Agents: read the raw Markdown of this page, or start at llms.txt.
~/wiki/ideas
Apps with a built-in Jev option
TL;DR Products that ship Jev as a setting or node, so you can use it without writing a client: what Jev decides there, how to turn it on, and each vendor's status. Community tier; vendor claims are labelled. Libraries and frameworks you code against stay on Framework integrations: Pydantic AI, LangChain, Spring AI, Vercel AI SDK, eve, LiteLLM, BAML, TanStack AI, DSPy, Langfuse and other framework-shipped Jev support.
Apps with a built-in Jev option
| App | What Jev does | Turn it on | Status | Note |
|---|---|---|---|---|
| Langfuse "Jev as a judge" (docs) | by Langfuse, marked experimental ("details of the UI and score format may still" change): a decision-model evaluator; you map observation fields into state, add 1 to 50 questions (Langfuse's cap), all answered in one call, each written back as one score (Choice → CATEGORICAL option, Score → NUMERIC expected level, Yes/no → NUMERIC P(true)); probabilities, confidence and resolved model under metadata.typesafe; templates for topic, out-of-scope, frustration and a seven-signal conversation check; runs on live rules, batch evaluation and experiments; execution traces in environment langfuse-llm-as-a-judge |
Settings → LLM Connections, adapter typesafe; upstream TypeSafe, Vercel AI Gateway or OpenRouter, billed by that upstream; models jev-1.13.0, jev-latest |
Langfuse docs · captured 2026-09-28 | Limits (255 options, 2 to 10 levels, ~32k for state plus longest question, input-only pricing) verified against Models, aliases, pricing, rate limits, context and Score questions. Stale on access: its FAQ says access "runs through a waitlist" and links the old /settings/keys path; TypeSafe reopened signup on 2026-09-27 and the docs link /keys (console.typesafe.ai (console + playground)). State goes to TypeSafe via the chosen upstream: check its data terms. Build and results: Builds: data, search and business (Langfuse trace judge) |
| MLflow (blog part 2, Databricks, 2026-09-25) | by MLflow: MLflow 3.17 (unreleased at 2026-09-25) previews Jev in built-in scorers and make_judge via typesafe:/jev-latest; the probability lands in feedback metadata typesafe.probability |
MLflow 3.17 | preview | Judge results: Head-to-head benchmarks: retrieval, reranking, screening and judging |
| n8n "TypeSafe AI" node (announcement, Oli Morris, n8n team, 2026-09-30; vendor) | by n8n, pitched as a "smart If node": operation "Route item by System One question" with State (mapped fields, or the whole item as JSON), Instructions and one Route per outcome, each a named output like If or Switch; Jev's confidence can send unsure items to a person (n8n's linked video gates at 95%) | n8n Cloud (Starter, Pro, new trial): built in, runs on n8n Gateway credits, no TypeSafe account or key; self-hosted or Community Edition: install the TypeSafe AI node with your own key | free on Cloud until 2026-10-10 23:59 UTC, per n8n; then n8n's standard Gateway credit rates | Node name and Route operation match TypeSafe's own node (@typesafe-ai/n8n-nodes-typesafe-ai, Evaluate and Route; first on npm 2026-09-29, our check 2026-10-01; typesafe-ai GitHub organisation and repos); the post does not say Cloud runs that same package (inferred, not confirmed). On Cloud your state goes through n8n's Gateway to TypeSafe: check n8n's data terms. $42 per Btok input, output free verified (Models, aliases, pricing, rate limits, context); "238x lower" than Fable 5.1 is TypeSafe's claim. n8n's advice matches the wiki's: a clean rule (amount > 10000) stays an If node; text output needs an LLM. Community nodes: Tools and integrations hub: MCP servers, plugins and connectors. An n8n email-routing build (Gmail to Jev to Slack): Builds: business ops and markets. P06, P22 |
Inbox Zero (elie222; open-source email assistant, 12314★ on 2026-09-23; decision-model code at de44248) |
optional "specialized decision model": Choice over the user's rules, cold-email yes/no, sender categories, thread status, unsubscribe pages, reply-memory selection; raw fetch to /v1/systemone, 30 s timeout |
operator sets DEFAULT_DECISION_MODEL=typesafe:<model> and TYPESAFE_API_KEY; DEFAULT_DECISION_MODEL_ENABLED (default false) decides whether users who have not chosen get it; users with their own AI key never get it by default; per-user Settings toggle |
merged to main 2026-09-19 and 2026-09-21, no tagged release (last web tag 2026-05-18); whether the hosted app runs it is not stated | our code read 2026-09-24, not captured (unverified beyond that). Any error, or a rule Choice under 0.3 (MIN_CHOICE_CONFIDENCE, compared with the answer's confidence, or with the chosen rule's probability when it pools content rules; our code read 2026-09-25), falls back to the normal LLM path. Its pricing table lists jev-latest at $42 per billion input tokens, output free (= $0.042/Mtok; our code read 2026-09-24; equals the list price, Models, aliases, pricing, rate limits, context). P22 |
MotherDuck prompt_jev() (walkthrough, @mehd_io, 2026-09-29; docs; vendor) |
by MotherDuck: a SQL function returning a struct with .choice, .confidence, .probabilities; Choice and Score. Walkthrough: gpt-5-mini proposed 7 labels from 40 sampled CFPB complaints (7.6 s), then Jev classified 100,000 rows in 82.2 s; gpt-5-nano via prompt() took 6 min 9 s for 10,000 rows (Jev 12.5 s). Mean confidence 0.83; 65.8% of rows ≥ 0.8, 10.9% below 0.5. On 300 rows: GPT-5 vs Opus 5.5 agree 72.7%, Jev vs GPT-5 77.3%, Jev vs Opus 74.0%; Jev matches on 90.4% of the 218 rows where both agree, 96.9% of the 161 of those at ≥ 0.8. MotherDuck's launch benchmark (100k short news articles): Jev 40 s, $0.50 vs LLMs 18–32 min, $1.58–$37.58 |
prompt_jev(text, question, ...) in SQL |
paid plans | text goes to TypeSafe; check data handling. Agreement is with other models, not labels. unverified |
DataGOL (document sorting): Jev available as of 2026-10-05; vendor claims only: Head-to-head: Jev against other models and methods.
Related
- Framework integrations: Pydantic AI, LangChain, Spring AI, Vercel AI SDK, eve, LiteLLM, BAML, TanStack AI, DSPy, Langfuse and other framework-shipped Jev support — framework-shipped Jev support you code against
- Platforms and gateways: Zapier, Cloudflare, Netlify, Vercel AI Gateway, OpenRouter, Pydantic AI Gateway, Opper, Fly.io, OpenCode Zen and other hosted routes to Jev — hosted routes to Jev
- Head-to-head benchmarks: retrieval, reranking, screening and judging — judge comparisons
- HTTP API: POST /v1/systemone and GET /v1/models — the wire contract these apps call
Sources
Files in frontmatter sources:, captured by 2026-10-05; original URLs inline.