Skip to content

api-key.md

Latest commit

 

History

History
102 lines (70 loc) · 3.8 KB

api-key.md

File metadata and controls

102 lines (70 loc) · 3.8 KB

Get an API Key

You need to generate an API key to access NVIDIA NIM Microservices from the NVIDIA RAG Blueprint. You need an API key to access models hosted in the NVIDIA API Catalog, and to download models on-premises. For more information, refer to NGC API Keys.

To generate an API key, use the following procedure.

  1. Go to https://org.ngc.nvidia.com/setup/api-keys.

  2. Click Generate Personal Key.

  3. Enter a Key Name.

  4. For Expiration, choose Never Expire.

  5. For Services Included, select NGC Catalog and Public API Endpoints.

  6. Click Generate Personal Key.

  7. Copy your key and save it somewhere safe and private.

  8. (Important) Export your key as an environment variable by using the following code.

    export NGC_API_KEY="<your-ngc-api-key>"

API Key Expiration

If your API key expires, do one of the following:

  • Create a new key by using the previous procedure, and then delete the expired key.
  • Use the Action menu to Rotate your key.

You must update the new key information in your environment variables and code.

Service-Specific API Keys

The RAG Blueprint supports service-specific API keys to provide fine-grained control over API access for different components. This enables separate billing accounts, different rate limits, security segregation, usage tracking per service, or mixed cloud/on-premises deployments.

How It Works

Service-specific API keys override global NVIDIA_API_KEY or NGC_API_KEY. The system uses the following fallback order to determine which API key to use:

service-specific key → NVIDIA_API_KEY → NGC_API_KEY → None

For example, if you set APP_LLM_APIKEY, the LLM service will use that key instead of the global NVIDIA_API_KEY.

Supported Service Keys

Service Key Purpose Used By
APP_LLM_APIKEY Main language model for RAG responses RAG Server (generate endpoint)
APP_EMBEDDINGS_APIKEY Text embedding generation Ingestor Server, RAG Server (search)
APP_RANKING_APIKEY Document reranking RAG Server (reranker)
APP_QUERYREWRITER_APIKEY Query rewriting and optimization RAG Server (query decomposition)
APP_FILTEREXPRESSIONGENERATOR_APIKEY Metadata filter generation from natural language RAG Server (search with filters)
APP_VLM_APIKEY Vision-Language Model for image understanding RAG Server (multimodal queries)
SUMMARY_LLM_APIKEY Document summarization Ingestor Server (summary generation)
REFLECTION_LLM_APIKEY Self-reflection and answer validation RAG Server (reflection mode)

Library Mode (Python Package)

Configure in config.yaml:

llm:
  api_key: "your-llm-api-key"  # Optional: overrides NVIDIA_API_KEY
embeddings:
  api_key: "your-embeddings-api-key"  # Optional: overrides NVIDIA_API_KEY

Docker Compose

export APP_LLM_APIKEY="your-llm-api-key"
export APP_EMBEDDINGS_APIKEY="your-embeddings-api-key"
# Additional service-specific keys can be set as needed

Helm

Use --set flags to pass API keys securely via command line:

helm upgrade --install rag -n rag <chart-path-or-url> \
  --set imagePullSecret.password=$NGC_API_KEY \
  --set ngcApiSecret.password=$NGC_API_KEY \
  --set apiKeysSecret.llmApiKey=$APP_LLM_APIKEY \
  --set apiKeysSecret.embeddingsApiKey=$APP_EMBEDDINGS_APIKEY

Additional service-specific keys can be configured as needed (e.g., rankingApiKey, vlmApiKey, summaryLlmApiKey).

Related Topics