@ggui-ai/ui-gen 0.13.0 → 0.15.0

package/dist/evaluation/axis-checks/registry.d.ts

@@ -1,4 +1,4 @@
-import { A as AxisCheck } from '../../types-public-DIpjC-OA.js';
+import { A as AxisCheck } from '../../types-public-CkNWEov9.js';
 import '@ggui-ai/protocol';
 import '../../axes-CzLEMDeB.js';
 import '../../types-BOvHNG7K.js';

package/dist/evaluation/index.d.ts

@@ -1,10 +1,10 @@
-export { A as AxisCheck, a as AxisCheckInput, C as CRITERIA, D as DEFAULT_QUALITY_CONFIG, d as EvalCategory, e as EvalCriterion, E as EvalIssue, f as EvalOutcome, b as EvalResult, c as EvalTier, P as Priority, Q as QualityConfig, g as QualityMode, h as buildCodingCriteriaSummary, i as getActionableIssues, j as getCriteriaByPriority, k as getCriterionById, l as getLLMCriteria, n as isBlocked, m as matches, p as priorityForIssue } from '../types-public-DIpjC-OA.js';
+export { A as AxisCheck, a as AxisCheckInput, C as CRITERIA, d as CriterionCoverage, e as CriterionRunStatus, D as DEFAULT_QUALITY_CONFIG, f as EvalCategory, g as EvalCriterion, E as EvalIssue, h as EvalOutcome, b as EvalResult, c as EvalTier, L as LLM_EVAL_STATIC_CRITERIA, P as Priority, Q as QualityConfig, i as QualityMode, j as buildCodingCriteriaSummary, k as getActionableIssues, l as getCriteriaByPriority, n as getCriterionById, o as getLLMCriteria, p as isBlocked, m as matches, q as notApplicableCoverage, r as priorityForIssue } from '../types-public-CkNWEov9.js';
 import * as _anthropic_ai_claude_agent_sdk from '@anthropic-ai/claude-agent-sdk';
 import { McpServerConfig } from '@anthropic-ai/claude-agent-sdk';
 import { EvaluationContext, EvaluationConfig, QualityMetadata, EvaluationResult, EvaluationIssue } from './types.js';
 export { DEFAULT_EVAL_MAX_ROUNDS, DimensionScores, MAX_EVAL_ROUNDS_HARD_LIMIT } from './types.js';
-export { R as RunAxisChecksInput, r as runAxisChecks } from '../dispatch-D75M-OfZ.js';
-import { A as AgentConfig, L as LLMResponse } from '../llm-router-BEVtbAb9.js';
+export { R as RunAxisChecksInput, r as runAxisChecks } from '../dispatch-BuZfTh9r.js';
+import { A as AgentConfig, L as LLMResponse } from '../llm-router-D7fjBhtr.js';
 import '@ggui-ai/protocol';
 import '../axes-CzLEMDeB.js';
 import '../types-BOvHNG7K.js';

package/dist/evaluation/index.js

@@ -3,16 +3,12 @@ import Anthropic from '@anthropic-ai/sdk';
 import { z } from 'zod';
 import { listContractGadgets, HOOK_NAME_RE, STDLIB_GADGETS_PACKAGE } from '@ggui-ai/protocol';
 import 'esbuild';
+import '@ggui-ai/design/rendering';
 
 var __defProp = Object.defineProperty;
 var __getOwnPropNames = Object.getOwnPropertyNames;
-var __esm = (fn, res, err) => function __init() {
-  if (err) throw err[0];
-  try {
-    return fn && (res = (0, fn[__getOwnPropNames(fn)[0]])(fn = 0)), res;
-  } catch (e) {
-    throw err = [e], e;
-  }
+var __esm = (fn, res) => function __init() {
+  return fn && (res = (0, fn[__getOwnPropNames(fn)[0]])(fn = 0)), res;
 };
 var __export = (target, all) => {
   for (var name in all)
@@ -132,6 +128,23 @@ var init_client = __esm({
 });
 
 // src/evaluation/types-public.ts
+var LLM_EVAL_STATIC_CRITERIA = [
+  { criterion: "functionality", tier: 1 },
+  { criterion: "crash", tier: 1 },
+  { criterion: "interactivity", tier: 2 },
+  { criterion: "accessibility", tier: 2 },
+  { criterion: "layout", tier: 2 },
+  { criterion: "loading", tier: 2 },
+  { criterion: "visual", tier: 2 }
+];
+function notApplicableCoverage(reason) {
+  return LLM_EVAL_STATIC_CRITERIA.map(({ criterion, tier }) => ({
+    criterion,
+    tier,
+    status: "not-applicable",
+    reason
+  }));
+}
 var DEFAULT_QUALITY_CONFIG = {
   quality: "fast",
   visualEval: false,
@@ -380,7 +393,16 @@ function getBedrockModelId(model) {
   const BEDROCK_MAP = {
     "anthropic/claude-haiku-4-5": "us.anthropic.claude-haiku-4-5-20251001-v1:0",
     "anthropic/claude-sonnet-4-6": "us.anthropic.claude-sonnet-4-6",
-    "anthropic/claude-opus-4-6": "us.anthropic.claude-opus-4-6-v1:0"
+    "anthropic/claude-opus-4-6": "us.anthropic.claude-opus-4-6-v1:0",
+    // Claude 5-family (ggui#706, strings verified 2026-09-02 against
+    // platform.claude.com; Bedrock base form `anthropic.<id>`, upcast to
+    // the cross-region profile). Explicit so the ids are pinned by test,
+    // not merely reachable through the generic fallback below.
+    "anthropic/claude-fable-5-1": "us.anthropic.claude-fable-5-1",
+    "anthropic/claude-opus-5": "us.anthropic.claude-opus-5",
+    "anthropic/claude-sonnet-5": "us.anthropic.claude-sonnet-5",
+    // Legacy (still available; registry state `legacy`, not deprecated).
+    "anthropic/claude-fable-5": "us.anthropic.claude-fable-5"
   };
   const normalized = model.replace(/^anthropic\./, "anthropic/");
   if (BEDROCK_MAP[normalized]) return BEDROCK_MAP[normalized];
@@ -619,6 +641,32 @@ var LLMAgent = class {
     }
   }
 };
+function normalizeAnthropicModelId(model) {
+  return model.replace(/^anthropic\//, "").replace(/^(?:[a-z]{2}\.)?anthropic\./, "");
+}
+function anthropicRejectsSamplingParams(model) {
+  return /^claude-(?:opus-4-7|opus-5|sonnet-5|fable-5)(?:-|$)/.test(
+    normalizeAnthropicModelId(model)
+  );
+}
+function anthropicRejectsForcedToolChoice(model) {
+  return /^claude-fable-5-1(?:-|$)/.test(normalizeAnthropicModelId(model));
+}
+function resolveAnthropicSampling(model, temperature) {
+  if (temperature === void 0) {
+    return { request: {}, applied: { temperature: "provider-default" } };
+  }
+  if (anthropicRejectsSamplingParams(model)) {
+    return {
+      request: {},
+      applied: {
+        temperature: "provider-default",
+        strippedReason: `${normalizeAnthropicModelId(model)} rejects non-default sampling params (HTTP 400 on Opus 4.7+/5-family); caller's temperature=${temperature} was not sent`
+      }
+    };
+  }
+  return { request: { temperature }, applied: { temperature } };
+}
 var AnthropicAgent = class extends LLMAgent {
   provider = "anthropic";
   resolveModel(model) {
@@ -650,6 +698,10 @@ var AnthropicAgent = class extends LLMAgent {
         cache_control: { type: "ephemeral" }
       }
     ];
+    const sampling = resolveAnthropicSampling(resolvedModel, temperature);
+    if (sampling.applied.strippedReason !== void 0) {
+      console.warn(`[anthropic] callText: ${sampling.applied.strippedReason}`);
+    }
     try {
       const response = await this.apiCall(
         () => client.messages.create({
@@ -657,7 +709,7 @@ var AnthropicAgent = class extends LLMAgent {
           max_tokens: maxTokens ?? 4096,
           system,
           messages: [{ role: "user", content: userPrompt }],
-          ...temperature !== void 0 && { temperature }
+          ...sampling.request
         })
       );
       const usage = response.usage;
@@ -689,7 +741,8 @@ var AnthropicAgent = class extends LLMAgent {
       return {
         text,
         inputTokens: response.usage.input_tokens,
-        outputTokens: response.usage.output_tokens
+        outputTokens: response.usage.output_tokens,
+        sampling: sampling.applied
       };
     } catch (e) {
       emitLlmTraceEvent({
@@ -775,6 +828,12 @@ var AnthropicAgent = class extends LLMAgent {
       ...i === tools.length - 1 ? { cache_control: { type: "ephemeral" } } : {}
     }));
     try {
+      const appliedToolChoice = toolChoice === "required" && anthropicRejectsForcedToolChoice(resolvedModel) ? "auto" : toolChoice;
+      if (appliedToolChoice !== toolChoice) {
+        console.warn(
+          `[anthropic] callTools: ${normalizeAnthropicModelId(resolvedModel)} rejects forced tool_choice (HTTP 400); caller's 'required' downgraded to 'auto'`
+        );
+      }
       const response = await this.apiCall(
         () => client.messages.stream({
           model: resolvedModel,
@@ -782,7 +841,7 @@ var AnthropicAgent = class extends LLMAgent {
           system,
           messages: [{ role: "user", content: userPrompt }],
           tools: cachedTools,
-          tool_choice: { type: toolChoice === "required" ? "any" : "auto" }
+          tool_choice: { type: appliedToolChoice === "required" ? "any" : "auto" }
         }).finalMessage()
       );
       const usage = response.usage;
@@ -824,7 +883,8 @@ var AnthropicAgent = class extends LLMAgent {
         inputTokens: response.usage.input_tokens,
         outputTokens: response.usage.output_tokens,
         cacheReadTokens: cacheRead,
-        cacheCreationTokens: cacheCreated
+        cacheCreationTokens: cacheCreated,
+        appliedToolChoice
       };
     } catch (e) {
       emitLlmTraceEvent({
@@ -958,7 +1018,8 @@ var OpenAIAgent = class extends LLMAgent {
     return {
       text,
       inputTokens: response.usage?.input_tokens ?? 0,
-      outputTokens: response.usage?.output_tokens ?? 0
+      outputTokens: response.usage?.output_tokens ?? 0,
+      sampling: { temperature: temperature ?? "provider-default" }
     };
   }
   async callTools(model, systemPrompt, userPrompt, tools, toolChoice = "required", _scopedTools) {
@@ -1127,7 +1188,8 @@ var GoogleAgent = class extends LLMAgent {
     return {
       text: interactionText(interaction),
       inputTokens: interaction.usage?.total_input_tokens ?? 0,
-      outputTokens: interaction.usage?.total_output_tokens ?? 0
+      outputTokens: interaction.usage?.total_output_tokens ?? 0,
+      sampling: { temperature: temperature ?? "provider-default" }
     };
   }
   async callVision(model, systemPrompt, userPrompt, image, maxTokens) {
@@ -1409,7 +1471,8 @@ var OpenRouterAgent = class extends LLMAgent {
     return {
       text,
       inputTokens: response.usage.prompt_tokens,
-      outputTokens: response.usage.completion_tokens
+      outputTokens: response.usage.completion_tokens,
+      sampling: { temperature: temperature ?? "provider-default" }
     };
   }
   async callTools(model, systemPrompt, userPrompt, tools, toolChoice, _scopedTools) {
@@ -3261,6 +3324,6 @@ Evaluate the screenshot of the generated component.`;
   );
 }
 
-export { CRITERIA, DEFAULT_EVAL_MAX_ROUNDS, DEFAULT_QUALITY_CONFIG, MAX_EVAL_ROUNDS_HARD_LIMIT, buildCodingCriteriaSummary, buildFixPrompt, callMultimodalLLM, computeEvaluationScore, createEvaluationToolsServer, extractCompiledCode, extractCompiledCodeFromMessage, extractEvalResult, extractSourceCode, extractSourceCodeFromMessage, extractToolResultTexts, getActionableIssues, getCriteriaByPriority, getCriterionById, getEvaluatorSystemPrompt, getLLMCriteria, isBlocked, matches, priorityForIssue, runAxisChecks, runEvaluation, runEvaluationLoop };
+export { CRITERIA, DEFAULT_EVAL_MAX_ROUNDS, DEFAULT_QUALITY_CONFIG, LLM_EVAL_STATIC_CRITERIA, MAX_EVAL_ROUNDS_HARD_LIMIT, buildCodingCriteriaSummary, buildFixPrompt, callMultimodalLLM, computeEvaluationScore, createEvaluationToolsServer, extractCompiledCode, extractCompiledCodeFromMessage, extractEvalResult, extractSourceCode, extractSourceCodeFromMessage, extractToolResultTexts, getActionableIssues, getCriteriaByPriority, getCriterionById, getEvaluatorSystemPrompt, getLLMCriteria, isBlocked, matches, notApplicableCoverage, priorityForIssue, runAxisChecks, runEvaluation, runEvaluationLoop };
 //# sourceMappingURL=index.js.map
 //# sourceMappingURL=index.js.map