CYBERARK.value:
try:
secret = client.sync_read_secret(secret_name=secret_name)
if secret is None:
raise ValueError(f"No secret found in CyberArk Secret Manager for {secret_name}")
except Exception as e:
print_verbose(f"An error occurred - {str(e)}")
raise e
elif key_manager == "local":
secret = os.getenv(secret_name)
else: # assume the default is infisicial client
secret = client.get_secret(secret_name).secret_value
return secret
+2 -1
View File
@@ -1,5 +1,5 @@
import enum
from typing import List, Literal, Optional, Dict
from typing import Dict, List, Literal, Optional
from litellm.types.llms.base import LiteLLMPydanticObjectBase
@@ -10,6 +10,7 @@ class KeyManagementSystem(enum.Enum):
AWS_SECRET_MANAGER = "aws_secret_manager"
GOOGLE_SECRET_MANAGER = "google_secret_manager"
HASHICORP_VAULT = "hashicorp_vault"
CYBERARK = "cyberark"
LOCAL = "local"
AWS_KMS = "aws_kms"
@@ -0,0 +1,54 @@
"""
Integration test for CyberArk Conjur Secret Manager.
"""
import os
import sys
import pytest
from dotenv import load_dotenv
load_dotenv()
sys.path.insert(0, os.path.abspath("../.."))
from unittest.mock import patch
from litellm._uuid import uuid
# Set up environment variables for testing
os.environ["CYBERARK_API_KEY"] = "2syke5r262b6je2f4et1x3jptmry3frfx83t65e6417zad632e5qq8a"
os.environ["CYBERARK_API_BASE"] = "http://0.0.0.0:8080"
os.environ["CYBERARK_ACCOUNT"] = "default"
os.environ["CYBERARK_USERNAME"] = "admin"
from litellm.secret_managers.cyberark_secret_manager import CyberArkSecretManager
@pytest.mark.asyncio
async def test_cyberark_write_and_read_secret():
"""
Integration test: Write a secret to CyberArk Conjur and read it back to validate.
"""
with patch("litellm.proxy.proxy_server.premium_user", True):
# Create CyberArk secret manager instance
cyberark_manager = CyberArkSecretManager()
# Generate unique secret name and value
secret_name = f"test-secret-{uuid.uuid4()}"
secret_value = f"test-value-{uuid.uuid4()}"
# Write the secret
write_response = await cyberark_manager.async_write_secret(
secret_name=secret_name,
secret_value=secret_value,
)
# Avoid logging write_response to prevent leaking secret names
# Validate write was successful
assert write_response["status"] == "success"
# Read the secret back
read_value = cyberark_manager.sync_read_secret(secret_name=secret_name)
# Don't log secret value in clear text
# Validate the secret exists and has the correct value
assert read_value is not None
assert read_value == secret_value