weavatrix 0.1.4 → 0.2.0

package/src/analysis/change-classification.js

@@ -0,0 +1,509 @@
+// Symbol-aware git-diff classification for change_impact. This module only decides WHAT changed and
+// which exact graph nodes are legitimate reverse-impact seeds; tools-impact owns traversal/formatting.
+// Pure additions intentionally produce no seeds by default, preventing a new export in a busy API file
+// from flooding the report with every legacy importer of that file.
+import { spawnSync } from "node:child_process";
+import { childProcessEnv } from "../child-env.js";
+
+const DEFAULT_LIMITS = Object.freeze({
+  maxDiffBytes: 2 * 1024 * 1024,
+  maxFiles: 500,
+  maxChangedLines: 20_000,
+  maxLineLength: 4_000,
+  maxSymbolsPerFile: 250,
+  maxSeeds: 1_000,
+});
+const CLASS_RANK = Object.freeze({
+  "metadata-only": 0,
+  added: 1,
+  "body-changed": 2,
+  "signature-changed": 3,
+  removed: 4,
+  unknown: 5,
+});
+const VERDICT_RANK = Object.freeze({ LOW: 0, MEDIUM: 1, HIGH: 2 });
+const STRUCTURAL_RELATIONS = new Set(["contains", "method"]);
+
+const normalizePath = (value) => String(value || "").replace(/\\/g, "/").replace(/^\.\//, "");
+const endpoint = (value) => String(value && typeof value === "object" ? value.id : value || "");
+const lineNumber = (value) => Number((String(value || "").match(/(?:^L|@)(\d+)$/) || [])[1] || 0);
+const bareLabel = (value) => String(value || "").replace(/\(.*$/, "").replace(/[^A-Za-z0-9_$].*$/, "").trim();
+const boundedNumber = (value, fallback, min, max) => Math.max(min, Math.min(max, Number(value) || fallback));
+
+function limitsOf(value = {}) {
+  return {
+    maxDiffBytes: boundedNumber(value.maxDiffBytes, DEFAULT_LIMITS.maxDiffBytes, 1_024, DEFAULT_LIMITS.maxDiffBytes),
+    maxFiles: boundedNumber(value.maxFiles, DEFAULT_LIMITS.maxFiles, 1, DEFAULT_LIMITS.maxFiles),
+    maxChangedLines: boundedNumber(value.maxChangedLines, DEFAULT_LIMITS.maxChangedLines, 10, DEFAULT_LIMITS.maxChangedLines),
+    maxLineLength: boundedNumber(value.maxLineLength, DEFAULT_LIMITS.maxLineLength, 80, DEFAULT_LIMITS.maxLineLength),
+    maxSymbolsPerFile: boundedNumber(value.maxSymbolsPerFile, DEFAULT_LIMITS.maxSymbolsPerFile, 1, DEFAULT_LIMITS.maxSymbolsPerFile),
+    maxSeeds: boundedNumber(value.maxSeeds, DEFAULT_LIMITS.maxSeeds, 1, DEFAULT_LIMITS.maxSeeds),
+  };
+}
+
+function decodeGitQuoted(value) {
+  const input = String(value || "").trim();
+  if (!input.startsWith('"')) return input.split("\t", 1)[0];
+  try { return JSON.parse(input); } catch { /* Git also emits octal escapes, which JSON rejects. */ }
+  const bytes = [];
+  for (let index = 1; index < input.length - 1; index += 1) {
+    const char = input[index];
+    if (char !== "\\") {
+      bytes.push(...Buffer.from(char));
+      continue;
+    }
+    const next = input[++index] || "";
+    if (/[0-7]/.test(next)) {
+      let octal = next;
+      while (octal.length < 3 && /[0-7]/.test(input[index + 1] || "")) octal += input[++index];
+      bytes.push(parseInt(octal, 8));
+    } else {
+      const escapes = { n: 10, r: 13, t: 9, b: 8, f: 12, v: 11, "\\": 92, '"': 34 };
+      bytes.push(escapes[next] ?? next.charCodeAt(0));
+    }
+  }
+  return Buffer.from(bytes).toString("utf8");
+}
+
+function diffPath(raw, prefix) {
+  const decoded = decodeGitQuoted(raw);
+  if (!decoded || decoded === "/dev/null") return null;
+  return normalizePath(decoded.startsWith(`${prefix}/`) ? decoded.slice(2) : decoded);
+}
+
+function headerPaths(line) {
+  const match = /^diff --git ("(?:\\.|[^"])*"|\S+) ("(?:\\.|[^"])*"|\S+)$/.exec(line);
+  if (!match) return { oldPath: null, newPath: null };
+  return { oldPath: diffPath(match[1], "a"), newPath: diffPath(match[2], "b") };
+}
+
+const emptyFile = (paths = {}) => ({
+  oldPath: paths.oldPath || null,
+  newPath: paths.newPath || null,
+  newFile: false,
+  deletedFile: false,
+  renamed: false,
+  binary: false,
+  hunks: [],
+  additions: [],
+  removals: [],
+});
+
+// Parse ordinary unified output, expecting --unified=0 but safely accepting context lines too.
+export function parseZeroContextDiff(diffText, options = {}) {
+  const limits = limitsOf(options);
+  const original = String(diffText ?? "");
+  const byteLength = Buffer.byteLength(original);
+  const oversized = byteLength > limits.maxDiffBytes;
+  const text = oversized ? original.slice(0, limits.maxDiffBytes) : original;
+  const files = [];
+  let file = null;
+  let hunk = null;
+  let changedLines = 0;
+  let truncated = oversized;
+  const finish = () => {
+    if (!file) return;
+    if ((file.oldPath || file.newPath) && files.length < limits.maxFiles) files.push(file);
+    else if (files.length >= limits.maxFiles) truncated = true;
+    file = null;
+    hunk = null;
+  };
+  const addChange = (kind, line, oldLine, newLine, mappedNewLine) => {
+    changedLines++;
+    if (changedLines > limits.maxChangedLines) { truncated = true; return; }
+    const change = {
+      kind,
+      text: String(line).slice(0, limits.maxLineLength),
+      ...(oldLine != null ? { oldLine } : {}),
+      ...(newLine != null ? { newLine } : {}),
+      mappedNewLine,
+    };
+    (kind === "added" ? file.additions : file.removals).push(change);
+  };
+
+  for (const line of text.split(/\r?\n/)) {
+    if (line.startsWith("diff --git ")) {
+      finish();
+      file = emptyFile(headerPaths(line));
+      continue;
+    }
+    if (!file && line.startsWith("--- ")) file = emptyFile();
+    if (!file) continue;
+    if (line.startsWith("new file mode ")) { file.newFile = true; continue; }
+    if (line.startsWith("deleted file mode ")) { file.deletedFile = true; continue; }
+    if (line.startsWith("rename from ")) { file.oldPath = normalizePath(decodeGitQuoted(line.slice(12))); file.renamed = true; continue; }
+    if (line.startsWith("rename to ")) { file.newPath = normalizePath(decodeGitQuoted(line.slice(10))); file.renamed = true; continue; }
+    if (line.startsWith("Binary files ") || line === "GIT binary patch") { file.binary = true; continue; }
+    if (!hunk && line.startsWith("--- ")) {
+      file.oldPath = diffPath(line.slice(4), "a");
+      if (!file.oldPath) file.newFile = true;
+      continue;
+    }
+    if (!hunk && line.startsWith("+++ ")) {
+      file.newPath = diffPath(line.slice(4), "b");
+      if (!file.newPath) file.deletedFile = true;
+      continue;
+    }
+    const match = /^@@ -(\d+)(?:,(\d+))? \+(\d+)(?:,(\d+))? @@/.exec(line);
+    if (match) {
+      hunk = {
+        oldStart: Number(match[1]), oldCount: match[2] == null ? 1 : Number(match[2]),
+        newStart: Number(match[3]), newCount: match[4] == null ? 1 : Number(match[4]),
+        oldCursor: Number(match[1]), newCursor: Number(match[3]),
+      };
+      file.hunks.push({ oldStart: hunk.oldStart, oldCount: hunk.oldCount, newStart: hunk.newStart, newCount: hunk.newCount });
+      continue;
+    }
+    if (!hunk || line.startsWith("\\ No newline")) continue;
+    if (line.startsWith("+")) {
+      addChange("added", line.slice(1), null, hunk.newCursor, hunk.newCursor);
+      hunk.newCursor++;
+    } else if (line.startsWith("-")) {
+      addChange("removed", line.slice(1), hunk.oldCursor, null, hunk.newCursor);
+      hunk.oldCursor++;
+    } else {
+      hunk.oldCursor++;
+      hunk.newCursor++;
+    }
+  }
+  finish();
+  return {
+    files,
+    byteLength,
+    changedLines: Math.min(changedLines, limits.maxChangedLines),
+    truncated,
+    oversized,
+    limits,
+  };
+}
+
+function graphIndex(graph, limits) {
+  const byFile = new Map();
+  for (const node of graph?.nodes || []) {
+    const file = normalizePath(node?.source_file || (!String(node?.id || "").includes("#") ? node?.id : ""));
+    if (!file) continue;
+    if (!byFile.has(file)) byFile.set(file, { path: file, fileNodeId: null, symbols: [] });
+    const record = byFile.get(file);
+    if (!String(node.id).includes("#")) record.fileNodeId = String(node.id);
+    else if (record.symbols.length < limits.maxSymbolsPerFile) {
+      const start = lineNumber(node.source_location) || lineNumber(node.id);
+      if (!start) continue;
+      record.symbols.push({
+        id: String(node.id),
+        label: String(node.label || node.id),
+        start,
+        end: lineNumber(node.source_end),
+        exported: node.exported === true,
+        symbolKind: node.symbol_kind || null,
+      });
+    }
+  }
+  for (const record of byFile.values()) {
+    record.symbols.sort((a, b) => a.start - b.start || a.id.localeCompare(b.id));
+    for (let index = 0; index < record.symbols.length; index += 1) {
+      const symbol = record.symbols[index];
+      if (!symbol.end || symbol.end < symbol.start) {
+        const next = record.symbols[index + 1]?.start;
+        symbol.end = next ? Math.max(symbol.start, next - 1) : symbol.start + 400;
+      }
+      symbol.end = Math.min(symbol.end, symbol.start + 2_000);
+    }
+  }
+  return byFile;
+}
+
+function isMetadataLine(text) {
+  const value = String(text || "").trim();
+  return !value || /^(?:\/\/|\/\*|\*|\*\/|#(?!include\b)|<!--|-->|"""|''')/.test(value);
+}
+
+function signatureText(text, symbol) {
+  const value = String(text || "").trim();
+  const name = bareLabel(symbol.label);
+  const hasName = name && new RegExp(`\\b${name.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")}\\b`, "i").test(value);
+  return hasName && /\b(?:export|default|declare|abstract|async|function|class|interface|type|enum|const|let|var|def|func|fn|struct|trait|impl|public|private|protected|static)\b/.test(value);
+}
+
+function signaturePosition(change, symbol) {
+  const candidates = change.kind === "removed" ? [change.mappedNewLine, change.oldLine] : [change.newLine];
+  if (candidates.some((line) => line === symbol.start)) return true;
+  const nearStart = candidates.some((line) => Number.isFinite(line) && line > symbol.start && line <= symbol.start + 4);
+  if (!nearStart) return false;
+  const value = String(change.text || "").trim();
+  // Multiline TS/Python/Java/Go parameter/type clauses. Keep statement keywords out so an ordinary
+  // first body line is not promoted to a public-contract change.
+  if (/^(?:return|throw|yield|if|for|while|switch|match|const|let|var|this\.|self\.|[A-Za-z_$][\w$]*\s*=)/.test(value)) return false;
+  return /^(?:[A-Za-z_$][\w$]*\??\s*:\s*[^;]+[,)]?|[A-Za-z_$][\w$<>,.?\[\] :*&]+\s+[A-Za-z_$][\w$]*\s*[,)]|[),:<>{}\[\]|&?]+)$/.test(value);
+}
+
+function moduleSignatureText(text) {
+  return /^\s*(?:import\b|export\s+(?:\*|\{)|(?:const|let|var)\s+\w+\s*=\s*require\b|using\b|package\b|#include\b|mod\b|pub\s+use\b)/.test(String(text || ""));
+}
+
+function chooseSymbol(record, change) {
+  if (!record?.symbols?.length) return null;
+  const lines = change.kind === "removed"
+    ? [...new Set([change.mappedNewLine, change.oldLine].filter((line) => Number.isFinite(line) && line > 0))]
+    : [change.newLine];
+  const candidates = [];
+  for (const symbol of record.symbols) {
+    if (!lines.some((line) => line >= symbol.start && line <= symbol.end)) continue;
+    const label = bareLabel(symbol.label);
+    const mentions = label && new RegExp(`\\b${label.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")}\\b`).test(change.text);
+    candidates.push({ symbol, mentions, span: symbol.end - symbol.start });
+  }
+  return candidates.sort((a, b) => Number(b.mentions) - Number(a.mentions) || a.span - b.span || a.symbol.id.localeCompare(b.symbol.id))[0]?.symbol || null;
+}
+
+function strongest(classes) {
+  return [...classes].sort((a, b) => (CLASS_RANK[b] ?? 5) - (CLASS_RANK[a] ?? 5) || a.localeCompare(b))[0] || "metadata-only";
+}
+
+function uniqueSorted(values, limit) {
+  const all = [...new Set(values.filter(Boolean).map(String))].sort((a, b) => a.localeCompare(b));
+  return { items: all.slice(0, limit), truncated: all.length > limit, total: all.length };
+}
+
+function analyzeParsedFile(parsed, indexed, { includeAddedSeeds }) {
+  const path = parsed.newPath || parsed.oldPath || "(unknown)";
+  const record = indexed.get(parsed.newPath) || indexed.get(parsed.oldPath) || null;
+  const grouped = new Map();
+  const unmapped = [];
+  for (const change of [...parsed.additions, ...parsed.removals]) {
+    const symbol = chooseSymbol(record, change);
+    if (!symbol) { unmapped.push(change); continue; }
+    if (!grouped.has(symbol.id)) grouped.set(symbol.id, { symbol, additions: [], removals: [] });
+    grouped.get(symbol.id)[change.kind === "added" ? "additions" : "removals"].push(change);
+  }
+
+  const symbols = [];
+  const seedIds = [];
+  for (const group of [...grouped.values()].sort((a, b) => a.symbol.start - b.symbol.start || a.symbol.id.localeCompare(b.symbol.id))) {
+    const addedCode = group.additions.filter((change) => !isMetadataLine(change.text));
+    const removedCode = group.removals.filter((change) => !isMetadataLine(change.text));
+    const addedDeclaration = addedCode.some((change) => signaturePosition(change, group.symbol) || signatureText(change.text, group.symbol));
+    const removedDeclaration = removedCode.some((change) => signaturePosition(change, group.symbol) || signatureText(change.text, group.symbol));
+    let classification;
+    if (parsed.newFile || (addedDeclaration && !removedCode.length)) classification = "added";
+    else if (parsed.deletedFile || (removedDeclaration && !addedCode.length)) classification = "removed";
+    else if (addedDeclaration || removedDeclaration) classification = "signature-changed";
+    else if (addedCode.length || removedCode.length) classification = "body-changed";
+    else classification = "metadata-only";
+
+    const reasons = {
+      added: "new declaration; existing callers cannot depend on it yet",
+      removed: "declaration removed; existing callers may break",
+      "signature-changed": "declaration/signature line changed",
+      "body-changed": "executable lines changed inside the symbol body",
+      "metadata-only": "only comment/blank lines changed in this symbol",
+    };
+    const symbolSeeds = [];
+    if (classification === "added") {
+      if (includeAddedSeeds) symbolSeeds.push(group.symbol.id);
+    } else if (classification !== "metadata-only") {
+      symbolSeeds.push(group.symbol.id);
+      if ((classification === "removed" || classification === "signature-changed") && group.symbol.exported && record?.fileNodeId) symbolSeeds.push(record.fileNodeId);
+    }
+    seedIds.push(...symbolSeeds);
+    symbols.push({
+      id: group.symbol.id,
+      label: group.symbol.label,
+      start: group.symbol.start,
+      end: group.symbol.end,
+      exported: group.symbol.exported,
+      classification,
+      reason: reasons[classification],
+      addedLines: group.additions.map((change) => change.newLine).filter(Boolean),
+      removedLines: group.removals.map((change) => change.oldLine).filter(Boolean),
+      seedIds: [...new Set(symbolSeeds)].sort(),
+    });
+  }
+
+  const unmappedCode = unmapped.filter((change) => !isMetadataLine(change.text));
+  const unmappedMetadata = unmapped.length > 0 && !unmappedCode.length;
+  let classification;
+  let reason;
+  if (parsed.binary) {
+    classification = "unknown";
+    reason = "binary diff has no line-level evidence";
+  } else if (parsed.deletedFile) {
+    classification = "removed";
+    reason = "file removed";
+  } else if (parsed.renamed) {
+    classification = "signature-changed";
+    reason = "file rename changes module identity";
+  } else if (parsed.newFile) {
+    classification = "added";
+    reason = "new file; no existing dependent can target it yet";
+  } else if (unmappedCode.some((change) => moduleSignatureText(change.text))) {
+    classification = "signature-changed";
+    reason = "module import/export surface changed outside a mapped symbol";
+  } else if (unmappedCode.length) {
+    classification = "unknown";
+    reason = "executable diff lines could not be mapped to a graph symbol";
+  } else if (symbols.length) {
+    classification = strongest(symbols.map((symbol) => symbol.classification));
+    reason = symbols.length === 1 ? symbols[0].reason : `${symbols.length} mapped symbols; strongest change is ${classification}`;
+  } else if (unmappedMetadata || parsed.hunks.length || parsed.additions.length || parsed.removals.length) {
+    classification = "metadata-only";
+    reason = "only comment/blank metadata changed outside symbols";
+  } else {
+    classification = "metadata-only";
+    reason = "file metadata changed without textual hunks";
+  }
+
+  if (parsed.binary || parsed.deletedFile || parsed.renamed || classification === "unknown") {
+    if (record?.fileNodeId) seedIds.push(record.fileNodeId);
+    if (parsed.binary || parsed.deletedFile || classification === "unknown") seedIds.push(...(record?.symbols || []).map((symbol) => symbol.id));
+  } else if (classification === "signature-changed" && !symbols.length && record?.fileNodeId) {
+    seedIds.push(record.fileNodeId);
+  } else if (classification === "added" && includeAddedSeeds) {
+    if (record?.fileNodeId) seedIds.push(record.fileNodeId);
+    seedIds.push(...symbols.map((symbol) => symbol.id));
+  }
+
+  return {
+    path,
+    oldPath: parsed.oldPath,
+    newPath: parsed.newPath,
+    classification,
+    reason,
+    binary: parsed.binary,
+    renamed: parsed.renamed,
+    addedLines: parsed.additions.length,
+    removedLines: parsed.removals.length,
+    symbols,
+    seedIds: [...new Set(seedIds)].sort(),
+  };
+}
+
+function unknownFile(path, indexed, reason) {
+  const normalized = normalizePath(path);
+  const record = indexed.get(normalized);
+  return {
+    path: normalized,
+    oldPath: normalized,
+    newPath: normalized,
+    classification: "unknown",
+    reason,
+    binary: false,
+    renamed: false,
+    addedLines: 0,
+    removedLines: 0,
+    symbols: [],
+    seedIds: [record?.fileNodeId, ...(record?.symbols || []).map((symbol) => symbol.id)].filter(Boolean).sort(),
+  };
+}
+
+function runGitDiff(repoRoot, base, _files, limits) {
+  const args = ["-C", repoRoot, "diff", "--no-ext-diff", "--find-renames", "--no-color", "--unified=0", String(base), "--"];
+  const result = spawnSync("git", args, {
+    encoding: "utf8",
+    windowsHide: true,
+    timeout: 12_000,
+    maxBuffer: limits.maxDiffBytes + 1,
+    env: childProcessEnv(),
+  });
+  if (result.status === 0) return { available: true, text: String(result.stdout || ""), error: null };
+  const oversized = result.error?.code === "ENOBUFS" || Buffer.byteLength(String(result.stdout || "")) > limits.maxDiffBytes;
+  return { available: false, text: String(result.stdout || ""), oversized, error: oversized ? "git diff exceeded the byte limit" : String(result.stderr || result.error?.message || "git diff unavailable").trim() };
+}
+
+const validExplicitFiles = (files) => [...new Set((Array.isArray(files) ? files : [])
+  .map(normalizePath)
+  .filter((file) => file && !file.startsWith("../") && !file.includes("/../") && !file.startsWith("-")))]
+  .sort((a, b) => a.localeCompare(b));
+
+export function classifyChangeImpact({
+  repoRoot = "",
+  graph = {},
+  base = "",
+  diffText,
+  files = [],
+  includeAddedSeeds = false,
+  limits: requestedLimits = {},
+} = {}) {
+  const limits = limitsOf(requestedLimits);
+  const explicitFiles = validExplicitFiles(files);
+  let source = "provided-diff";
+  let available = typeof diffText === "string";
+  let text = available ? diffText : "";
+  let unavailableReason = "";
+  let gitOversized = false;
+  if (!available && repoRoot && base) {
+    source = "git-diff";
+    const result = runGitDiff(repoRoot, base, explicitFiles, limits);
+    available = result.available;
+    text = result.text;
+    unavailableReason = result.error || "";
+    gitOversized = result.oversized === true;
+  } else if (!available) {
+    source = "files-only";
+    unavailableReason = "no unified diff was provided and no repoRoot/base pair was available";
+  }
+
+  const indexed = graphIndex(graph, limits);
+  const parsed = available ? parseZeroContextDiff(text, limits) : { files: [], changedLines: 0, byteLength: 0, truncated: gitOversized, oversized: gitOversized, limits };
+  const analyzed = parsed.files.map((file) => analyzeParsedFile(file, indexed, { includeAddedSeeds }));
+  const represented = new Set(analyzed.flatMap((file) => [file.oldPath, file.newPath].filter(Boolean)));
+  for (const file of explicitFiles) {
+    if (!represented.has(file)) analyzed.push(unknownFile(file, indexed, available ? "explicitly changed file had no textual hunk" : unavailableReason));
+  }
+  if (!available && !explicitFiles.length) analyzed.push(unknownFile("(diff unavailable)", indexed, unavailableReason));
+  analyzed.sort((a, b) => a.path.localeCompare(b.path));
+
+  if (parsed.truncated || parsed.oversized || gitOversized) {
+    for (const file of analyzed) {
+      file.classification = "unknown";
+      file.reason = "diff was truncated/oversized; symbol-level classification is incomplete";
+      const record = indexed.get(file.newPath) || indexed.get(file.oldPath);
+      file.seedIds = [...new Set([record?.fileNodeId, ...(record?.symbols || []).map((symbol) => symbol.id), ...file.seedIds].filter(Boolean))].sort();
+    }
+  }
+
+  const rawSeeds = analyzed.flatMap((file) => file.seedIds);
+  const seeds = uniqueSorted(rawSeeds, limits.maxSeeds);
+  let verdict = "LOW";
+  for (const file of analyzed) {
+    const next = ["removed", "signature-changed", "unknown"].includes(file.classification)
+      ? "HIGH"
+      : file.classification === "body-changed" ? "MEDIUM" : "LOW";
+    if (VERDICT_RANK[next] > VERDICT_RANK[verdict]) verdict = next;
+  }
+  if (!available || parsed.truncated || seeds.truncated) verdict = "HIGH";
+
+  const counts = Object.fromEntries(Object.keys(CLASS_RANK).map((name) => [name, analyzed.filter((file) => file.classification === name).length]));
+  const reasons = [];
+  if (!available) reasons.push(`Diff unavailable: ${unavailableReason || "unknown error"}; using conservative file/symbol seeds.`);
+  if (parsed.truncated || parsed.oversized || gitOversized) reasons.push("Diff exceeded a safety bound; incomplete evidence is classified HIGH/unknown.");
+  if (counts.removed || counts["signature-changed"]) reasons.push(`${counts.removed} removed and ${counts["signature-changed"]} signature/module-surface file change(s) can break existing callers.`);
+  if (counts["body-changed"]) reasons.push(`${counts["body-changed"]} file(s) contain mapped executable body changes.`);
+  if (counts.added && !includeAddedSeeds) reasons.push(`${counts.added} purely additive file change(s) create no dependent seeds by default.`);
+  if (counts["metadata-only"]) reasons.push(`${counts["metadata-only"]} metadata-only file change(s) create no dependent seeds.`);
+  if (counts.unknown) reasons.push(`${counts.unknown} file change(s) remain unknown and are seeded conservatively.`);
+  if (!reasons.length) reasons.push("No changed files were present in the supplied diff.");
+
+  return {
+    ok: available && !parsed.truncated && !seeds.truncated,
+    source,
+    verdict,
+    reasons,
+    seedIds: seeds.items,
+    files: analyzed,
+    summary: {
+      files: analyzed.length,
+      symbols: analyzed.reduce((sum, file) => sum + file.symbols.length, 0),
+      counts,
+      seeds: seeds.items.length,
+      totalSeedsBeforeCap: seeds.total,
+    },
+    bounds: {
+      ...limits,
+      diffBytes: parsed.byteLength,
+      changedLines: parsed.changedLines,
+      truncated: !!parsed.truncated || seeds.truncated,
+    },
+  };
+}
+
+export { DEFAULT_LIMITS as CHANGE_CLASSIFICATION_LIMITS };

package/src/analysis/cycle-route.js

@@ -0,0 +1,14 @@
+export const MAX_REPRESENTATIVE_CYCLE_FILES = 64;
+
+// Keep the complete closed route for normal cycles. Very large SCCs still get a closed, deterministic
+// head/tail route without letting one finding consume the whole audit response.
+export function formatRepresentativeCycle(cycle, maxFiles = MAX_REPRESENTATIVE_CYCLE_FILES) {
+  const raw = Array.isArray(cycle) ? cycle.map(String).filter(Boolean) : [];
+  if (!raw.length) return "";
+  const closed = raw.length > 1 && raw[0] === raw[raw.length - 1] ? raw : [...raw, raw[0]];
+  const cap = Math.max(4, Number(maxFiles) || MAX_REPRESENTATIVE_CYCLE_FILES);
+  if (closed.length <= cap + 1) return closed.join(" → ");
+  const side = Math.max(2, Math.floor(cap / 2));
+  const omitted = closed.length - side * 2;
+  return `${closed.slice(0, side).join(" → ")} → … ${omitted} file(s) omitted … → ${closed.slice(-side).join(" → ")}`;
+}

package/src/analysis/dead-check.js

@@ -9,13 +9,15 @@ import { readFileSync } from "node:fs";
 import { posix } from "node:path";
 import { createRepoBoundary } from "../repo-path.js";
 import { isStructuralRelation } from "../graph/relations.js";
+import { createPathClassifier, hasPathClass } from "../path-classification.js";
 
 const IDENT_RE = /[A-Za-z_$][\w$]*/g;
 const bareName = (label) => String(label || "").replace(/\s*\(.*$/, "").replace(/[()]/g, "").trim();
 // entry surfaces are never dead even with no inbound edge (framework/CLI/HTTP enter them externally).
 // Exported for internal-audit.js (reachability entry set) — keep the two in lockstep.
 export const ENTRY_FILE = /(^|[\\/])(index|main|app|server|cli|cmd|bootstrap|entry|run|__main__|manage|wsgi|asgi|setup|conftest)\.[a-z0-9]+$|(^|[\\/])(bin|cmd)[\\/]|(^|[\\/])main\.go$/i;
-const TEST_FILE = /(^|[\\/])[^\\/]*[._-](test|spec)\.[a-z0-9]+$|(^|[\\/])(test|tests|__tests__|spec)[\\/]/i;
+const defaultPathClassifier = createPathClassifier(null);
+const isTestFile = (file) => hasPathClass(defaultPathClassifier.explain(file), "test", "e2e");
 
 // Framework-owned entry modules are invoked by convention rather than a source import. Keep this narrow:
 // these are Next.js App/Pages Router surfaces and framework metadata files, not every file under `app/`.
@@ -144,7 +146,7 @@ export function computeDead(graph, sources, { entrySet = new Set() } = {}) {
   const deadSymbols = [];
   for (const n of symById.values()) {
     if (isReferenced(n)) continue;
-    const test = TEST_FILE.test(n.source_file);
+    const test = isTestFile(n.source_file);
     deadSymbols.push({ id: n.id, file: n.source_file, label: n.label, test, reason: "no inbound edge and name unreferenced outside its file" });
   }
   const deadSet = new Set(deadSymbols.map((s) => s.id));
@@ -202,7 +204,7 @@ export function computeUnusedExports(graph, sources, { dynamicTargets = new Set(
   for (const { n, nm } of candidates) {
     const occ = occursIn.get(nm);
     if (occ && [...occ].some((f) => f !== n.source_file)) continue; // referenced beyond its own file → alive
-    out.push({ id: n.id, file: n.source_file, label: n.label, test: TEST_FILE.test(n.source_file), reason: "exported but never imported or referenced outside its own file" });
+    out.push({ id: n.id, file: n.source_file, label: n.label, test: isTestFile(n.source_file), reason: "exported but never imported or referenced outside its own file" });
   }
   return out;
 }