fix(eval): accept target id in targets files

.agentv/targets.yaml

@@ -11,36 +11,36 @@ targets:
   # redirects to a named target, controlled via AGENT_TARGET env var.
   # One env var switches the entire provider config (auth, model, etc.).
   # Example: AGENT_TARGET=copilot-cli  or  AGENT_TARGET=claude
-  - label: default
+  - id: default
     use_target: ${{ AGENT_TARGET }}
 
-  - label: agent
+  - id: agent
     use_target: ${{ AGENT_TARGET }}
 
   # ── LLM target (text generation, no agent binary needed) ────────────
   # Delegates to LLM_TARGET — same provider used for grading and LLM evals.
-  - label: llm
+  - id: llm
     use_target: ${{ LLM_TARGET }}
 
   # ── Grader (LLM-as-judge) ──────────────────────────────────────────
   # Used by agent targets via grader_target. Switch provider via GRADER_TARGET.
-  - label: grader
+  - id: grader
     use_target: ${{ GRADER_TARGET }}
 
   # ── Named agent targets ───────────────────────────────────────────
-  - label: copilot
+  - id: copilot
     provider: copilot-cli
     model: ${{ COPILOT_MODEL }}
     grader_target: grader
     stream_log: raw
 
-  - label: copilot-sdk
+  - id: copilot-sdk
     provider: copilot-sdk
     model: ${{ COPILOT_MODEL }}
     grader_target: grader
     stream_log: raw
 
-  - label: copilot-sdk-azure
+  - id: copilot-sdk-azure
     provider: copilot-sdk
     model: ${{ AZURE_DEPLOYMENT_NAME }}
     subprovider: azure
@@ -49,22 +49,22 @@ targets:
     grader_target: grader
     stream_log: raw
 
-  - label: claude
+  - id: claude
     provider: claude-cli
     grader_target: grader
 
-  - label: claude-sdk
+  - id: claude-sdk
     provider: claude-sdk
     grader_target: grader
 
-  - label: pi
+  - id: pi
     provider: pi-cli
     subprovider: openrouter
     model: ${{ OPENROUTER_MODEL }}
     api_key: ${{ OPENROUTER_API_KEY }}
     grader_target: grader
 
-  - label: pi-sdk
+  - id: pi-sdk
     provider: pi-coding-agent
     subprovider: openai
     base_url: ${{ OPENAI_ENDPOINT }}
@@ -75,15 +75,15 @@ targets:
     stream_log: raw
 
 
-  - label: pi-azure
+  - id: pi-azure
     provider: pi-cli
     subprovider: azure
     base_url: ${{ AZURE_OPENAI_ENDPOINT }}
     model: ${{ AZURE_DEPLOYMENT_NAME }}
     api_key: ${{ AZURE_OPENAI_API_KEY }}
     grader_target: grader
 
-  - label: pi-sdk-azure
+  - id: pi-sdk-azure
     provider: pi-coding-agent
     subprovider: azure
     base_url: ${{ AZURE_OPENAI_ENDPOINT }}
@@ -93,7 +93,7 @@ targets:
     thinking: low
     stream_log: raw
 
-  - label: codex
+  - id: codex
     provider: codex
     executable: ${{ CODEX_EXECUTABLE }}
     model: ${{ CODEX_MODEL }}
@@ -104,7 +104,7 @@ targets:
     stream_log: raw
 
   # ── LLM targets (direct model access) ─────────────────────────────
-  - label: gh-models
+  - id: gh-models
     provider: openai
     base_url: https://models.github.ai/inference
     api_key: ${{ GH_MODELS_TOKEN }}
@@ -115,40 +115,40 @@ targets:
   # overridden via AZURE_OPENAI_API_VERSION. Chat-completions-only Azure
   # deployments must use `provider: openai` with a deployment-scoped
   # `base_url` instead.
-  - label: azure
+  - id: azure
     provider: azure
     endpoint: ${{ AZURE_OPENAI_ENDPOINT }}
     api_key: ${{ AZURE_OPENAI_API_KEY }}
     model: ${{ AZURE_DEPLOYMENT_NAME }}
     version: ${{ AZURE_OPENAI_API_VERSION }}
 
-  - label: gemini
+  - id: gemini
     provider: gemini
     api_key: ${{ GOOGLE_GENERATIVE_AI_API_KEY }}
     model: ${{ GEMINI_MODEL_NAME }}
 
-  - label: openai
+  - id: openai
     provider: openai
     endpoint: ${{ OPENAI_ENDPOINT }}
     api_key: ${{ OPENAI_API_KEY }}
     model: ${{ OPENAI_MODEL }}
 
   # Local OpenAI-compatible endpoint. Useful for dogfood against a local proxy
   # without changing provider-specific target labels.
-  - label: local-openai
+  - id: local-openai
     provider: openai
     base_url: ${{ LOCAL_OPENAI_PROXY_BASE_URL }}
     api_key: ${{ LOCAL_OPENAI_PROXY_API_KEY }}
     model: ${{ LOCAL_OPENAI_PROXY_MODEL }}
     grader_target: local-openai-grader
 
-  - label: local-openai-grader
+  - id: local-openai-grader
     provider: openai
     base_url: ${{ LOCAL_OPENAI_PROXY_BASE_URL }}
     api_key: ${{ LOCAL_OPENAI_PROXY_API_KEY }}
     model: ${{ LOCAL_OPENAI_PROXY_MODEL }}
 
-  - label: pi-cli-openai
+  - id: pi-cli-openai
     provider: pi-cli
     subprovider: openai
     base_url: ${{ LOCAL_OPENAI_PROXY_BASE_URL }}
@@ -158,7 +158,7 @@ targets:
     thinking: low
     stream_log: raw
 
-  - label: codex-sdk-openai
+  - id: codex-sdk-openai
     provider: codex
     base_url: ${{ LOCAL_OPENAI_PROXY_BASE_URL }}
     api_key: ${{ LOCAL_OPENAI_PROXY_API_KEY }}
@@ -168,7 +168,7 @@ targets:
     grader_target: local-openai-grader
     stream_log: raw
 
-  - label: copilot-sdk-openai
+  - id: copilot-sdk-openai
     provider: copilot-sdk
     subprovider: openai
     base_url: ${{ LOCAL_OPENAI_PROXY_BASE_URL }}
@@ -177,25 +177,25 @@ targets:
     grader_target: local-openai-grader
     stream_log: raw
 
-  - label: openrouter
+  - id: openrouter
     provider: openrouter
     api_key: ${{ OPENROUTER_API_KEY }}
     model: ${{ OPENROUTER_MODEL }}
 
   # ── MiMo (Xiaomi) via OpenRouter ───────────────────────────────────
-  - label: mimo
+  - id: mimo
     provider: openrouter
     api_key: ${{ OPENROUTER_API_KEY }}
     model: xiaomi/mimo-v2.5-pro
     grader_target: grader
 
-  - label: mimo-flash
+  - id: mimo-flash
     provider: openrouter
     api_key: ${{ OPENROUTER_API_KEY }}
     model: xiaomi/mimo-v2-flash
     grader_target: grader
 
-  - label: mimo-direct
+  - id: mimo-direct
     provider: openai
     base_url: https://token-plan-sgp.xiaomimimo.com/v1
     api_key: ${{ XIAOMI_MIMO_API_KEY }}

apps/cli/src/commands/eval/task-bundle.ts

@@ -625,14 +625,20 @@ function uniqueTargetDefinitions(
 }
 
 function serializeTargetDefinition(definition: TargetDefinition): Record<string, unknown> {
-  const target: Record<string, unknown> = { label: definition.name };
+  const target: Record<string, unknown> = { id: definition.name };
   if (definition.id !== undefined) {
     target.id = definition.id;
   }
   const config: Record<string, unknown> = {};
 
   for (const [key, value] of Object.entries(definition)) {
-    if (value === undefined || key === 'name' || key === 'id' || key === 'config') {
+    if (
+      value === undefined ||
+      key === 'name' ||
+      key === 'id' ||
+      key === 'label' ||
+      key === 'config'
+    ) {
       continue;
     }
     if (AUTHORING_TOP_LEVEL_TARGET_FIELDS.has(key)) {

apps/cli/src/commands/runs/rerun.ts

@@ -122,6 +122,9 @@ async function readTargetDefinitions(
 }
 
 function targetName(definition: Record<string, unknown>): string | undefined {
+  if (typeof definition.id === 'string' && definition.id.trim().length > 0) {
+    return definition.id.trim();
+  }
   if (typeof definition.label === 'string' && definition.label.trim().length > 0) {
     return definition.label.trim();
   }

apps/cli/src/templates/.agentv/targets.yaml

@@ -4,14 +4,14 @@
 # Agent and CLI targets use grader_target to reference an LLM target for scoring.
 
 targets:
-  - label: default
+  - id: default
     provider: azure
     endpoint: ${{ AZURE_OPENAI_ENDPOINT }}
     api_key: ${{ AZURE_OPENAI_API_KEY }}
     model: ${{ AZURE_DEPLOYMENT_NAME }}
     # version: ${{ AZURE_OPENAI_API_VERSION }}  # Optional: uncomment to override default (v1)
 
-  - label: codex
+  - id: codex
     provider: codex
     grader_target: azure-llm
     # Uses the Codex CLI (defaults to `codex` on PATH)
@@ -28,7 +28,7 @@ targets:
     stream_log: raw                    # Optional: 'summary' for consolidated logs or 'raw' for per-event logs
 
   # Claude - Anthropic's Claude Agent SDK
-  - label: claude
+  - id: claude
     provider: claude
     grader_target: azure-llm
     # Uses the @anthropic-ai/claude-agent-sdk
@@ -40,19 +40,19 @@ targets:
     stream_log: raw                           # Optional: 'summary' for consolidated logs or 'raw' for per-event logs
     # system_prompt: optional override (default instructs agent to include code in response)
 
-  - label: azure-llm
+  - id: azure-llm
     provider: azure
     endpoint: ${{ AZURE_OPENAI_ENDPOINT }}
     api_key: ${{ AZURE_OPENAI_API_KEY }}
     model: ${{ AZURE_DEPLOYMENT_NAME }}
     version: ${{ AZURE_OPENAI_API_VERSION }}
 
-  - label: gemini-llm
+  - id: gemini-llm
     provider: gemini
     api_key: ${{ GOOGLE_GENERATIVE_AI_API_KEY }}
     model: ${{ GEMINI_MODEL_NAME }}
 
-  - label: local_cli
+  - id: local_cli
     provider: cli
     grader_target: azure-llm
     # Passes the fully rendered prompt and any attached files to a local Python script
@@ -75,12 +75,12 @@ targets:
   #   mimo-v2.5        — 1M context, ~131K output, multimodal
   #   mimo-v2-flash    — 262K context, 65K output, fast MoE (open-source)
   #   mimo-v2-omni     — 262K context, 65K output, omni-modal
-  - label: mimo
+  - id: mimo
     provider: openrouter
     api_key: ${{ OPENROUTER_API_KEY }}
     model: xiaomi/mimo-v2.5-pro
 
-  - label: mimo-flash
+  - id: mimo-flash
     provider: openrouter
     api_key: ${{ OPENROUTER_API_KEY }}
     model: xiaomi/mimo-v2-flash

apps/cli/test/commands/eval/bundle.test.ts

@@ -54,11 +54,11 @@ describe('agentv eval bundle', () => {
     await writeFile(
       path.join(sourceDir, '.agentv', 'targets.yaml'),
       `targets:
-  - label: inherited
+  - id: inherited
     provider: mock
     response: '{"answer":"Mock provider response from inherited target"}'
     fallback_targets: [backup]
-  - label: backup
+  - id: backup
     provider: mock
     response: '{"answer":"Backup mock response"}'
 `,
@@ -153,8 +153,8 @@ tests: ../data/cases.yaml
     expect(input[0]?.content[0]).toEqual({ type: 'file', value: 'files/data/input.txt' });
 
     const bundledTargets = await readFile(path.join(bundleDir, 'targets.yaml'), 'utf8');
-    expect(bundledTargets).toContain('label: inherited');
-    expect(bundledTargets).toContain('label: backup');
+    expect(bundledTargets).toContain('id: inherited');
+    expect(bundledTargets).toContain('id: backup');
 
     await rm(sourceDir, { recursive: true, force: true });
     const run = await runCli(bundleDir, [
@@ -178,7 +178,7 @@ tests: ../data/cases.yaml
     await writeFile(
       path.join(sourceDir, 'evals', 'inline.eval.yaml'),
       `targets:
-  - label: candidate
+  - id: candidate
     provider: mock
     response: '{"answer":"inline bundled response"}'
 tests:
@@ -201,7 +201,7 @@ tests:
 
     expect(bundle.exitCode).toBe(0);
     const bundledTargets = await readFile(path.join(bundleDir, 'targets.yaml'), 'utf8');
-    expect(bundledTargets).toContain('label: candidate');
+    expect(bundledTargets).toContain('id: candidate');
     expect(bundledTargets).toContain('provider: mock');
     expect(bundledTargets).toContain('inline bundled response');
   }, 30_000);
@@ -214,7 +214,7 @@ tests:
     await writeFile(
       path.join(sourceDir, '.agentv', 'targets.yaml'),
       `targets:
-  - label: default
+  - id: default
     provider: mock
 `,
       'utf8',

apps/cli/test/commands/eval/targets.test.ts

@@ -18,15 +18,15 @@ describe('eval target selection', () => {
     await rm(tempDir, { recursive: true, force: true });
   });
 
-  it('resolves authored target ids through targets.yaml while keeping labels for display', async () => {
+  it('resolves authored target ids through targets.yaml', async () => {
     const agentvDir = path.join(tempDir, '.agentv');
     await mkdir(agentvDir, { recursive: true });
     await writeFile(
       path.join(agentvDir, 'targets.yaml'),
       [
         '$schema: agentv-targets-v2.2',
         'targets:',
-        '  - label: openai:gpt-5.4-mini',
+        '  - id: openai:gpt-5.4-mini',
         '    provider: mock',
         '',
       ].join('\n'),
@@ -38,7 +38,6 @@ describe('eval target selection', () => {
         'name: target-label-suite',
         'targets:',
         '  - id: openai:gpt-5.4-mini',
-        '    label: mini',
         'tests:',
         '  - id: target-case',
         '    input: hello',
@@ -60,7 +59,7 @@ describe('eval target selection', () => {
 
     expect(selections).toHaveLength(1);
     expect(selections[0]?.targetName).toBe('openai:gpt-5.4-mini');
-    expect(selections[0]?.targetLabel).toBe('mini');
+    expect(selections[0]?.targetLabel).toBeUndefined();
     expect(selections[0]?.resolvedTarget.kind).toBe('mock');
   });
 });

apps/cli/test/commands/grade/grade-prepared.test.ts

@@ -83,7 +83,7 @@ console.log(JSON.stringify({
     path.join(root, '.agentv', 'targets.yaml'),
     `
 targets:
-  - label: codex
+  - id: codex
     provider: cli
     command: bun ./scripts/target.ts
 `,