Docs
CommunityJev Cookbooks
Cut LLM Cost with a Jev-Verified Cascade
An OpenRouter cookbook recipe for a draft-verify-escalate cascade: a cheap model drafts an answer from retrieved context, Jev votes supported, unsupported, or declined, and a frontier model rewrites only when verification fails.
- Format
- Docs
- Reading time
- 2 min
- Published by
- Community
- Last verified
- Oct 2026
- Languages
- TypeScript
- Platforms
- OpenRouterAPI
- Capabilities
- Requires API key
Highlights
- Draft with a cheap model, verify the draft against retrieved context with Jev, and escalate to a frontier model only when the check fails.
- One Jev Choice question returns supported, unsupported, or declined, with a confidence number and a probability per label.
- answerWithCascade returns route: 'send' when a tier passes the threshold, and route: 'handoff' when context is missing or no tier is accepted.
- On the page's 50-question benchmark, the cascade shipped the same zero wrong answers as running the frontier model everywhere, at about 7% of the cost.
- Every tier is paid: the cheap draft plus one Jev check per question, and an escalation adds a frontier call plus a second check.
Watch out
Grounded in the truncated cookbook capture from 2026-10-15; the verifyAnswer implementation is cut off mid-definition, the 50-question result is the page's own worked example, and the model aliases resolve to concrete models at request time.
Questions about this entry
Short answers drawn from the linked project or page.
A draft-verify-escalate pattern for retrieval apps: a cheap model writes an answer from retrieved context, Jev checks whether that context supports it, and a frontier model rewrites the answer only when the check fails.
A verdict over one Choice question with three labels: supported, unsupported, or declined. Supported means the draft's facts, numbers, and policies all appear in the excerpts; declined means it says the excerpts do not cover the question.
The route ('send' or 'handoff'), the model that produced the answer, the answer text, and Jev's verdict for each tier it ran.
On the page's 50-question benchmark, the cascade shipped the same zero wrong answers as running the frontier model on every question, at about 7% of the cost.
Every tier is paid: a question costs the cheap draft plus one Jev check, and an escalated question adds a frontier call plus a second Jev check.
A TypeScript app on Node.js or Bun making OpenRouter chat completions with retrieved context, an OPENROUTER_API_KEY, @openrouter/sdk, and typed process.env values.
The worked example sets DRAFT_MODEL to ~openai/gpt-luna-latest and ESCALATION_MODEL to ~openai/gpt-astra-latest, OpenRouter aliases that resolve to concrete models at request time.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Reddit: Localization QA with a decision model: meaning, terminology, and placeholders as separate checks.
Hacker News: Pick Verify: a single decision at the warehouse pick station catches confusable picks.
Show HN: Verifying robot picks with a decision modelRead the thread on Hacker NewsMore Jev resources like this
An OpenRouter cookbook recipe that replaces a blanket human-approval rule with a per-call Jev check: the proposed tool call, ticket, and policy go to Jev, whose probabilities route to approve, block, or review.
Jev CookbooksDocs#community#catalog-import#cobanov
Community
A host-agnostic routing layer that sits between a coding harness and its models, choosing capacity per round from trajectory evidence like tool calls, failures and context pressure, running as a user-scoped daemon and local OpenAI-compatible proxy.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
vizuh/sabiTypeScript
vizuh/sabiTypeScriptA lightweight, self-hosted LLM proxy that gives Hack Club's teen community free chat, embedding, and image models over OpenAI-compatible endpoints, with Hack Club auth, API keys, moderation, spending limits, and analytics.
Jev Repos & SDKs#community#catalog-import#logicrw
Community
hackclub/aiTypeScript
hackclub/aiTypeScriptClaude Code plugin that asks Jev a few typed questions before every turn and sets reasoning effort, subagent model and effort, strategy, and the skill a prompt needs; you keep Opus for the conversation.
Jev Agents & MCP#community#catalog-import#logicrw
Community
akramovic1/jev-pilotTypeScript
akramovic1/jev-pilotTypeScriptA small, idiomatic Elixir client for Jev: Jev.ask?/2 returns a real boolean, choose/3 picks among supplied options, score/3 rates ordered levels, and analyze/3 sends several typed questions in one provider request.
Jev Repos & SDKs#community#catalog-import#madewithjev
Community
elue-dev/jev_elixirElixir
elue-dev/jev_elixirElixirToolkit, knowledge base and local dashboard for building with Jev via OpenRouter Alpha Decisions or the TypeSafe Direct API: a zero-dependency SQLite request log, a localhost REST server and Python CLI demos.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
rileycarney/jevtoolsHTML
rileycarney/jevtoolsHTMLFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.