feat(worker): Add ALWAYS_INDEX_FILE_PATTERNS env var to specify files that should always be indexed (#631)

Author: brendan-kellam

CHANGELOG.md

@@ -7,6 +7,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+- Added `ALWAYS_INDEX_FILE_PATTERNS` environment variable to allow specifying a comma seperated list of glob patterns matching file paths that should always be indexed, regardless of size or # of trigrams. [#631](https://github.com/sourcebot-dev/sourcebot/pull/631)
+
 ### Fixed
 - Fixed issue where single quotes could not be used in search queries. [#629](https://github.com/sourcebot-dev/sourcebot/pull/629)
 

docs/docs/configuration/environment-variables.mdx

@@ -35,6 +35,7 @@ The following environment variables allow you to configure your Sourcebot deploy
 | `SOURCEBOT_STRUCTURED_LOGGING_FILE` | - | <p>Optional file to log to if structured logging is enabled</p> | 
 | `SOURCEBOT_TELEMETRY_DISABLED` | `false` | <p>Enables/disables telemetry collection in Sourcebot. See [this doc](/docs/overview.mdx#telemetry) for more info.</p> |
 | `DEFAULT_MAX_MATCH_COUNT` | `10000` | <p>The default maximum number of search results to return when using search in the web app.</p> |
+| `ALWAYS_INDEX_FILE_PATTERNS` | - | <p>A comma separated list of glob patterns matching file paths that should always be indexed, regardless of size or number of trigrams.</p> |
 
 ### Enterprise Environment Variables
 | Variable | Default | Description |

docs/docs/connections/overview.mdx

@@ -69,6 +69,26 @@ To learn more about how to create a connection for a specific code host, check o
 
 <Note>Missing your code host? [Submit a feature request on GitHub](https://github.com/sourcebot-dev/sourcebot/issues/new?template=feature_request.md).</Note>
 
+## Indexing Large Files
+
+By default, Sourcebot will skip indexing files that are larger than 2MB or have more than 20,000 trigrams. You can configure this by setting the `maxFileSize` and `maxTrigramCount` [settings](/docs/configuration/config-file#settings).
+
+These limits can be ignored for specific files by passing in a comma separated list of glob patterns matching file paths to the `ALWAYS_INDEX_FILE_PATTERNS` environment variable. For example:
+
+```bash
+# Always index all .sum and .lock files
+ALWAYS_INDEX_FILE_PATTERNS=**/*.sum,**/*.lock
+```
+
+Files that have been skipped are assigned the `skipped` language. You can view a list of all skipped files by using the following query:
+```
+lang:skipped
+```
+
+## Indexing Binary Files
+
+Binary files cannot be indexed by Sourcebot. See [#575](https://github.com/sourcebot-dev/sourcebot/issues/575) for more information.
+
 
 ## Schema reference
 ---

packages/backend/src/zoekt.ts

@@ -1,5 +1,5 @@
 import { Repo } from "@sourcebot/db";
-import { createLogger } from "@sourcebot/shared";
+import { createLogger, env } from "@sourcebot/shared";
 import { exec } from "child_process";
 import { INDEX_CACHE_DIR } from "./constants.js";
 import { Settings } from "./types.js";
@@ -11,6 +11,8 @@ export const indexGitRepository = async (repo: Repo, settings: Settings, revisio
     const { path: repoPath } = getRepoPath(repo);
     const shardPrefix = getShardPrefix(repo.orgId, repo.id);
 
+    const largeFileGlobPatterns = env.ALWAYS_INDEX_FILE_PATTERNS?.split(',').map(pattern => pattern.trim()) ?? [];
+
     const command = [
         'zoekt-git-index',
         '-allow_missing_branches',
@@ -21,6 +23,7 @@ export const indexGitRepository = async (repo: Repo, settings: Settings, revisio
         `-tenant_id ${repo.orgId}`,
         `-repo_id ${repo.id}`,
         `-shard_prefix ${shardPrefix}`,
+        ...largeFileGlobPatterns.map((pattern) => `-large_file ${pattern}`),
         repoPath
     ].join(' ');
 

packages/shared/src/env.server.ts

@@ -219,6 +219,9 @@ export const env = createEnv({
 
         // Configure the default maximum number of search results to return by default.
         DEFAULT_MAX_MATCH_COUNT: numberSchema.default(10_000),
+
+        // A comma separated list of glob patterns that shwould always be indexed regardless of their size.
+        ALWAYS_INDEX_FILE_PATTERNS: z.string().optional(),
     },
     runtimeEnv,
     emptyStringAsUndefined: true,