agentv 2.5.4 → 2.5.6

package/README.md

@@ -31,13 +31,9 @@ evalcases:
   - id: addition
     expected_outcome: Correctly calculates 15 + 27 = 42
 
-    input_messages:
-      - role: user
-        content: What is 15 + 27?
+    input: What is 15 + 27?
 
-    expected_messages:
-      - role: assistant
-        content: "42"
+    expected_output: "42"
 
     execution:
       evaluators:
@@ -108,8 +104,8 @@ See [AGENTS.md](AGENTS.md) for development guidelines and design principles.
 For large-scale evaluations, AgentV supports JSONL (JSON Lines) format as an alternative to YAML:
 
 ```jsonl
-{"id": "test-1", "expected_outcome": "Calculates correctly", "input_messages": [{"role": "user", "content": "What is 2+2?"}]}
-{"id": "test-2", "expected_outcome": "Provides explanation", "input_messages": [{"role": "user", "content": "Explain variables"}]}
+{"id": "test-1", "expected_outcome": "Calculates correctly", "input": "What is 2+2?"}
+{"id": "test-2", "expected_outcome": "Provides explanation", "input": "Explain variables"}
 ```
 
 Optional sidecar YAML metadata file (`dataset.yaml` alongside `dataset.jsonl`):
@@ -184,7 +180,7 @@ execution:
       script: ./validators/check_answer.py
 ```
 
-For complete templates, examples, and evaluator patterns, see: [custom-evaluators.md](.claude/skills/agentv-eval-builder/references/custom-evaluators.md)
+For complete templates, examples, and evaluator patterns, see: [custom-evaluators](https://agentv.dev/evaluators/custom-evaluators/)
 
 ### Compare Evaluation Results
 
@@ -238,7 +234,7 @@ Write validators in any language (Python, TypeScript, Node, etc.):
 ```
 
 For complete examples and patterns, see:
-- [custom-evaluators skill](.claude/skills/agentv-eval-builder/references/custom-evaluators.md)
+- [custom-evaluators](https://agentv.dev/evaluators/custom-evaluators/)
 - [code-judge-sdk example](examples/features/code-judge-sdk)
 
 ### LLM Judges
@@ -264,9 +260,7 @@ evalcases:
   - id: quicksort-explain
     expected_outcome: Explain how quicksort works
 
-    input_messages:
-      - role: user
-        content: Explain quicksort algorithm
+    input: Explain quicksort algorithm
 
     rubrics:
       - Mentions divide-and-conquer approach
@@ -281,7 +275,7 @@ Auto-generate rubrics from expected outcomes:
 agentv generate rubrics evals/my-eval.yaml
 ```
 
-See [rubric-evaluator skill](.claude/skills/agentv-eval-builder/references/rubric-evaluator.md) for detailed patterns.
+See [rubric evaluator](https://agentv.dev/evaluation/rubrics/) for detailed patterns.
 
 ## Advanced Configuration
 
@@ -310,9 +304,15 @@ Automatically retries on rate limits, transient 5xx errors, and network failures
 - AI agents: Ask Claude Code to `/agentv-eval-builder` to create and iterate on evals
 
 **Detailed Guides:**
-- [Evaluation format and structure](.claude/skills/agentv-eval-builder/SKILL.md)
-- [Custom evaluators](.claude/skills/agentv-eval-builder/references/custom-evaluators.md)
-- [Structured data evaluation](.claude/skills/agentv-eval-builder/references/structured-data-evaluators.md)
+- [Evaluation format and structure](https://agentv.dev/evaluation/eval-files/)
+- [Custom evaluators](https://agentv.dev/evaluators/custom-evaluators/)
+- [Rubric evaluator](https://agentv.dev/evaluation/rubrics/)
+- [Composite evaluator](https://agentv.dev/evaluators/composite/)
+- [Tool trajectory evaluator](https://agentv.dev/evaluators/tool-trajectory/)
+- [Structured data evaluators](https://agentv.dev/evaluators/structured-data/)
+- [Batch CLI evaluation](https://agentv.dev/evaluation/batch-cli/)
+- [Compare results](https://agentv.dev/tools/compare/)
+- [Example evaluations](https://agentv.dev/evaluation/examples/)
 
 **Reference:**
 - Monorepo structure: `packages/core/` (engine), `packages/eval/` (evaluation logic), `apps/cli/` (commands)