1. Official documentation map
September 18, 2026 · View on GitHub
The official docs at https://docs.typesafe.ai are authoritative and change without notice
(index: https://docs.typesafe.ai/llms.txt; append .md to any page path for Markdown). This
repository no longer restates them. The table says where each topic's operative version lives
here: the condensed rules a coding agent needs while designing, served by the skill
(skills/tenbin/reference/) and the MCP server (tenbin://guide/*, from tenbin/resources/).
| Topic | Official pages | In this repository |
|---|---|---|
| What System One / Jev is, RLCD, state basics | introduction, concepts/system-one, concepts/state, introduction/machine-learning-primer | README.md "TypeSafe in one minute"; skills/tenbin/reference/state.md |
| Choice / Score / Noul, request shape, limits | primitives, primitives/{choice,score,noul,advanced} | reference/primitives.md (= tenbin://guide/primitives), reference/question-writing.md (structured instructions, levels, criteria) |
| Confidence and thresholds | confidence, patterns/confidence-routing | reference/confidence.md, reference/thresholds.md |
| Patterns (fan-out, composite, intent routing, two-stage, verification, guardrail) | patterns/* , demos/smart-home | reference/patterns.md; diagrams in 02-workflows.md |
| How to build a workflow, use-case map | concepts/how-to-build-with-system-one, concepts/use-case-map | skills/tenbin/SKILL.md steps 1–9; 02-workflows.md |
| Jev 1.13 failure modes (jaggedness) | model-jaggedness/jev-1.13 | reference/jaggedness.md (carries the official "last reviewed" date) |
| HTTP API, models, pricing, rate limits, language support, data handling | api, models, legal | reference/sdk.md "Limits and price"; tenbin/src/config.ts (limits as constants); 04-runbook.md §1 (data handling) |
| Python / JavaScript SDK | sdk, sdk/python/, sdk/javascript/ | reference/sdk.md; skills/tenbin/templates/*.py, questions.ts |
| Cookbooks (18) | cookbooks/* | reference/cookbooks.md (= tenbin://guide/cookbooks): one line + URL each; techniques folded into patterns.md, thresholds.md, question-writing.md |
| Official agent skill | agent-skill, https://github.com/typesafe-ai/skills | skills/tenbin/ is a superset with execution (see 03-mcp-server-and-skill-design.md) |
Official site structure
introduction/ Introduction, Quick start, AI primer
concepts/ System One, State, How to build, Use case map
primitives/ Choice, Score, Noul, Advanced: structure
confidence
patterns/ fan-out, confidence-routing, composite-scoring, intent-routing
demos/ smart-home
cookbooks/ 18 cookbooks (consistency_noul, consistency_choice, parallel_questions, rerank,
semantic_find, autoformat, function_calling, skill_suggestion, entity_alignment,
classifying_rag_passages, citation_check, llm_guardrails, sde_cascade,
date_extraction, pre_parsed_value_extraction, hierarchical_classification,
autoresearch_feature_discovery, classification_using_confidence)
sdk/ python (usage, changelog, api/**), javascript (changelog, api/**)
models, api, agent-skill, legal, model-jaggedness/jev-1.13
Keeping the repository in step
When an official page changes, update tenbin/resources/*.md (then scripts/sync_resources.sh)
or the skill-only files under skills/tenbin/reference/, and the lint rules if a limit moved.
The maintenance checklist is in 04-runbook.md §5.