Registry / ai-ml / harbor-rewardkit

harbor-rewardkit

library0.1.4pypypi✓ verified 86d ago

Lightweight grading toolkit for environment-based tasks, part of the Harbor framework. Current version 0.1.4, updated Apr 2025. Pre-release with monthly releases.

ai-ml testing

pip install harbor-rewardkit

github ↗homepage ↗

INSTALL

IMPORT

SIG · HARBOR-REWARDKIT

harbor-rewardkit

ai-mlpythonv0.1.4

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

AgentCLI

✓ from rewardkit import AgentCLI

✗ from harbor_rewardkit import Aggregator

LLMJudge

✓ from rewardkit import LLMJudge

Reward

✓ from rewardkit import Reward

Creates an Aggregator with default score and reward, then grades a trajectory.

from harbor_rewardkit import Aggregator from harbor_rewardkit.scores import EpsilonGreedyScore from harbor_rewardkit.rewards import BinaryReward agg = Aggregator(score=EpsilonGreedyScore(), reward=BinaryReward()) result = agg.grade(trajectory=[("state1", "action1", 1.0), ("state2", "action2", 0.0)]) print(result)

Debug

Known issues

breakingPython 3.12+ required. Install on 3.11 or earlier will fail with syntax errors due to new language features.

fix

Upgrade Python to 3.12 or later.

affects: all

gotchaImport paths use underscores (harbor_rewardkit) not hyphens (harbor-rewardkit). This is a common mistake when copy-pasting from the project name.

fix

Use from harbor_rewardkit import ...

affects: all

deprecatedThe `grade` method signature may change. Currently it accepts `trajectory` as a list of tuples; future versions may require a structured dict.

fix

Check documentation before upgrade; wrap trajectory creation in a helper.

affects: <=0.1.4

gotchaDefault score and reward classes may change defaults in minor versions (e.g. EpsilonGreedyScore epsilon value). Always specify parameters explicitly.

fix

Instantiate with explicit arguments: EpsilonGreedyScore(epsilon=0.1)

affects: >=0.1.0

Upgrade

Version history

0.1.4latest on PyPI · released May 8, 2026

Audit

Dependencies

No dependency data recorded yet.

Agent activity

21 hits · last 30 days

node

Amazon

OpenAI (training)

Resources

githubhttps://github.com/harbor-framework/harbor ↗

homepagehttps://github.com/harbor-framework/harbor ↗