Wizard: Which Memory Evaluation Approach Fits Your Stack?
There isn't one right way to evaluate or seed an agent-memory system — the right approach depends on how close your use case is to general financial-advisor decision memory, how much engineering time you have to invest, and whether you need data today or can build toward it. This wizard routes to one of four honest paths: run the free open benchmark, build a custom harness, buy a ready-made decision corpus, or request a custom-spec batch — each with its real tradeoffs stated plainly, not oversold.
What you walk away with
~2 min · 3 questions · 4 possible outcomes- A specific recommended path: run the open bench, build custom, buy the corpus, or request a custom spec.
- A one-paragraph rationale for why that path fits what you described.
- Direct links to the relevant guide, template, or dataset for that path.
- A 'talk to us' fallback on every outcome, for the cases this wizard doesn't cover exactly.
What are you trying to do right now?
Pick the closest fit — the next question (if any) narrows further.
FAQ
What if our use case spans more than one of these paths?
Pick the one that unblocks the most urgent need first. It's common to run the open bench for an initial read on vendor landscape, then buy a ready-made corpus to seed a specific agent, then build custom evaluation later once the domain-specific gaps are clear from real use.
Is running the open bench really free?
Yes — the schema, harness, and public dev set are free and openly licensed. The only cost is your own compute and, if testing a vendor's system, whatever that vendor charges — nothing is gated behind this wizard's other three paths.
Can we revisit our path?
Yes — the wizard saves your path to localStorage, and the Back button steps back without losing state, so re-answering a question doesn't mean starting over.