Jev directory
CommunityJev Agents & MCP
Dsh Skill Auto Load TypeSafe
A DeepSeek Harness host plugin that calls TypeSafe AI through the SDK to score which visible skills apply to a user turn, then injects the selected skill instructions into the same model request.
- Reading time
- 2 min
- Published by
- Community
- Added
- Oct 2026
Highlights
- Scores visible skills per user turn via one systemOne() request.
- Default threshold 0.75 on the Noul yes probability, max 3 skills.
- Verified against Harness 0.1.5-rc.3, Cordis 4.0.2, TypeSafe SDK 0.6.0.
- Requires Node ^22.19.0 or >=24.0.0 and a TYPESAFE_API_KEY.
- Temporary network or timeout errors default to continue.
Questions about this entry
Short answers drawn from the linked project or page.
Run dsh plugin --profile web add dsh-skill-auto-load-typesafe, which pulls the package from npm through pnpm. For local development you can instead add an absolute directory path after building with npm ci and npm run build.
It uses a separate TYPESAFE_API_KEY credential, resolved by the Harness credential service from the launch environment, the credentials file, or .env sources. It does not route through the Harness ctx.llm or change the main model's provider.
By default at most 3 skills are added, chosen by descending yes probability with ties broken by name. Only skills at or above the 0.75 threshold are selected; if none qualify, nothing is injected.
With the default onSelectionError of continue, temporary network, parsing, or timeout failures degrade gracefully and the original task proceeds. Missing keys, auth failures, and invalid config fail explicitly.
Only agents in the standard preset are covered out of the box. The presetIds config option widens or narrows that, and an empty list means all presets including headless agents.
The total pre-step budget is 5000 ms by default, which is also the SDK request timeout. Requests are capped at 65536 bytes and injected skill bodies at 32768 bytes; over-limit requests error rather than truncate.
It writes started and completed records to a skill_auto_load_typesafe storage domain, including session ID, user message ID, endpoint, the JSON request, byte counts, candidate count, and selection outcomes.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
X: A group chat bot with decision room routing
More Jev resources like this
Zero-dependency supervisor extension for the Pi coding agent. Blocks destructive commands and hardcoded secrets with a 0 ms regex gate, warns on unverified edits, and routes diffs of 10 lines or more through Jev, OpenRouter, or Cloudflare Clef.
Jev Agents & MCP#coding-agent#guardrails#routing
Community
A Model Context Protocol server that exposes TypeSafe AI's Jev decision model as typed tools over stdio, runnable with npx or bunx. Ships jev_decide, jev_choice, jev_score, jev_noul, jev_box and jev_list_models for MCP clients such as Claude Code, Cursor and VS Code.
Jev Agents & MCP#mcp#typescript#agents
Communitydevbux/typesafe-mcp
A Juspay TypeScript SDK that treats Jev as a third inference type beside generate and stream: providers declare which kinds they serve, and tryDecide returns typed boolean, choice, or score answers, null when no decision provider is set.
Jev Repos & SDKs#community#typescript#sdk
Community
juspay/neurolinkTypeScript
juspay/neurolinkTypeScriptGitHub issue triage agent built with Eve and Jev. It takes one typed decision per issue through a pipeline of classify, validate, duplicate and apply steps, with dry-run defaults, a label allow-list, and raw Jev answers stored for threshold tuning.
Jev Agents & MCP#agents#github#madewithjev
Communitybenjamincanac/tia
A model router for the Codex CLI: on every user turn Jev judges how hard the task is and a YAML policy picks one of three lanes — flash, craft or forge — so premium quota goes to the prompts that need it.
Jev Agents & MCP#community#typescript#codex
Community
romanmeclazcke/codex-siftTypeScript
romanmeclazcke/codex-siftTypeScriptA DeepSeek Harness plugin that exposes Jev's noul, choice and score primitives as agent tools, plus configurable guardrails, retries, response caching, a circuit breaker and Web UI result cards.
Jev Agents & MCP#community#typescript#deepseek
Community
jackie-cqz/dsh-jev-pluginTypeScript
jackie-cqz/dsh-jev-pluginTypeScriptFrom 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.