User repository on GitHub
Typed Jev relation governance and safe archive plans for Hermes Curator
- Forks
- 3
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Agents & MCP
Hermes Jev Curator
A Hermes Agent plugin that puts typed Jev judgments in front of the background skill curator: per-pair relation choices and coverage probabilities feed star-shaped merge plans, while core keeps ownership of every mutation.
- Stars
- 3
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- anpicasso
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Python
- Capabilities
- Open sourceRequires API key
Highlights
- Each pair gets one Jev request answering an 8-way relation choice — duplicate, subset, complementary, conflict, unrelated — plus coverage probabilities.
- Candidates are deterministic: a read-only inventory plus lexical top-k neighbor pairs, no model call, identical outputs for unchanged inputs.
- Merge plans are star-shaped: A~B plus B~C never merges C into A; each absorbed member needs its own direct containment judgment bound to digests.
- The plugin writes no skill text; apply only archives sources a validated plan says the canonical preserves, via core's ledgered skill_manage.
- Default mode is read-only observe, and a profile-local 60-second debounce refreshes evidence; automatic runs never apply or archive.
Quickstart
bash
hermes curator run --consolidateWatch out
Copy reviewed against the truncated repo README on 2026-10-15; the capture ends inside the curator configuration section. This is a 0.1.0 experimental release and its bundled synthetic corpus is a contract self-check, not a model-quality benchmark, as the README states.
Questions about this entry
Short answers drawn from the linked project or page.
A plugin for Hermes Agent's background skill curator that adds typed, per-pair skill-relationship judgments served by Jev. It finds candidates deterministically, asks Jev about each pair, and builds safe archive plans without writing skill text itself.
Each candidate pair uses one bounded request when both packages fit, or deterministic per-direction chunk requests otherwise. Jev answers a versioned contract: an 8-way relation Choice plus coverage, preservation and conflict probability questions.
No. It complements stock Hermes Curator: core still owns the managed-skill population, consolidation agent, snapshots, skill_manage, archive ledger and rollback. The plugin adds typed judgments and hash-bound guard plans on top.
The plugin writes no skill text and umbrella prose stays core's job. Its apply path only archives sources that a validated plan says the canonical already preserves, through core's ledgered skill_manage, after a pre-apply snapshot.
Every successful skill lifecycle mutation resets a profile-local 60-second debounce; after one quiet minute the plugin runs read-only when mode is not off and content egress is allowed. Repeated mutations collapse into one run, and automatic runs never apply or archive.
No. The README says the bundled synthetic corpus is a contract self-check, not a model-quality benchmark. The plugin has been exercised against the live TypeSafe endpoint with whole, chunked and locally unavailable pairs.
It is a 0.1.0 experimental release, and the 60-second debounce is described as a pragmatic workaround over Hermes' current per-event hook, with a cleaner upstream API intended. Default mode is read-only observe; apply requires explicit opt-in.
More Jev resources like this
A native Hermes Agent tool plugin - not a chat-model provider - where Jev supplies bounded semantic judgments and an explicit model recommendation while the main agent keeps planning, generation, execution and approvals. Backends: TypeSafe, Cloudflare and OpenRouter.
Jev Agents & MCP#community#catalog-import#logicrw
Community
ourines/hermes-jevPython
ourines/hermes-jevPythonAn intent compiler for AI agents: it probes what files and tickets already answer, asks only what they can't, and emits a typed IntentSpec YAML saved to .intent/<intent>.intent.yaml before work is routed. The README positions it as the input layer for routers and typed-decision models like Jev and Laya.
Jev Agents & MCP#community#catalog-import#madewithjev
Community
angel291592/intent-routerPython
angel291592/intent-routerPythonA Claude Code plugin that keeps agents honest: Claude lays work out as phases with a definition of done and a check, and whenever it tries to stop, Jevflow runs the checks and asks Jev whether the work is really done. Several agents can share one plan.
Jev Agents & MCP#community#catalog-import#logicrw
Community
parth1811/jevflowPython
parth1811/jevflowPythonA Hermes Agent skill that turns an intention into a written finish line: machine-checkable requirements run as shell commands, fuzzy ones go to Jev for calibrated probabilities, and exit 1 names the weakest unmet requirement.
Jev Agents & MCP#community#catalog-import#logicrw
Official MCP server for JevGuard and TypeSafe AI: zero-dependency guardrails over JSON-RPC 2.0 on stdio, gating shell commands and code patches, caching decisions in local SQLite and pruning state.
Jev Agents & MCP#community#catalog-import#cobanov
Community
seb4ez/jevguard-mcpPython
seb4ez/jevguard-mcpPythonA Claude Code plugin and zero-dependency Python CLI that judges items in bulk with calibrated Jev answers, resolves the confident ends in code, and sends only the uncertain band to Claude or a human.
Jev Agents & MCP#community#catalog-import#logicrw
From 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.