docs: Describe filesystem search exclusions

Document the shared relative-glob exclusion behavior for find and grep, and add release metadata for the large-tree performance changes.

Author: aron-cf

.changeset/faster-large-tree-operations.md

@@ -0,0 +1,5 @@
+---
+"@cloudflare/computer": minor
+---
+
+Add subtree exclusions to filesystem searches and reduce database and RPC work for sync, Worker shell tree walks, grep, and scoped Git diffs.

docs/04_filesystem_interface.md

@@ -392,6 +392,7 @@ interface GrepOptions {
   limit?: number;
   offset?: number;
   include?: string;
+  exclude?: string[];
 }
 
 interface WorkspaceGrepContextLine {
@@ -417,8 +418,9 @@ grep(
 Matching is literal and case-sensitive by default. Set `regex: true` to
 interpret `pattern` as a regular expression and `ignoreCase: true` to ignore
 letter case. `context` adds that many lines before and after each match.
-`include` is a glob relative to a searched directory. `limit` and `offset`
-paginate matching lines.
+`include` and `exclude` are globs relative to a searched directory. Exclusions
+use the same rules as `find` and prune matching directories before reading
+their contents. `limit` and `offset` paginate matching lines.
 
 `path` may be a directory or a single file. Directory searches return matches
 in deterministic depth-first discovery order, then line order within each

docs/09_tool_interface.md

@@ -197,7 +197,7 @@ The AI tool defaults to literal, case-sensitive matching. Set `regex: true` to i
 
 The tool passes `include`, `limit`, and `offset` through one `workspace.fs.grep` call. The storage search pages matching files and stops after the requested matches, so an included search does not build the full file or match list in the tool layer. Directory searches return matches in deterministic depth-first discovery order, then line order within each file. They are not globally sorted by full path.
 
-The lower-level `workspace.fs.grep` uses the same literal, case-sensitive defaults. Its options also accept `limit`, `offset`, `include`, `context`, `regex`, and `ignoreCase`.
+The lower-level `workspace.fs.grep` uses the same literal, case-sensitive defaults. Its options also accept `limit`, `offset`, `include`, `exclude`, `context`, `regex`, and `ignoreCase`. `exclude` uses the same relative exclusion globs as `find` and prunes matching directories before reading their contents.
 
 ## `write`
 

packages/dofs/README.md

@@ -47,7 +47,8 @@ export class WorkspaceDO extends DurableObject {
 - All filesystem primitives listed above are implemented and unit-tested.
   `readdir` returns size and modification time and supports stable
   `limit`/`offset` pages, including files held in pending write buffers.
-  `find` supports `*`, `**`, and `?` globs. `grep` supports bounded pages,
+  `find` supports `*`, `**`, and `?` globs. `find` and `grep` can prune
+  directories with relative exclusion globs. `grep` supports bounded pages,
   regular expressions or fixed strings, explicit case handling, and numbered
   context lines.
 - `SQLiteWorkspaceProvider` (the `@platformatic/vfs` adapter) implemented and exported from the package entrypoint; consumed by `@cloudflare/computerd`.

packages/dofs/src/fs/grep.ts

@@ -32,10 +32,9 @@ export interface GrepOptions {
   /** Glob relative to a searched directory that limits files. */
   include?: string;
   /**
-   * Whole-segment names to skip during traversal. Excluded directories
-   * are never descended into, so their files are never read. Unlike
-   * `include`, which filters files after the walk has already visited
-   * them, this prunes the walk itself.
+   * Globs relative to the searched directory that exclude entries.
+   * Matching directories are never descended into, so their files are
+   * not read. Exclusions use the same glob syntax as `include`.
    */
   exclude?: string[];
 }
@@ -125,9 +124,9 @@ function normalizeOptions(options: GrepOptions): {
 
 interface ScanTarget {
   path: string;
-  // Undefined when the caller grepped a single file directly: that
-  // path still needs a normal resolve. Traversal-produced targets
-  // carry the inode the walk already read.
+  // Targets normally carry the inode and size already read during path
+  // resolution or traversal. The optional fields keep the streaming
+  // path available to callers that only have a path.
   inode?: number;
   size?: number;
 }