B1 – Model Statistics
← Back to index | ← B – Chat | → B2 – Document Generation
B1 · Model exploration
Natural language queries against the open EA model work reliably.
Test query:
What is the name of current repository?
Kernaro returned a structured model overview including all root models with package hierarchy, diagram inventory, and a cost estimation for document generation.
Important: For large repositories, always specify a GUID or package name to limit context scope. Without scoping, Kernaro sends large portions of the model as context with each request – see B2.
B2 · Token cost – context injection
Test: Generate a Word document with diagram list for the full repository (~800 MB, MariaDB [1], 4,734 diagrams).
Result: ~6,000,000 tokens consumed. Cost: ~$2.42 USD for a partial run.
Root cause: Kernaro uses context injection – model data is serialised and sent as part of each LLM prompt. The entire model (or large portions) travels to the Anthropic API [2] with each request.
| Scope | Approx. tokens | Approx. cost |
|---|---|---|
| Full model, no scope limit | ~6,000,000 | ~$2.00–$3.50 |
| GUID-scoped package | ~3,000 | ~$0.34 |
| Single element agent | ~1,144 | ~$0.05 |
Recommendation: Always include a GUID or package name in chat queries.
GUID ={8208DC8D-B283-10B3-86D0-E8D891DD8861}.
List all elements in this package that have empty Notes field.
Show element name and type.
Product links
[1] MariaDB – https://mariadb.org
[2] Anthropic Console – https://console.anthropic.com
← Back to index | ← B – Chat | → B2 – Document Generation
Images
[IMG-01]
[IMG-02]
[IMG-03]
[IMG-04]
[IMG-05]
[IMG-06]
[IMG-07]
[IMG-08]
[IMG-09]
[IMG-10]
