Kavvo Documentation
The documentation comes in three categories—don’t mix them up:
User documentation — tutorials organized around workflows. Install the software, build a model, run the analysis, generate a calculation report: just follow along.
Materials handbook — how to fill in material parameters. Each material is covered clearly: where it applies, what each parameter means, where the values come from, and what happens if you get them wrong.
AI knowledge base (guide) — domain knowledge cited by Kavvo’s AI in conversation, one entry per class of “pitfall the model itself will hit.” They are written for humans too. Wherever the AI cites an entry, you can read the same entry to see the basis behind what it says—the 📖 badge in the desktop app conversation links straight to the corresponding page.
Materials Handbook
Section titled “Materials Handbook”Material documentation in an analysis tool shouldn’t just be a “parameter list.” The same f′c entered as 30 versus 20.1 differs by a factor of two in capacity, and the software won’t stop you. That’s why these four pages are organized as “how to choose → how to fill in → how to check”:
- Materials System Overview — uniaxial vs. multiaxial, where parameters come from, the iron rules of the unit system
- Steel & Reinforcement — Steel01 or Steel02, what the hardening ratio b is, why the Bauschinger parameters shouldn’t be touched casually
- Concrete & Confined Concrete — whether the tension branch is needed, how to split cover vs. core zones, how to use the Mander model
- Material Selection & Full Catalog — lookup tables by member scenario, with an index of 68 uniaxial + 23 multiaxial materials
Every parameter value is annotated with its source and applicable conditions. Code values, test values, and software defaults must never be mixed—the constitutive model needs the actual strength at peak stress, not the design value. This is the most common source of errors in the materials section.
Why the AI Knowledge Base Is Public
Section titled “Why the AI Knowledge Base Is Public”Because “verifiable results” is a fundamental promise of this product—the computed results must check out to the right order of magnitude, and what the AI says must check out against its sources. Hiding the AI’s knowledge sources amounts to asking users to trust a conclusion with no citation.
These entries come from the exact same files the desktop application reads: when the software updates the guides, the documentation site updates in sync, so you never get “one story in the software, another on the web.”