spice-caching
Spice Caching
Configure in-memory caching for SQL query results, search results, and embeddings in the Spice runtime.
Overview
Spice caches results from SQL queries (/v1/sql), search (/v1/search), and embeddings requests. All three caches are enabled by default with a 1-second TTL and 128 MiB max size. Caching applies to HTTP and Arrow Flight APIs.
Configuration
Caching is configured under runtime.caching in spicepod.yaml:
version: v1
kind: Spicepod
name: app
More from spiceai/skills
spice-data-connector
Configure individual data source connectors in Spice — PostgreSQL, MySQL, S3, Databricks, Snowflake, DuckDB, GitHub, Kafka, and 25+ more. Use this skill whenever the user wants to add a dataset, connect to a specific database or data source, load data from S3 or files, configure connector-specific parameters, understand file formats (Parquet, CSV, PDF, DOCX), or set up hive partitioning. This skill is the reference for the `from:` and `params:` fields in dataset configuration. For cross-source federation, views, and catalogs, see spice-connect-data.
22spicepod-config
Create and configure Spicepod manifests (spicepod.yaml) — the central configuration file for Spice applications. Use this skill whenever the user wants to create a new spicepod.yaml from scratch, understand the overall spicepod structure and available sections, configure runtime settings (ports, caching, telemetry/observability), set up a complete Spice application combining datasets + models + search, or understand deployment models and use cases. This is the "glue" skill that shows how all Spice components fit together in one manifest. For details on specific sections (datasets, models, search, etc.), see the dedicated skills.
16spice-models
Configure AI/LLM model providers and connections in Spice — OpenAI, Anthropic, Azure, Google, xAI, Bedrock, Perplexity, Databricks, HuggingFace, and local GGUF models. Use this skill whenever the user wants to add a model, configure a specific LLM provider, set up an OpenAI-compatible endpoint (e.g. Groq, Ollama), serve a local model, configure system prompts, set parameter overrides (temperature, response format), or understand which providers are available. This skill is the model connector reference. For AI features like tools, memory, workers, and NSQL, see spice-ai.
16spice-accelerators
Choose and configure the right acceleration engine — Arrow, DuckDB, SQLite, Cayenne, PostgreSQL, or Turso. Use this skill whenever the user needs to pick an accelerator engine, compare engines (e.g. "should I use DuckDB or Cayenne?"), configure engine-specific parameters (duckdb_file, sqlite_file), tune memory vs file mode, or understand engine capabilities and limitations. This skill is the engine selection and tuning guide. For the broader acceleration feature (refresh modes, retention, snapshots, indexes), see spice-acceleration.
15spice-secrets
Configure secret stores in Spice — environment variables, Kubernetes, AWS Secrets Manager, and OS keyring. Use this skill whenever the user needs to manage credentials, API keys, passwords, or tokens in Spice, reference secrets in spicepod.yaml params with ${ store:KEY } syntax, set up .env files, configure secret store precedence, or understand how the `secrets:` section works. Also use when the user asks how to pass database passwords or API keys securely to Spice datasets or models.
12spice-acceleration
Accelerate data locally for sub-second query performance — the feature and its configuration. Use this skill whenever the user asks about data acceleration concepts, enabling acceleration on a dataset, choosing refresh modes (full, append, changes, caching), configuring retention policies, setting up snapshots for cold-start, adding indexes and constraints, or understanding the difference between federated and accelerated queries. This skill covers the "what and why" of acceleration. For choosing which acceleration engine to use (Arrow vs DuckDB vs SQLite vs Cayenne), see spice-accelerators.
10