@kevinrabun/judges 3.117.8 → 3.119.0

package/dist/tools/register-workflow.js

@@ -8,6 +8,7 @@ import { evaluateProject, evaluateDiff, analyzeDependencies, runAppBuilderWorkfl
 import { evaluateFilesBatch } from "../api.js";
 import { getGlobalSession } from "../evaluation-session.js";
 import { generatePublicRepoReport } from "../reports/public-repo-report.js";
+import { evaluateGitDiff, evaluateUnifiedDiff } from "../git-diff.js";
 import { configSchema, toJudgesConfig } from "./schemas.js";
 import { validateCodeSize } from "./validation.js";
 import { benchmarkGate, formatBenchmarkReport, formatBenchmarkMarkdown, runBenchmarkSuite, } from "../commands/benchmark.js";
@@ -20,6 +21,7 @@ export function registerWorkflowTools(server) {
     registerAppBuilderFlow(server);
     registerEvaluateProject(server);
     registerEvaluateDiff(server);
+    registerEvaluateGitDiff(server);
     registerAnalyzeDependencies(server);
     registerBenchmarkGate(server);
     registerBenchmarkDashboard(server);
@@ -919,3 +921,117 @@ function registerRecordFeedback(server) {
         };
     });
 }
+// ─── evaluate_git_diff ───────────────────────────────────────────────────────
+function registerEvaluateGitDiff(server) {
+    server.tool("evaluate_git_diff", "Evaluate code changes from a git diff. Parses the unified diff from a git repository, identifies changed files and lines, and runs the full tribunal on each changed file — filtering findings to only those on changed lines. Supports both live git repos (provide repoPath + base ref) and pre-computed diffs (provide diffText).", {
+        repoPath: z
+            .string()
+            .optional()
+            .describe("Absolute path to the git repository. Required when not providing diffText."),
+        base: z
+            .string()
+            .optional()
+            .describe("Git ref to diff against (e.g., 'main', 'HEAD~1', 'origin/main'). Default: 'HEAD~1'"),
+        diffText: z
+            .string()
+            .optional()
+            .describe("Pre-computed unified diff text. When provided, repoPath is used only for reading file contents."),
+        confidenceFilter: z
+            .number()
+            .min(0)
+            .max(1)
+            .optional()
+            .describe("Minimum confidence threshold for findings (default: no filter)"),
+        autoTune: z
+            .boolean()
+            .optional()
+            .describe("Apply feedback-driven auto-tuning to reduce false positives (default: false)"),
+        maxPromptChars: z
+            .number()
+            .min(0)
+            .optional()
+            .describe("Maximum character budget for LLM prompts. Controls truncation of deep-review prompts. Set to 0 to disable all truncation. Default: 100000."),
+        config: configSchema,
+    }, async ({ repoPath, base, diffText, confidenceFilter, autoTune, maxPromptChars, config }) => {
+        try {
+            const evalOptions = {
+                confidenceFilter,
+                autoTune,
+                maxPromptChars,
+                config: toJudgesConfig(config),
+            };
+            let result;
+            if (diffText) {
+                result = evaluateUnifiedDiff(diffText, repoPath ?? ".", evalOptions);
+            }
+            else if (repoPath) {
+                result = evaluateGitDiff(repoPath, base ?? "HEAD~1", evalOptions);
+            }
+            else {
+                return {
+                    content: [
+                        {
+                            type: "text",
+                            text: "Error: Provide either `repoPath` (for live git diff) or `diffText` (for pre-computed diff).",
+                        },
+                    ],
+                    isError: true,
+                };
+            }
+            let md = `# Git Diff Analysis\n\n`;
+            md += `**Files changed:** ${result.files.length}\n`;
+            md += `**Total findings:** ${result.totalFindings}\n\n`;
+            for (const file of result.files) {
+                md += `## ${file.filePath}\n`;
+                md += `**Verdict:** ${file.verdict.verdict} · **Score:** ${file.verdict.score}/100 · `;
+                md += `**Changed lines:** ${file.verdict.linesAnalyzed} · **Findings:** ${file.verdict.findings.length}\n\n`;
+                if (file.verdict.findings.length > 0) {
+                    for (const f of file.verdict.findings) {
+                        const conf = f.confidence !== undefined && f.confidence !== null ? ` (${Math.round(f.confidence * 100)}%)` : "";
+                        md += `- **${f.ruleId}** ${f.severity}${conf}: ${f.title}`;
+                        if (f.lineNumbers && f.lineNumbers.length > 0) {
+                            md += ` (L${f.lineNumbers.join(", L")})`;
+                        }
+                        md += `\n`;
+                    }
+                    md += `\n`;
+                }
+            }
+            const structured = {
+                filesAnalyzed: result.files.length,
+                totalFindings: result.totalFindings,
+                fileVerdicts: result.files.map((fv) => ({
+                    filePath: fv.filePath,
+                    verdict: fv.verdict.verdict,
+                    score: fv.verdict.score,
+                    changedLineCount: fv.verdict.linesAnalyzed,
+                    findingCount: fv.verdict.findings.length,
+                    findings: fv.verdict.findings.map((f) => ({
+                        ruleId: f.ruleId,
+                        severity: f.severity,
+                        confidence: f.confidence,
+                        title: f.title,
+                        lineNumbers: f.lineNumbers,
+                    })),
+                })),
+            };
+            return {
+                content: [
+                    { type: "text", text: md },
+                    { type: "text", text: "```json\n" + JSON.stringify(structured, null, 2) + "\n```" },
+                ],
+            };
+        }
+        catch (error) {
+            return {
+                content: [
+                    {
+                        type: "text",
+                        text: error instanceof Error ? `Error: ${error.message}` : "Error: Failed to evaluate git diff",
+                    },
+                ],
+                isError: true,
+            };
+        }
+    });
+}

package/dist/types.d.ts

@@ -670,6 +670,28 @@ export interface TribunalVerdict {
         /** Recommended action */
         recommendation: string;
     };
+    /**
+     * LLM deep-review prompt section. Present when `deepReview: true` is set
+     * in evaluation options. Contains a structured prompt that downstream LLM
+     * consumers can use for a second-pass analysis of the findings.
+     */
+    deepReviewPrompt?: string;
+    /**
+     * Auto-tune metadata. Present when `autoTune: true` is set and feedback
+     * data was applied. Records how many findings were suppressed or downgraded.
+     */
+    autoTuneApplied?: {
+        suppressed: number;
+        downgraded: number;
+    };
+    /**
+     * Confidence filter metadata. Present when `confidenceFilter` is set.
+     * Records how many findings were filtered out due to low confidence.
+     */
+    confidenceFilterApplied?: {
+        threshold: number;
+        filteredOut: number;
+    };
 }
 /**
  * Must-fix gate configuration for high-risk findings.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kevinrabun/judges",
-  "version": "3.117.8",
+  "version": "3.119.0",
   "description": "45 specialized judges that evaluate AI-generated code for security, cost, and quality.",
   "mcpName": "io.github.KevinRabun/judges",
   "type": "module",

package/server.json

@@ -7,12 +7,12 @@
     "url": "https://github.com/kevinrabun/judges",
     "source": "github"
   },
-  "version": "3.117.8",
+  "version": "3.119.0",
   "packages": [
     {
       "registryType": "npm",
       "identifier": "@kevinrabun/judges",
-      "version": "3.117.8",
+      "version": "3.119.0",
       "transport": {
         "type": "stdio"
       }