jevbooks

← All projects

harness-router

Protocol-Lattice/harness-router · Homepage

Decision layer for tool selection inside agent harness loops.

Agent Decisions1.00Runner-up: Classification & RoutingTraffic light
Automated README summary

Automatically summarized from the README on 30 Sept 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.

What it is

Harness Router is a framework-agnostic tool router for coding agents. It takes a goal, observation, and available tools, then returns a next-tool choice, confidence, or a planner fallback, keeping obvious calls on a fast path.

How it uses Jev

Jev makes discrete next-tool decisions from compact state and tool descriptions via OpenRouter. The router accepts choices above a confidence threshold, caches reusable high-confidence choices, and otherwise falls back to the host planner. MCTS can optionally use a Jev prior.

Primitives:choice

Technique worth stealing

Route only genuinely ambiguous tool choices through Jev; keep obvious calls on a fast path and fall back to the planner on low confidence.

Try it

uv tool install --force --with 'mcp>=2' harness-router, then run har route --help.

View on GitHub

Start using Jev: API key and your first request →

Model judgments from README excerpts

Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.

Models & evaluation dates

Inclusion gate
jev-1.13.0
Project signals
jev-1.13.0
Category
jev-1.13.0
Pattern tags
jev-1.13.0

Evidence

These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.

  • Jev is core to the projectnot found in the excerpt0.19
  • Shows a System One patternFound in material0.85
  • Describes handling uncertaintyunclear0.47
  • Reports measured numbersnot found in the excerpt0.05
  • Is a runnable toolFound in material0.90
  • Top-10 learning shortlist signalnot found in the excerpt0.26
  • Is a replica of the modelnot found in the excerpt0.04
  • Solves a common problem (0–2)score on a 0–2 scale1.57 / 2
  • About JevFound in material0.94
Signals
Jev is core to the project0.19Shows a System One pattern0.85Describes handling uncertainty0.47Reports measured numbers0.05Is a runnable tool0.90Top-10 learning shortlist signal0.26Is a replica of the model0.0400.51Solves a common problem (0–2)1.57 / 2012

no-judgment band (0.40–0.60)

Design patterns
Traffic light0.96Funnel0.65Fail open, fail closed0.39Viewfinder0.36Thermostat0.34Grey zone0.29Scorecard0.2000.51

no-judgment band (0.40–0.60)

The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 30 Sept 2026.