Choosing the Right YJTOON Endpoint for Your Agent

YJTOON serves the same data three ways. Picking the right path for a given agent run is a small decision that compounds over millions of runs.

The three options

• REST API at /api/dataset/<slug>. Content-negotiated, rate-limited, served from origin.
• Static file at /static-data/dataset/<slug>.<fmt>. CDN-fronted, no rate limit, byte-stable until the next publish.
• Inline snippet in your own system prompt. Zero round trip, no freshness.

Pick the REST API when

• The agent runs once and needs the freshest data.
• You are writing the agent in a language where HTTP is a one-liner.
• You want format negotiation without committing to one format in your prompt.

Pick the static file when

• The agent runs in CI or on a schedule and caches well.
• You need a deterministic byte stream for hashing or diffing.
• You want to avoid the rate limiter entirely.

Pick the inline snippet when

• The data is small, stable, and tightly coupled to the prompt.
• You want zero network dependency in the agent run.
• You are willing to re-emit the prompt on every dataset update.

The hybrid pattern

Inline the smallest, most stable part of the reference. Fetch the larger or fresher part from the static file at agent boot. Combine inside the prompt, attribute both. This is the pattern most of our internal agents end up using.