fix(vscode): respect Git diff collapse attributes

.changeset/git-diff-attributes.md

@@ -0,0 +1,5 @@
+---
+"kilo-code": patch
+---
+
+Collapse files marked `linguist-generated` in repository `.gitattributes` rules while keeping explicitly visible files such as English and German translations expanded by default.

packages/kilo-vscode/src/agent-manager/local-diff-cache.ts

@@ -69,6 +69,7 @@ export function createDiffCache(load: Loader) {
       meta.additions,
       meta.deletions,
       meta.binary,
+      meta.generatedLike,
       meta.stamp,
     )
 

packages/kilo-vscode/src/agent-manager/local-diff.ts

@@ -1,4 +1,5 @@
 import * as fs from "fs/promises"
+import { classifyGenerated, generatedLike, gitGeneratedFiles } from "../diff/shared/git-attributes"
 import { imageMime, loadImage, readImageFile } from "../diff/shared/image"
 import { resolveInside } from "../diff/shared/path"
 import type { GitOps } from "./GitOps"
@@ -32,54 +33,7 @@ const MAX_SUMMARY_FILES = 32
 
 /** Ported from `packages/opencode/src/file/ignore.ts` — identical patterns,
  *  no runtime dependency on minimatch/picomatch. */
-const FOLDERS = new Set([
-  "node_modules",
-  "bower_components",
-  ".pnpm-store",
-  "vendor",
-  ".npm",
-  "dist",
-  "build",
-  "out",
-  ".next",
-  "target",
-  "bin",
-  "obj",
-  ".git",
-  ".svn",
-  ".hg",
-  ".vscode",
-  ".idea",
-  ".turbo",
-  ".output",
-  "desktop",
-  ".sst",
-  ".cache",
-  ".webkit-cache",
-  "__pycache__",
-  ".pytest_cache",
-  "mypy_cache",
-  ".history",
-  ".gradle",
-])
-
-const SUFFIXES = [".swp", ".swo", ".pyc", ".log"]
-const BASENAMES = new Set([".DS_Store", "Thumbs.db"])
-const CONTAINS_SEGMENTS = ["logs", "tmp", "temp", "coverage", ".nyc_output"]
-
-export function generatedLike(file: string): boolean {
-  const parts = file.split(/[/\\]/)
-  for (const part of parts) {
-    if (FOLDERS.has(part)) return true
-    if (CONTAINS_SEGMENTS.includes(part)) return true
-  }
-  for (const suffix of SUFFIXES) {
-    if (file.endsWith(suffix)) return true
-  }
-  const base = parts[parts.length - 1] ?? ""
-  if (BASENAMES.has(base)) return true
-  return false
-}
+export { generatedLike } from "../diff/shared/git-attributes"
 
 const BASE_CANDIDATES = ["main", "master", "dev", "develop"]
 
@@ -173,6 +127,18 @@ function statusFromCode(code: string): Status {
 }
 
 async function list(git: GitOps, dir: string, anc: string, log?: Log): Promise<Meta[]> {
+  const markGenerated = async (entries: Meta[]) => {
+    const configured = await gitGeneratedFiles(
+      git,
+      dir,
+      entries.map((entry) => entry.file),
+    )
+    return entries.map((entry) => ({
+      ...entry,
+      generatedLike: classifyGenerated(entry.file, configured),
+    }))
+  }
+
   const [tracked, untracked] = await Promise.all([
     git.execGit(["-c", "core.quotepath=false", "diff", "--raw", "--numstat", "--no-renames", anc], dir, {
       priority: true,
@@ -224,11 +190,11 @@ async function list(git: GitOps, dir: string, anc: string, log?: Log): Promise<M
 
   if (untracked.code !== 0) {
     log?.("git ls-files --others failed", { code: untracked.code, stderr: untracked.stderr.trim() })
-    return result
+    return markGenerated(result)
   }
 
   const files = untracked.stdout.trim()
-  if (!files) return result
+  if (!files) return markGenerated(result)
   const paths = files.split("\n").filter((file) => file && !seen.has(file))
 
   for (let index = 0; index < paths.length; index += MAX_SUMMARY_FILES) {
@@ -255,7 +221,7 @@ async function list(git: GitOps, dir: string, anc: string, log?: Log): Promise<M
     }
   }
 
-  return result
+  return markGenerated(result)
 }
 
 /**
@@ -295,6 +261,7 @@ async function detailMeta(
 ): Promise<Meta | undefined> {
   const full = resolveInside(dir, file)
   if (!full) return undefined
+  const configured = await gitGeneratedFiles(git, dir, [file], { signal })
   const tracked = await git.execGit(["ls-files", "--error-unmatch", "--", file], dir, { signal, priority: true })
   check(signal)
   if (tracked.code !== 0) {
@@ -312,7 +279,7 @@ async function detailMeta(
       deletions: 0,
       status: "added",
       tracked: false,
-      generatedLike: generatedLike(file),
+      generatedLike: classifyGenerated(file, configured),
       binary: value.binary,
       stamp: value.stamp,
     }
@@ -341,7 +308,7 @@ async function detailMeta(
     deletions: stat.deletions,
     status,
     tracked: true,
-    generatedLike: generatedLike(pathPart),
+    generatedLike: classifyGenerated(pathPart, configured),
     binary: stat.binary,
     stamp:
       status === "deleted"

packages/kilo-vscode/src/diff/shared/git-attributes.ts

@@ -0,0 +1,92 @@
+import type { GitOps } from "../../agent-manager/GitOps"
+
+const ATTRIBUTE = "linguist-generated"
+const FOLDERS = new Set([
+  "node_modules",
+  "bower_components",
+  ".pnpm-store",
+  "vendor",
+  ".npm",
+  "dist",
+  "build",
+  "out",
+  ".next",
+  "target",
+  "bin",
+  "obj",
+  ".git",
+  ".svn",
+  ".hg",
+  ".vscode",
+  ".idea",
+  ".turbo",
+  ".output",
+  "desktop",
+  ".sst",
+  ".cache",
+  ".webkit-cache",
+  "__pycache__",
+  ".pytest_cache",
+  "mypy_cache",
+  ".history",
+  ".gradle",
+])
+
+const SUFFIXES = [".swp", ".swo", ".pyc", ".log"]
+const BASENAMES = new Set([".DS_Store", "Thumbs.db"])
+const CONTAINS_SEGMENTS = ["logs", "tmp", "temp", "coverage", ".nyc_output"]
+
+export type GeneratedAttributes = ReadonlyMap<string, boolean>
+export type GeneratedFiles = (files: readonly string[]) => Promise<GeneratedAttributes>
+
+export function generatedLike(file: string): boolean {
+  const parts = file.split(/[/\\]/)
+  for (const part of parts) {
+    if (FOLDERS.has(part)) return true
+    if (CONTAINS_SEGMENTS.includes(part)) return true
+  }
+  for (const suffix of SUFFIXES) {
+    if (file.endsWith(suffix)) return true
+  }
+  const base = parts[parts.length - 1] ?? ""
+  return BASENAMES.has(base)
+}
+
+export function classifyGenerated(file: string, attrs?: GeneratedAttributes): boolean {
+  return attrs?.get(file) ?? generatedLike(file)
+}
+
+/**
+ * Read the repository's generated-file attributes for a set of paths.
+ * GitHub Linguist uses `linguist-generated`, and repositories already keep
+ * those rules in `.gitattributes` for pull-request diffs.
+ */
+export async function gitGeneratedFiles(
+  git: GitOps,
+  dir: string,
+  files: readonly string[],
+  options: { cached?: boolean; signal?: AbortSignal } = {},
+): Promise<Map<string, boolean>> {
+  const paths = [...new Set(files.filter(Boolean))]
+  if (paths.length === 0) return new Map()
+
+  const args = ["check-attr"]
+  if (options.cached) args.push("--cached")
+  args.push("-z", "--stdin", ATTRIBUTE)
+  const result = await git.execGit(args, dir, {
+    stdin: `${paths.join("\0")}\0`,
+    signal: options.signal,
+    priority: true,
+  })
+  if (result.code !== 0) return new Map()
+
+  const attrs = new Map<string, boolean>()
+  const fields = result.stdout.split("\0")
+  for (let field = 0; field + 2 < fields.length; field += 3) {
+    if (fields[field + 1] !== ATTRIBUTE) continue
+    const value = fields[field + 2]
+    if (value === "true" || value === "set") attrs.set(fields[field]!, true)
+    if (value === "false" || value === "unset") attrs.set(fields[field]!, false)
+  }
+  return attrs
+}

packages/kilo-vscode/src/diff/sources/catalog.ts

@@ -1,6 +1,7 @@
 import * as vscode from "vscode"
 import type { KiloConnectionService } from "../../services/cli-backend"
 import { GitOps } from "../../agent-manager/GitOps"
+import { gitGeneratedFiles } from "../shared/git-attributes"
 import { resolveLocalDiffTarget } from "../shared/target"
 import { appendOutput, getWorkspaceRoot } from "../../review-utils"
 import type { BranchListItem } from "../../agent-manager/git-import"
@@ -73,6 +74,8 @@ export class DiffSourceCatalog implements vscode.Disposable {
   // owned by the catalog so it survives source swaps.
   private branchGit: GitOps | undefined
   private branchOutput: vscode.OutputChannel | undefined
+  private attributeGit: GitOps | undefined
+  private attributeOutput: vscode.OutputChannel | undefined
 
   constructor(
     private readonly connection: KiloConnectionService,
@@ -100,6 +103,10 @@ export class DiffSourceCatalog implements vscode.Disposable {
 
   build(id: string, ctx: PanelContext): DiffSource {
     const opts = { dir: () => ctx.dir, strictDir: ctx.strictDir, git: ctx.git, log: ctx.log }
+    const dir = ctx.dir ?? ctx.workspaceRoot
+    const generated = dir
+      ? (files: readonly string[]) => gitGeneratedFiles(ctx.git ?? this.ensureAttributeGit(), dir, files)
+      : undefined
     if (id === WORKSPACE_SOURCE_ID) {
       return createWorktreeDiffSource({
         ...opts,
@@ -118,18 +125,13 @@ export class DiffSourceCatalog implements vscode.Disposable {
       if (!sessionId || !messageId) {
         throw new Error(`DiffSourceCatalog.build: malformed turn id "${id}" (expected turn:<sessionId>:<messageId>)`)
       }
-      return createTurnDiffSource(sessionId, messageId, this.turnFetch, ctx.workspaceRoot)
+      return createTurnDiffSource(sessionId, messageId, this.turnFetch, dir, generated)
     }
 
     if (id.startsWith(SESSION_PREFIX)) {
       const sessionId = id.slice(SESSION_PREFIX.length)
       if (!sessionId) throw new Error(`DiffSourceCatalog.build: empty session id in "${id}"`)
-      return createSessionDiffSource(
-        sessionId,
-        this.sessionFetch,
-        ctx.dir ?? ctx.workspaceRoot,
-        this.checkSnapshotsEnabled,
-      )
+      return createSessionDiffSource(sessionId, this.sessionFetch, dir, this.checkSnapshotsEnabled, generated)
     }
 
     throw new Error(`DiffSourceCatalog.build: unknown source id "${id}"`)
@@ -164,8 +166,12 @@ export class DiffSourceCatalog implements vscode.Disposable {
   dispose(): void {
     this.branchGit?.dispose()
     this.branchGit = undefined
+    this.attributeGit?.dispose()
+    this.attributeGit = undefined
     this.branchOutput?.dispose()
     this.branchOutput = undefined
+    this.attributeOutput?.dispose()
+    this.attributeOutput = undefined
   }
 
   private readonly branchLog = (...args: unknown[]) => {
@@ -179,4 +185,13 @@ export class DiffSourceCatalog implements vscode.Disposable {
     this.branchGit = new GitOps({ log: this.branchLog })
     return this.branchGit
   }
+
+  private ensureAttributeGit(): GitOps {
+    if (this.attributeGit) return this.attributeGit
+    this.attributeOutput = vscode.window.createOutputChannel("Kilo Diff: Attributes")
+    this.attributeGit = new GitOps({
+      log: (...args) => appendOutput(this.attributeOutput!, "DiffSourceCatalog", ...args),
+    })
+    return this.attributeGit
+  }
 }

packages/kilo-vscode/src/diff/sources/git-status.ts

@@ -3,7 +3,7 @@
 
 import * as fs from "fs/promises"
 import type { GitOps } from "../../agent-manager/GitOps"
-import { generatedLike } from "../../agent-manager/local-diff"
+import { classifyGenerated, generatedLike, gitGeneratedFiles } from "../shared/git-attributes"
 import { imageMime, readImageFile } from "../shared/image"
 import { resolveInside } from "../shared/path"
 import type { DiffFile } from "../types"
@@ -19,9 +19,43 @@ export interface FileEntry {
   deletions: number
   tracked: boolean
   binary: boolean
+  generatedLike?: boolean
   stamp?: string
 }
 
+export async function applyGeneratedAttributes(
+  git: GitOps,
+  dir: string,
+  entries: FileEntry[],
+  cached = false,
+): Promise<FileEntry[]> {
+  const configured = await gitGeneratedFiles(
+    git,
+    dir,
+    entries.map((entry) => entry.file),
+    { cached },
+  )
+  return entries.map((entry) => ({
+    ...entry,
+    generatedLike: classifyGenerated(entry.file, configured),
+  }))
+}
+
+export function createFileEntry(
+  item: { file: string; status: Status },
+  stats: Map<string, { additions: number; deletions: number; binary: boolean }>,
+): FileEntry {
+  const stat = stats.get(item.file)
+  return {
+    file: item.file,
+    status: item.status,
+    additions: stat?.additions ?? 0,
+    deletions: stat?.deletions ?? 0,
+    tracked: true,
+    binary: stat?.binary ?? false,
+  }
+}
+
 /** Parse `git diff --name-status` output into entries (status code + path). */
 export function parseNameStatus(stdout: string): { file: string; status: Status }[] {
   const out: { file: string; status: Status }[] = []
@@ -85,7 +119,7 @@ export function summarize(entry: FileEntry): DiffFile {
     deletions: entry.deletions,
     status: entry.status,
     tracked: entry.tracked,
-    generatedLike: generatedLike(entry.file),
+    generatedLike: entry.generatedLike ?? generatedLike(entry.file),
     // Binary metadata is complete because no deferred text body exists.
     // Images are the exception: their encoded sides load lazily on expansion.
     summarized: image || !entry.binary,