RolePlatform engineer & builderStatusShippedDocket No.012Opened2026-01
Structured evaluation for LLM systems

Agentic Evaluation SDK

An evaluation SDK that turns LLM interactions into structured cases, scored with independent metrics so agent behavior is testable.

LLM evaluationAgenticPython
Risk tierGate

Case summary

Problem

Model interactions are difficult to release safely when evaluation is a one-off report instead of a repeatable gate.

Outcome

LLM interactions become structured evaluation cases, captured, scored with independent metrics, and mapped to a deployment risk tier that can gate a release.

Chain of evidence

The first figure follows an interaction from capture to a risk tier.

01
Interactionapp or CLIcaptured as a case
02
Casestructured recordnormalized shape
03
Scoreheuristics or judgeindependent metrics
04
Aggregateone combined resultweighted across metrics
05
Risk tierGate
Risk tierlow to criticalgates the release

An interaction is captured, scored by independent metrics, aggregated, and mapped to a risk tier that can gate a release.

Runtime evaluation, without a rewrite

The second shows the runtime path: an existing Bedrock client wrapped and evaluated in the background, feeding the same classification.

1
Existing callunmodified code
2
Instrumented clientwraps the SDK
3
Trace + redactPII scrubbed
4
Background evaluatorGate
Background evaluatorbounded queue
5
Risk tiersame classification

An existing Bedrock client is wrapped, not replaced. Calls are traced, redacted, and scored in the background.

Evaluation zones

Capture zone

Any interaction, offline or live, normalizes into the same evaluation-case shape.

Scoring zone

Heuristic metrics and LLM-as-judge share one scoring contract, with an external framework as an optional layer.

Runtime zone

A wrapped Bedrock client evaluates calls in the background without touching invocation code.

Governance zone

Metric scores and data sensitivity together decide the risk tier that can gate a release.

Case narrative

I treat evaluation as a chain of evidence: capture an interaction as a structured evaluation case, score it with heuristic metrics or LLM-as-judge, aggregate into a single result, then map scores and data sensitivity to a risk tier that can gate a release. Metrics cover RAG grounding, agent trajectories, and safety. Existing model calls can be wrapped, traced, redacted, and evaluated in the background, with no rewrite of the invocation code. Tightrope is the contract-testing cousin of the same verify-before-trust idea. Nobody should both write and merge is the policy cousin of eval gating.

Scoring criteria

01
LLM evaluation

Structured rubrics and fixtures replace vibe checks on agent demos.

02
Python

Eval cases, rubrics, and scorers for Agentic Eval are Python.

03
Platform engineering

Eval fixtures and scorers are a shared path teams can reuse without reinventing rubrics.

04
Agentic

Captured trajectories become eval cases; scoring stays independent of the agent under test.

05
AI governance

Independent scoring keeps agent demos from self-grading their own trajectories.

06
Security

Eval corpora and scorers assume adversarial or mistaken agent output.

07
CLI

Operators run eval cases and rubrics from the CLI against captured trajectories.

08
Amazon Bedrock

Agentic Evaluation SDK runs foundation models through Bedrock with explicit model and eval hooks.

09
RAG

Agentic Evaluation SDK grounds answers in retrieved context and evaluates whether retrieval helped.

10
GitHub Copilot SDK

CopilotJudge drives LLM-as-judge calls through Copilot SDK sessions.

11
OpenAI SDK

OpenAIJudge is one of the ensemble judges for agent trajectory evaluation.