Guides — field manual for System One.
Conceptual ladder first, then hands-on recipes. Use Cases stay on their own index.
- 4 foundation pages
- 5 recipe pages
- Paraphrased from public TypeSafe docs
Foundations
What Jev is → vs chat/code → Choice / Score / Noul → first small task.
01
What is Jev?
Jev is TypeSafe’s System One decision model: send state + typed questions, get structured answers your code can branch on—illustrated with a support email, not a chat.
- Paraphrased from public TypeSafe docs
- See Sources below
02
Jev, chat models, and code — who owns what
Code owns control flow and side effects; Jev owns narrow semantic judgments; LLMs write prose when you need text. Prefer workflows + gates over open agent loops.
- Paraphrased from TypeSafe how-to-build guidance
- See Sources below
03
Choice, Score, or Noul — how to pick
Choice selects from a closed set; Score rates ordered levels; Noul returns P(yes). Mix them in one call; each question runs in parallel against the same state.
- Paraphrased from public primitives docs
- See Sources below
04
Pick a small first task
Shrink “AI handles support” to a reviewable label on desensitized mail. Define input, options, success checks, and what you refuse to automate.
- Field-manual template; not an official mandate
- Cite TypeSafe use-case map for patterns
Recipes
Prepare materials → support routing → RAG filtering → n8n → extract and validate.
Step 01
Prepare materials
API key, SDK or HTTP, optional Gateway; decide atomic judgments, state fields, thresholds, and escalate paths before you code.
Step 02
Hands-on: support email routing
Route desensitized mail into shipping/refund/billing/other for human review—never refund or send mail from the model path.
Step 03
RAG passage filtering
Retrieve candidates with your search stack; ask Jev which snippets are relevant enough to keep—code enforces the cut.
Step 04
n8n automation
Call Jev from n8n via community node or HTTP; gate side effects on confidence — community node ≠ TypeSafe product.
Step 05
Extract and validate
Let Jev propose or select candidate field values; deterministic validators own schema, ranges, and rejects.