APRIL 23, 2025

Sourcegraph 6.2

A newer version of Sourcegraph is available: Sourcegraph 6.12.

Highlights

Personalized search ranking now live

Sourcegraph search results are now more relevant and tailored to each user.

We’ve improved how Sourcegraph ranks results in the main search bar to deliver more relevant results tailored to each user. When you search for something like data router, Sourcegraph now applies best-in-class ranking strategies that better understand your full query - not just keywords in isolation. In addition, we boost results from repositories you’ve recently contributed to, making it easier to find the code you care about faster.

Flexible code search queries - where you are looking for something, but don’t 100% remember the name - has been improved. Below is a before and after for the query contributors job that highlights this improvement and our use of BM25 ranking.

Before	After

Second, here is an example demonstrating the new personalized ranking. For a user who regularly contributes to sourcegraph/zoekt, a search would often include results from low-traffic repos that they don’t contribute to.

With the new personalized ranking, the top result is now from the sourcegraph/zoekt repository.

This improvement helps address a long-standing pain point for users in large codebases, especially new users who may otherwise get overwhelmed by results from unfamiliar or irrelevant repositories. With personalized search ranking, your search results are more likely to reflect the work that’s most relevant to you.

Batch spec library now customizable via API

Site admins can now customize batch spec library examples using new GraphQL APIs.

Site admins can now fully customize the examples shown in the batch spec library - visible when defining a batch spec for a new batch change - using three new GraphQL APIs:

createBatchSpecLibraryRecord
updateBatchSpecLibraryRecord
deleteBatchSpecLibraryRecord

With this change customers can now automate how they manage and display batch spec examples across their organization. With this API, they can now create their own examples, or even sync examples from their internal source of truth into Sourcegraph automatically.

Only site admins can modify the library, and there’s currently no UI for managing examples (though we’re open to building one based on feedback). End users can still view the examples, but won’t be able to make edits.

By default, the library displays up to 20 examples. However, admins can raise this limit to 1000 by toggling the batches-library-limit feature flag - though we recommend caution, as the UI isn’t optimized for that many entries yet.

Learn more about the batch spec library in our docs.

Find Batch Changes with title-based search

You can now search batch changes by title, making it easier to find the one you’re looking for - no more scrolling through long lists.

The Batch Changes list page now supports searching by title, letting you filter down results by typing keywords that match the batch change’s name. Previously, the list view only showed the 15 most recent batch changes by default, with the option to load 15 more at a time. For customers with long histories of batch changes - sometimes hundreds deep - this made it time-consuming to locate a specific one.

Now you can simply search by title to quickly narrow the list and jump straight to the change you care about, significantly improving the experience for teams managing large volumes of batch changes.

Batch Changes support for GitHub fine-grained access tokens (experimental)

Batch Changes now has experimental support for GitHub fine-grained access tokens for user and admin credentials.

Batch Changes now supports using GitHub fine-grained access tokens as credentials for creating and managing changesets. This support is experimental and is especially useful for customers who prefer not to use GitHub Apps due to their broad repo access, and who also disallow personal access tokens (PATs) because of their wide permissions and lack of expiration controls. Fine-grained access tokens strike a better balance: they’re more secure than PATs and more flexible than GitHub Apps.

They can be used for both user-level and site-admin credentials in Batch Changes. However, note that fine-grained tokens can be more complex to set up and may require admin approval in your GitHub organization.

Read more on how to use fine-grained access tokens in our docs.

Bulk rebase changesets available in Batch Changes

Batch Changes now supports a new bulk action to help teams working in fast-moving codebases - rebase changesets.

Batch Changes now supports a new bulk action to help teams working in fast-moving codebases: rebase changesets.

This makes it significantly easier for teams to keep changesets up to date. Instead of rebasing each changeset manually, users can now do it all at once with a single action—streamlining workflows and reducing friction when managing large-scale changes.

Learn more about Batch Changes in our docs.

Commit signing with SSH keys in Batch Changes (experimental)

Experimental support for SSH key–based commit signing in Batch Changes is now available for GitHub and GitLab.

You can now sign commits with SSH keys in Batch Changes - giving enterprise teams a more secure, user-level alternative to GitHub Apps for commit verification. This feature is available in experimental form and currently supports GitHub and GitLab.

To get started, users add a personal access token (PAT) and check a box to enable commit signing. When they do, Sourcegraph generates an SSH key pair, surfaces the public key, and guides users to add it to their code host as a signing key. Site admins can also add PATs, and they function as a global credential to all users who don't set their own.

Once set up, any commit made through Batch Changes will be automatically signed, and GitHub (or GitLab) will display the familiar “Verified” badge on the resulting pull or merge request.

SSH key-based signing is a great fit for orgs that:

• Disallow GitHub Apps or personal access tokens with broad scopes

• Want a clear and secure audit trail on code changes

• Need to enable commit signing at the user level, not just globally

Learn more about commit signing in our docs.

Claude 3.7 Sonnet now available in Cody

Claude 3.7 Sonnet is now available for Cody in two variants - standard and "extended-thinking" - to support both deep reasoning and fast, responsive edit workflows.

Developers using Cody chat now have access to two variations of Claude 3.7 Sonnet: the standard model useful for faster, more responsive interactions, and another optimized for reasoning ("extended thinking").

With this update, Claude 3.7 Sonnet can now be used in different contexts depending on whether long-form reasoning is required or for tasks where speed and performance is a priority.

The standard version of Claude 3.7 Sonnet will be the default chat model in the 6.2 release for Cloud customers. Self-hosted customers are encouraged to follow this recommendation, as Claude 3.7 Sonnet outperforms 3.5 in most scenarios.

This change applies across Cody Web, VS Code, JetBrains IDEs and you can learn more about model use in our docs.

Details

Improvements

Code Search Enable personalized ranking by default
We've enabled personalized ranking for search results in the new web app—marking the first step in integrating user-specific ranking signals into our search pipeline. As a result, queries spanning multiple repositories will prioritize results from repositories the user has frequently contributed to.

User can disable the feature by setting boostRelevantRepositories to false in the settings
```
{
  "experimentalFeatures": {
    "boostRelevantRepositories": false,
  },
}
```
Code Search Use `patterntype=nls` for Cody context
The GetCodyContext endpoint is now deprecated. Instead, clients should use the search API with patterntype=nls. Here's an example of how Cody calls the search API: (nlsSearchQuery).
Batch Changes Find batch changes with title-based search
Batch Changes Bulk rebase changesets available in Batch Changes
Batch Changes New api for batch spec examples
feat(batches): new batch spec examples library api
Batch Changes Add fine-grained access tokens to credential modal
feat(batches): Batch Changes now supports fine-grained access tokens as user and site-admin credentials
Batch Changes Commit signing with ssh keys
Feature is behind batches-commit-signing feature flag.

Adds the UI form elements needed for submitting a private SSH key and passphrase for git commit signing.

note: only UI elements and graphql changes in this PR, backend changes are still required.
Cody Claude 3.7 Sonnet now available in Cody
Cody Add a site config for "enforced" Guardrails and transmit it to Cody clients
- Adds a site config setting, "attribution.mode": "enforced", which configures Cody IDE extensions after v1.82 (VSCode) or v7.82 (JetBrains) to not display code until attribution checks have finished.
Agents Add rule stats to the `./api/rules` API
Added rule stats to the .api/agents/{agent_id}/rules REST endpoint when the query string ?include=revision.stats is present
Agents Auto-generate feedback from changeset conversations
When a pull request is merged, diagnostic feedback is auto-generated via LLM judge from the comments and reactions
Agents Add multiple table view UIs
- New UIs to view an agent repos, conversations, PRs, reviews, and diagnostics
Agents Mirror pull requests in our database
- It's now possible to POST /.api/agents/{agent_id}/runs to trigger a background review of a pull request
- It's now possible to list pull requests for an agent with GET /.api/agents/{agent_id}/changesets
- It's now possible to list repositories for an agent with GET /.api/agents/{agent_id}/repos
- It's now possible to trigger a background sync of pull requests via POST /.api/agents/{agent_id}/runs.
Agents Link to original rule URL in GitHub review comments
- Review comments on GitHub now link to the original rule. Previously, it only displayed the rule slug (short ID like logic) and it was difficult to find the source of the rule. Now, it's easy to find both the rule instructions or identify the rule author (via git blame).
Agents Add $-based cost estimates in agent overview
- There is now a new "Spend" tab in the agent overview that gives insights into how expensive it is to run an operate an agent, which is helpful among other reasons when making a decision about what LLM model to use.
- The LLM API /.api/completions/stream now returns the used "modelref" (Sourcegraph canonical representation) alongside the used "model" (LLM provider canonical representation).
Agents Rename all-inferred to repo-rules
- Use the setting rules: ["repo-rules"] instead of rules: ["all-inferred"] to pick up *.rule.md files in the repository based on the changed files in the diff.
Dev Add CLAUDE.md file for claude code
N/A
Smart Apply Add instant apply model for smart apply
- feat(smart-apply): Add instant apply models
Source Graphql: add databaseID field to repository graphql field
The repository graphql API now supports fetchinig the databaseID field.
Backport cac91705f82f1221766f1576cdadd0b5c2c7502a from #4953
Source Create different proxy interpretation modes for sub repo ip perms matching
authz: Add Perforce IP rule interpretation modes (unified (default), directOnly, proxyOnly) in site configuration via rulesInterpretationMode field to provide more predictable sub-repo permissions behavior
Backport 774ac0c653371b0acbae62e233764dc6729fddf3 from #4713
Source Create graphql endpoint for viewing sub repo perms for a repository

Backport 34dd46a20d9d9d3a52abaaeaa845b968bf781435 from #4528
Source Add projectQuery option to Gerrit code host connections
Gerrit code host connections now support a projectQuery option that allows for any arbitrary query parameters supported by the Gerrit API.
Source Allow adding details of an already existing GitHub App
Site admins can now add their own pre-existing GitHub Apps to Sourcegraph. This also allows site admins to supply the App details of Enterprise GitHub Apps.
Uncategorized Configurable runtime for codemonitor jobs
feat: configurable runtime for codemonitor jobs
Uncategorized Do-init-only flag in `sg release create`
NA
Uncategorized Patch release creates git branch for version
na
Uncategorized Implement commit search on searcher
- Added a feature flag to run commit search on searcher, goal being to improve stability and resource usage of gitserver
Uncategorized New GraphQL query evaluateFeatureFlags
feat: new GraphQL query evaluateFeatureFlags

Fixes

Code Search Fix bug that wasn't setting a default tab selection
- Fixes a bug where default selections were not being set.
Code Search Disallow rev filters with repo predicates
Fixes a search bug where queries like repo:has.file(...) rev:1.0 would completely ignore the rev filter. Now, these types of searches are explicitly disallowed.
Code Search Ensure we ignore context canceled errors in searcher client
Fix a rare bug in repo:has.file queries where some searches could return no results.
Code Search Truncate all Rockskip tables
With this change we delete all Rockskip indexes. In Sourcegraph v5.5.0 we shipped a bug in Rockskip which causes some unindexed symbol searches to return no results. The bug has been fixed in 6.2 and was backported to 6.1. However indexes that have been created prior to the fix are corrupted. Hence, we have to reindex all Rockskip repositories. A new index job will automatically be triggered. This requires no further action. However, during the reindex, symbol search at older revisions of very large repos will likely time out, so we encourage all customers to communicate this to their users ahead of the rollout of the new version.
Analytics Deprecate admin analytics
- Deprecate the admin analytics pages (/site-admin/analytics) and redirect users to Sourcegraph Analytics (analytics.sourcegraph.com).
Cody Add new bedrock fields to legacy completions config

Backport 0d9f70dd54a4c20621fe7404d3c7598e60264f59 from #4771
Cody Update auditlog API to have cursor-based pagination
- Improve usability of auditlog API
Cody Prevent panic in chat completions API by checking error first
- The API POST /.api/llm/chat/completions now does not panic anymore when sending a bad request (HTTP 400), which was possible to trigger if you used tool with an incorrect schema.
Agents Filter out empty diagnostic paths
- Fixed error "empty string is not a valid pathspec" when loading diagnostics page
  Backport ac2034bc8b71c5983a94946777cc1f75b168c287 from #4407
Agents Automatically add feature-flag trigger
- When creating a new agent program, it now automatically includes a feature flag trigger for the feature flag named agents-auto-review. You must create this feature flag to have the trigger activate.
Agents Skip draft PRs with the `feature_flag_enabled` trigger
- The review agent no longer automatically reviews draft PRs with the feature_flag_enabled trigger. Instead, users must explicitly request reviews on draft PRs through a comment (comment_starts_with) or a label (label_exists)
Agents Correctly chunk diffs wrt generated files
- Fix a bug where the review agent would error if a generated file had more than 100k tokens.
- Updated docs to describe the algorithm to detect and ignore generated files
Agents Mark run as "errored" on panic
- Fixed a bug where an agent run would be stuck in "running" mode forever. Now, these runs get correctly marked as "errored"
Agents Several small fixes for agents
- The review agent now posts a comment reply when it's triggered by a comment trigger (example: @sourcegraph review). This is makes the agent behave more like a colleague would.
- The agent run page now links back to the pull request at the top of the page
- "Duration: 0s" is no longer shown on the agent run page while it's running.
- The links to the "applied_rules" in the agent run page now work correctly. Previously, the links were a 404 because they were one long ", " joined string. We plan to further improve the design of how rules are displayed in this page, so this is not a final solution, just a quick fix for the most immediate bug.
Ci Make minor_nightly_cut.sh executable
na
Perf Remove quadratic behavior in rune counting
- (perf) Fixed quadratic behavior in commit and diff search
Release Add runtype names
na
Rr v2 and v1 next patch return same values
na
Smart Apply Change the deployment id for the custom model
- fix(smart-apply): Change deployment id
Source Sub-repo-perms: adjust ipv4 string prefix matcher to use stack instead of heap allocations
N/A
Uncategorized Fetch user by username for Bitbucket Server
- Fixes an issue where we OAuth would fail for Bitbucket Server users with special characters in their usernames
  Backport 28317328924ae0139edacc3711d169b37179c31f from #4490
Uncategorized Enforce file size limits for highlighting
- Enforces limits of 1.5 MiB and 50K lines for highlighting code files.
Uncategorized Unset cxx instead of NoCC
na
Uncategorized Use pathspec literal to avoid colons
- Fixed a bug where paths that start with colons may be reported as not found.
Uncategorized Batch workspaces to reduce memory pressure
fix(batches): batch workspaces to reduce memory pressure
Uncategorized Mark releases as development in the releaseregistry
NA
Uncategorized Remove vertical scrollbar from pre block
- (minor bugfix) Fixed an issue where we would show scrollbars on unscrollable items in the dynamic search filters sidebar
Uncategorized Nightly Minor Release Pipeline
NA
Uncategorized Call bazel directly for generating stitched migration graph in nightly release pipeline
NA
Uncategorized /latest releaseregistry endpoint returns latest by semver
NA
Uncategorized Nightly release pipeline must test do-init step
na