Examples and conventions
The gRPC examples require the downloadable API definitions and SHRIKEDB_API_DIR setup. The examples assume a supplied ShrikeDB executable is installed on your command path as shrikedb. If your package uses another filename, substitute that executable. Command examples run in a shell; blocks marked Cypher or SPARQL are query text to submit through a query client.
Local connection settings
| Interface | Address in this guide | Purpose |
|---|---|---|
gRPC |
|
Database operations and query execution. |
HTTP |
REST and MCP requests. |
|
Admin |
Health probes and diagnostics. |
The quick start sets these addresses explicitly. For an existing server, get its addresses and logical database ID from your administrator.
Examples using grpcurl -plaintext connect to the local unencrypted gRPC listener. For a TLS endpoint, omit -plaintext and supply the CA certificate required by your deployment. HTTP examples use loopback addresses; use your deployment’s HTTPS endpoint for remote authenticated requests.
Tools and placeholders
Shell examples use Bash, curl, and grpcurl. Pages that process JSON also use jq; byte payload examples use base64. grpcurl uses the API definitions supplied with this guide. Download them and set SHRIKEDB_API_DIR before running gRPC examples.
Replace sample paths, index names, database IDs, and credentials with values from your deployment. A path sent to a server API usually refers to the server’s filesystem; the onboarding chapter explains how to upload a client-local file instead.
Protobuf JSON represents byte fields as base64 strings and commonly returns 64-bit integers as strings. Preserve those strings in client code to avoid precision loss.
Authentication and database selection
The initial walkthrough uses an isolated single-tenant server. Shared deployments require an access token or API key. Connection examples show the headers for both and explain the x-shrikedb-database selector.
Query examples describe the labels, relationships, indexes, and fields they expect. If you have different data, adapt those names before running the query. An empty result is a valid response when no data matches.
Reading feature limits
An unavailable operation returns an error. A disabled feature may require an administrator to enable it before use. Restrictions such as finite path bounds, vector dimensions, or read-only lakehouse mappings affect how you model and query data; the relevant chapter lists those restrictions beside its examples.