Subscribe: Our changelog includes an RSS feed that can integrate with Slack, email, Discord bots like Readybot or RSS Feeds to Discord Bot, and other subscription tools.
langsmith-0.14.0
LangSmith Self-Hosted v0.14 brings Polly (our AI assistant for traces and runs) to self-hosted, takes ABAC and audit logs GA (on by default), and enables the LLM Auth Proxy by default with URL allowlisting and richer JWT claims. Admins get unified model configurations shared across Agent Builder, Polly, Insights, Playground, and Evaluators, and fine-grained Prompt Owners for locking down who can promote or delete individual prompts. Evaluators gain multi-modal support and workspaces can now set cost alerts on tracing projects. Playground model support expands (Anthropic via Vertex AI, custom Azure models, Bedrock inference profiles, Gemini 3.1 Pro, GPT-5.3 / 5.4, Baseten + GLM-5), and new agent tools and triggers land for Google Sheets & Docs, Outlook, Teams, and Salesforce SOQL. On the infrastructure side, v0.14 adds GCS Workload Identity support for blob storage, Valkey as a drop-in Redis replacement, and a pre-upgrade migration hook for safer rollouts.Follow the upgrade instructions to get access to everything. To book time with LangChain support for your upgrade, contactsupport@langchain.dev.Breaking changes
- Fixed an issue where
host-backendwasn’t picking upcommonEnv. This may result in duplicate environment variables that need to be removed.
Infrastructure changes
- Migrations now run as a
Pre-upgradehook prior to image versions rolling out. This will prevent issues when migrations fail. - GCS Workload Identity support — authenticate to GCS blob storage using cloud-native workload identity instead of long-lived credentials.
- Valkey support — Valkey can now be used as a drop-in replacement for Redis.
New features
- Polly on self-hosted — our AI assistant for understanding traces, runs, and evaluator feedback is now available in self-hosted.
- ABAC and audit logs GA — Attribute-Based Access Control and audit logs are enabled by default for self-hosted deployments.
- LLM Auth Proxy on by default — URL allowlist prevents credential forwarding to unintended hosts, and JWTs now carry
organization_nameandworkspace_nameclaims. - Unified model configurations — Agent Builder, Polly, Insights, Playground, and Evaluators now share a single set of model configs, with workspace-admin controls over model access across all AI features.
- Prompt Owners — designate a specific group of users with fine-grained permission to promote or delete individual prompts, without granting broader org access.
- Multi-modal evaluators — pass attachments and base64 content (images, audio, PDFs) directly into evaluators.
- Cost alerts on tracing projects — set alerts on tracing project-level costs alongside existing LangSmith alerts.
- Expanded playground model support — Anthropic via Vertex AI, custom Azure models, Bedrock inference profiles and configurable base URLs, Gemini 3.1 Pro, GPT-5.3 / 5.4 (now default), and Baseten + GLM-5.
- New agent tools and triggers — Google Sheets and Docs, Outlook mail and calendar, Microsoft Teams, Salesforce SOQL, Gmail OAuth v2 with refresh tokens, and an Outlook Trigger.
- Insights enhancements — scheduled Insights reports, categories trending over time, full feedback comments in analysis, and a lower minimum job interval (6h → 1h).
- Annotation and review upgrades — required reviewers per queue, pairwise queues that honor
reviewer_access_mode, an “Assigned to me” filter, per-annotator CSV export, and bulk table actions. - Prompt Hub and tool registry — commit tag search, model select in template creation, a workspace-scoped tool registry API, and a private registry UI.
- Evaluator workflow improvements — prebuilt LLM evaluators use strict structured outputs by default, evaluators support tagging and reuse, retries no longer lose scores, and a new API runs playground experiments programmatically.
- Custom iframe output renderer — drag-to-resize HTML chart outputs in experiments and trace views.
- Thread and inbox UX — auto-generated thread titles, redesigned run details in threads, session-level feedback stats, keyboard shortcuts, and filtering out internal helper threads.
Admin changes
- Granular usage reporting — granular billable usage APIs that allow you to retrieve detailed trace usage data broken down by workspace, project, user, or API key.
langsmith-auth-proxy-0.0.10
- Internal improvements and maintenance updates
langsmith-auth-proxy-0.0.10.tgzlangsmith-0.13.43
- Fixed a bug to recognize
json.Numberas a primitive in metadata filtering, improving data handling in self-hosted deployments.
langsmith-0.13.43.tgzlangsmith-auth-proxy-0.0.10
- Fixed recognition of
json.Numberas a primitive in metadata filtering, improving data ingestion accuracy.
langsmith-auth-proxy-0.0.10.tgzlangsmith-0.13.42
- Fixed issue in metadata filtering to recognize json.Number as a primitive type, improving data ingestion accuracy.
langsmith-0.13.42.tgzlangsmith-0.13.41
- Internal improvements and maintenance updates
langsmith-0.13.41.tgzlangsmith-0.13.40
- Added support for mTLS configuration to enhance self-hosted security.
- Improved the loading speed of the Fleet interface.
- Fixed a bug in the tracing UI that caused intermittent display issues.
- Added support for Redis Cluster, improving scalability for self-hosted deployments.
- Improved PostgreSQL IAM integration for better database management in self-hosted instances.
langsmith-0.13.40.tgzlangsmith-0.13.39
- Users can now run experiments without
projects:create, decoupling experiment execution from project governance controls. - Added skeleton loading state when switching between agent chat threads instead of a blank chat input.
- Improved the Fleet Arcade integrations page to display correct actions and clearer backend states.
- Arcade gateway installs now automatically sanitize invalid MCP server names before adding them to a workspace.
- Users can now see assigned reviewers as name chips in the annotation queue list and filter to queues they are assigned to via an “Assigned to me” button.
- Improved run details hover on the trace tree for less flicker when clicking or moving quickly between rows.
- Fixed a timeout when generating Insights reports with feedback enabled on high-volume workspaces.
- Arcade integrations now show a permission error instead of a generic upstream failure when a connected user lacks access to the configured Arcade project.
- Added a clickable ID badge to the experiment detail page header for easy copying of the experiment ID.
- LLM auth proxy JWTs now include
organization_nameandworkspace_nameclaims. - Added dataset split selection to the evaluator playground, allowing users to run evaluator experiments on specific dataset splits.
- Fixed experiment comparison view showing contradictory improvement arrows and regression cell colors for composite scores.
- Fixed a bug where deleting traces from pre-compression multipart blob storage objects could corrupt byte ranges for other traces in the same object, causing 416 errors when reading their payloads.
- Fixed crash when viewing a tool run in an in-progress trace.
- Lowered Insights job schedule minimum interval from 6 hours to 1 hour, configurable via
CLIO_SCHEDULE_MIN_INTERVAL_SECONDSenvironment variable. - LLM auth proxy now supports Insights (CLIO) service identity for JWT-based LLM authentication.
- Fixed subagent files not being deleted from hub repo memory when subagents are removed in the agent editor.
- Simplified Arcade workspace connection status to show “Workspace configured” instead of the confusing “Connected account” label with a redundant badge.
- Fixed a bug where expanding a tool call in the run detail view and scrolling away caused it to collapse back when scrolled into view again.
- Agent file reads (clone, inspect, startup) now correctly use the hub as the source of truth when hub memory is enabled.
- Fixed custom output rendering (HTML charts via iframe) not working in the redesigned experiment detail panes.
- Allowed usage of LLM Auth Proxy in self-hosted by default.
- Sessions facets raw path now returns input/output KV facets when
RUNS_LITE_STATS_TENANTSis enabled, matching the Python backend behavior. - Click-to-copy tooltips (e.g. project ID) now respond to clicks on the tooltip content itself, not just the trigger badge.
- Improved MCP server authorization enforcement in the platform backend.
- Added Baseten as a model provider with GLM-5 support.
- Improved LLM inference efficiency by reducing date precision in system prompts from minute-level to date-only.
- Fixed Polly losing trace context when a trace page is expanded to full page view.
- Added a unified files sidebar to Fleet chat, accessible via a “Files” button in the header, to browse, search, create, rename, move, and preview all agent-generated files in one panel.
- Added
SSRF_ALLOW_PRIVATE_IPS_WEBHOOKS,SSRF_ALLOW_PRIVATE_IPS_MCP_SERVERS, andSSRF_ALLOW_PRIVATE_IPS_TOOLSenvironment variables to allow self-hosted deployments to connect to services on private IP ranges. - Added
get_current_timetool to Polly so relative time expressions in filter queries resolve correctly. - Reference outputs are now always visible in the experiment trace detail view, fixing an issue where they were hidden for some organizations.
- Fixed agent OAuth connections failing with “Unknown provider” errors for non-personal agents.
- Added Base URL configuration support for Bedrock models in the playground and model configurations, enabling custom endpoint URLs for proxy or gateway deployments.
- Added sign out button back to the settings page sidebar.
- The run rules list endpoint now returns
backfill_idwhen backfill progress is requested. - SSO users can no longer see or accept pending invites to organizations other than their own SSO organization.
- Fleet webhook execution now uses a dedicated endpoint instead of the MCP proxy.
- Added session-level feedback stats to the sessions API for parity with the Python backend.
- Improved MCP proxy authorization and URL safety checks for agent runtime requests.
langsmith-0.13.39.tgzlangsmith-0.13.38
- Fixed MCP OAuth tools (e.g., Hex, Notion) failing on self-hosted deployments when
HOST_BACKEND_ENDPOINT_PUBLIClacked anhttps://scheme. - Insights agent now includes full feedback comments when analyzing traces.
- Removed the legacy Feed page in Fleet; Inbox is now the default thread view for all tenants.
- Fixed a permission error that blocked users from creating evaluators when using the evaluator reuse feature.
- Org admins can now grant Model Configuration management to workspace editors and custom roles via Settings > Roles.
- Pairwise annotation queues now respect
reviewer_access_mode: completion/archive logic gates on assigned reviewers only, and GET responses includeassigned_reviewers. - Agent chat messages are no longer grouped into a collapsible “Completed N steps” container. A single copy dropdown on the last AI message of each turn lets you copy just the response or all steps including tool outputs.
- Fixed pagination in the waterfall view for thread traces, where threads with more than 20 turns now load additional traces when scrolling.
- Fixed “Empty Message” text appearing in the agent editor chat when AI invokes tools without accompanying text.
- Added Salesforce SOQL query tool to Fleet, enabling agents to query Salesforce data via OAuth.
- Annotation queue run list items now show reviewer names and avatars when hovering over the review stats badge.
- Go session stats now return
feedback_key,feedback_key_score,feedback_value, andfeedback_sourcefacets inrun_facets, matching the Python backend. - Fixed
/infoendpoint returning 401 wheninfoEndpointAuthRequiredis enabled with SSO authentication. - Added a documentation link button to the “Trigger Webhook” section in the Save prompt dialog for easier access to webhook docs.
- Fixed layout shift in agent chat caused by the copy button unmounting during streaming.
langsmith-0.13.38.tgzlangsmith-0.13.37
- Added URL allowlist for the LLM Auth Proxy to prevent credential forwarding to unintended hosts.
- Enabled audit logs by default for self-hosted deployments.
- MCP servers now respect granular RBAC permissions in the UI; users only see actions their role allows.
- Enabled ABAC by default for self-hosted deployments.
- Fixed low-frequency SmithDB ingestion errors (“incomplete chunked payload”) during multipart uploads when the HTTP body is truncated.
- Fixed a bug with OpenAI tools rendering.
langsmith-0.13.37.tgzlangsmith-0.13.36
- Internal improvements and maintenance updates.
langsmith-0.13.36.tgzlangsmith-0.13.35
- Internal improvements and maintenance updates.
langsmith-0.13.35.tgzlangsmith-0.13.34
- Internal improvements and maintenance updates.
langsmith-0.13.34.tgzlangsmith-0.13.33
- Internal improvements and maintenance updates.
langsmith-0.13.33.tgzlangsmith-0.13.32
- Internal improvements and maintenance updates.
langsmith-0.13.32.tgzlangsmith-0.13.31
- Internal improvements and maintenance updates.
langsmith-0.13.31.tgzlangsmith-0.13.30
- Internal improvements and maintenance updates.
langsmith-0.13.30.tgzlangsmith-0.13.29
- Internal improvements and maintenance updates.
langsmith-0.13.29.tgzlangsmith-0.13.28
- Internal improvements and maintenance updates.
langsmith-0.13.28.tgzlangsmith-0.13.27
- Internal improvements and maintenance updates.
langsmith-0.13.27.tgzlangsmith-0.13.26
- Sub-agents spawned during Fleet conversations now display real-time status cards inline in the chat, with a detail sidebar showing the sub-agent’s live timeline, tool calls, and results.
- Prebuilt LLM evaluators now use strict structured output mode by default. Strict mode automatically toggles when switching between OpenAI and non-OpenAI model providers.
- Fixed a bug where online evaluator scores were lost when the evaluation succeeded on a retry but the total job time exceeded the queue timeout.
- Deleting a run from an annotation queue now fully removes it instead of incorrectly marking it as completed.
- Included per-annotator feedback in experiments CSV export.
- Uploading dataset examples with invalid UTF-8 in inputs or outputs now returns a 422 error instead of a 500.
- Reduced CPU and memory requirements for dev and dev_free self-hosted deployments.
- Reject insecure default JWT secret at startup for improved security.
- Updated brand colors for neutral backgrounds and surfaces.
- Renamed prompt usage example label from “Use object in LangChain” to “Use Programmatically”.
- Fixed agent zip upload to correctly place cron schedules in the Schedule section.
- Inbox now sorts the All tab by recency and properly wraps long messages in preview.
- Fixed chat assistant tooltips rendering behind the chatbox.
- Added ABAC authorization middleware.
langsmith-0.13.26.tgzlangsmith-0.13.25
Helm chart update. Same application version as langsmith-0.13.24.Download the Helm chart:langsmith-0.13.25.tgzlangsmith-0.13.24
- Added rich markdown editor with toolbar and slash commands in Fleet.
- Added skill creation flow with page entry and navigation in Fleet.
- Annotation queue CSV exports now include per-annotator feedback scores and a reviewer notes column.
- Fixed dataset metadata filters not matching number fields across pages.
- Fixed datasets table only showing first page of results on tall screens.
- Alert like/notlike filters on error, inputs, and outputs now correctly match individual tokens instead of the full phrase.
- Fixed
list_runstool crashing when the LangSmith API returns an error. - Fixed stray artifact in system prompt for Gemini models.
- Accepting an organization invite now navigates to the newly joined organization.
- Improved Playground auto-scroll during streaming output.
- Added workspace scope display for personal API keys.
- Bumped Python to 3.13 and pinned OpenSSL to resolve security vulnerabilities.
- Blocked shell injection characters in build/install commands.
- Improved Polly assistant understanding of traces and runs.
- Fixed baseline experiment stats not showing on initial page load.
- Fixed duplicated x-axis date labels on the insights time series chart.
- Re-enabled ABAC for listing datasets.
- Added ABAC runs delete endpoint.
langsmith-0.13.24.tgzlangsmith-0.13.23
- Patched security vulnerabilities in smith-frontend.
- Patched security vulnerabilities in smith-polly.
- Fixed a code injection vulnerability.
- Restricted
--allow-runto only the deno binary in smith-ace. - Fixed XSS vulnerability by escaping URLs in the RichTextEditor.
- Fixed Playground functionality in self-hosted environments.
langsmith-0.13.23.tgzlangsmith-0.13.20
- Added JSON/YAML syntax highlighting to experiment comparison for better readability.
- Improved thread trace opening behavior in the frontend, removing the need for an expand button.
- Eliminated n+1 query issue in the backend for listing personal access tokens, improving performance.
- Fixed support for OpenAI compatible endpoints with smith-polly integration.
- Timed out bulk exports stuck in
CREATEDstatus to avoid indefinite processing. - Addressed issue where service identity access was blocked from creating repository endpoints.
- Recorded hub prompt commit in experiment session metadata for better session tracking.
- Improved authentication for /sessions shadow queries.
- Updated backend deployments with ABAC (Attribute-Based Access Control).
- Enhanced UI with projects and runs write permissions support.
- Added support for new models: GPT-5.4 and GPT-5.4 pro.
- Fixed large attachment image preview issue for better UI experience.
- Made GPT-5.4 the default OpenAI playground model, simplifying model selection.
- Increased maximum tags displayed in
RunTagscomponent for better visibility. - Added models and prompts columns to experiments table, enhancing data insights.
- Resolved agent builder runs rejection issue when limit settings were changed.
- Fixed float errors in /sessions go endpoint for improved data handling.
- Returned fetched value when Redis cache
SETfails, improving reliability. - Enabled AWS IAM role support for agent builder, Polly, and Insights features.
- Redesigned custom chart CRUD in the frontend, enhancing user satisfaction.
- Introduced prompt filtering in the experiments table for targeted data analysis.
- Updated inbox counts and thread fetching logic in Agent Builder for real-time information.
- Added a feature to group experiments by prompt for streamlined data management.
langsmith-0.13.20.tgzlangsmith-0.13.19
- Internal improvements and maintenance updates.
langsmith-0.13.19.tgzlangsmith-0.13.18
- Introduced a redesigned run details view in threads for improved user experience.
- Fixed an issue where popovers were covering other content in the UI.
- Added Microsoft Outlook Calendar Tools to the Agent Builder for integration.
- Addressed bugs related to agent chat popups and placeholders in the Agent Builder.
- Improved support for disabling feedback comment filtering in self-hosted instances.
- Enhanced performance with improved code splitting in the frontend.
- Added model support for GPT 5.3 instant and GPT-5.3-chat-latest.
- Introduced a new single_run filter type for more refined querying.
- Added ability to read insights reports and show insights categories over time.
- Added drag-to-resize functionality with persistence in custom iframe output renderer.
- Enhanced security with more robust user migration processes.
- Enabled tagging support for evaluators and enhanced their reuse functionality.
- Fixed issues with session expired warnings after logout.
- Enhanced UI components for better user interaction in the playground and feedback tagging.
- Improved metadata handling in datasets and fixed overflow issues.
- Introduced support for Microsoft Teams Tools in the Agent Builder.
- Implemented better handling for OAuth provider updates.
- Added new /orgs/current/info endpoint to the platform-backend for more robust organizational information retrieval.
- Introduced compatibility testing for session API with added safety checks for PostgreSQL and Redis connections.
- Added functionality to bind Slack agents dynamically, enhancing the integration experience.
langsmith-0.13.18.tgzlangsmith-0.13.17
- Fixed a bug in the executor deployment handling for new operator versions.
- Added a setting to filter out internal helper threads from the inbox.
- Improved the evaluator’s page by providing context to Polly’s feedback.
- Forced trace filtering for dataset code evaluators to enhance stability.
- Updated OAuth mode management, restricting changes during updates.
- Fixed an issue with experiment cell colors to enhance user clarity.
- Improved the usage configuration modal to utilize a new TTL endpoint for trace retention.
- Addressed a bug where workspace invites were not displaying correctly in the UI.
- Applied brand color adjustments in dark mode and various UI elements.
- Enhanced OAuth callback security by preventing potential reflected XSS vulnerabilities.
- Added a dynamic OAuth feature for user management.
- Fixed a bug preventing filter updates when certain conditions were met.
- Implemented rebranding updates for the auth screen.
- Added a feature to collapse sidebar automatically on small viewports.
- Fixed issues with variable handling in playground evaluate mode.
- Enhanced the Agent Builder with infinite scroll and improved inbox fetching.
- Added a new Outlook Trigger feature in the Agent Builder.
- Upgraded agent-builder to use websockets and new OpenAI model API (gpt-5.3-codex).
- Fixed auto-save on API key during onboarding process.
- Resolved issues causing errors in the playground due to empty placeholders.
- Updated frontend with a new logo for favicon.
- Fixed authorization bugs in cron deployment for Gmail/Outlook.
- Updated styling in various UI components, including studio button and index column behavior.
- Enhanced onboarding snippets for better integration with Langchain Python.
- Added support for custom separators in SCIM group names.
langsmith-0.13.17.tgzlangsmith-0.13.16
- Internal improvements and maintenance updates.
langsmith-0.13.16.tgzlangsmith-0.13.15
- Added rebranded primary colors to button under feature flag in the frontend UI.
- Replaced dataset autocomplete with tag input to improve user experience.
- Auto-hide and position Models column in the frontend based on data.
- Fixed revalidation conflict in the Smith frontend.
- Improved workspace model configurations to prevent text overflow with tooltips.
- Surfaced Models option in Group By popover under a feature flag.
- Supported loading ChatAnthropicVertex model configs in Smith-Polly.
- Added “No matching filters” message for empty search results in Filter Component Select V2.
- Enabled navigating automatically to insights with global scroll support.
- Resolved issues with playground and evaluators provider selector not filtering out disabled providers.
- Improved messaging mode user experience and styling.
- Implemented Raw Query Mode for Inline Filters.
- Allow
secret_key_refto beNoneinK8sEnvVarSourcefor backend improvements. - Fixed agent builder UI to wrap question text on narrow viewports and dismiss “Add API Key to Get Started” dialog.
- Updated UX to match evaluator button height with tool button pattern.
- Persisted selected model in local storage for a consistent UI experience.
- Auto-generated thread titles for improved thread management.
- Enhanced backend by gating secrets access with granular RBAC permissions.
- Implemented Outlook Email Tools in the Agent Builder.
- Improved keyboard shortcuts in the inbox feature of the Agent Builder UI.
langsmith-0.13.15.tgzlangsmith-0.13.14
- Fixed agent generation interruptions and handling, improving stability in the user experience.
- Fixed long feedback header text overflow when dragged to the last column.
- Added OAuth connections for built-in tools and providers on the tool page.
- Fixed crashes occurring on the run details page.
- Fixed onboarding dialog not fetching tools unnecessarily.
- Updated agent builder frontend to show real-time run count.
- Added private registry UI to the frontend.
- Enhanced support for SerializedConstructor model configs in playground and insights.
- Added Gemini 3.1 Pro model to playground and backend model lists.
- Fixed tool registry crash in the playground.
- Added support for Gmail authentication improvements, including refresh token capability.
- Added new API endpoints for running playground experiments using a new service.
- Improved UI for trace filters with version 2 UX using Filterbar.
- Enhanced syntax highlighting to match Figma design for standardization.
- Supported Gmail OAuth v2 with cron logic for higher reliability.
- Added new models column in the experiment view with updated filtering options.
- Supported multiple paths for query shadowing log improvements.
- Added UIs for managing and editing model API key names in the playground.
langsmith-0.13.14.tgzlangsmith-0.13.13
- Reverted the PostgreSQL version to v14.7 and the Redis version to v7. This fixes breaking changes introduced in langsmith-0.13.10.
- Fixed internal error details being leaked in 5xx responses to enhance security.
- Improved View UI by moving SaveViewButton from ViewDropdown and changing SaveForm to a modal for better usability.
- Added model select dropdown to the template creation flow for enhanced user experience.
- Added warning for duplicate URLs when creating MCP server to prevent configuration errors.
- Added user context to agents and sub-agents for better feature functionality.
- Added support for SerializedConstructor model configs in the playground for improved flexibility.
- Enhanced UI by showing categorical feedback in experiment view config and hiding the sort icon.
- Improved playground and experiment views by fixing cell alignment.
- Added image upload support to facilitate better asset management.
- Added onboarding dialog to general-purpose agent for improved user guidance.
- Added spinner to loading triggers skeleton for better loading indication.
langsmith-0.13.13.tgzlangsmith-0.13.12
- Improved button sizes and filter chip alignment in the InlineFilters UX.
- Added commit tags search and display to the Prompt Hub.
- Fixed issue with viewing experiments having objects for feedback scores.
- Enhanced tracing for the deploy_image task.
- Added a search bar for the new consolidated filter dropdown.
- Added environment variable for globally disabling personal access token creation.
- Added cost charts feature.
- Improved homepage styling and fixed related design issues.
- Fixed issues with rerendering in General Purpose API (GPA).
- Improved system to count PENDING, RETRY, and FAILED transactions in self-hosted offline usage reporting.
- Enhanced the agent builder to localize the current date to the user’s timezone.
- Added Bedrock inference profile dropdown to the playground.
- Improved error detection and messaging for server issues in agent-chat.
- Fixed styling issues including email count in invite modal and load state display in the agent editor.
- Implemented initial design for a tools page with feature flags.
- Added icon-only filter popover mode to the frontend filter UI.
- Added beacon endpoint for Self Hosted Agent Builder Runs Limiting.
- Enable new Granular Usage tab for reporting billable usage by workspace, project, user, and API key (enable with
DEFAULT_ORG_FEATURE_ENABLE_GRANULAR_USAGE_REPORTING=trueandGRANULAR_USAGE_TABLE_ENABLED=trueenvironment variables incommonEnv)
langsmith-0.13.12.tgzlangsmith-0.13.11
- Improved Agent Builder by using persisted simple model config.
- Fixed UI for Playground with better message block and tool button consistency.
- Added a search bar for the new consolidated filter dropdown.
- Fixed agent builder model selector for users without ‘workspaces:manage’ permission.
- Added file upload feature for General Purpose Agent.
- Added button to create General Purpose Agent.
- Enhanced the Playground by preserving baseline setting in URL on page reload.
- Improved Playground experiment table UI and alignment.
- Fixed bulk deletion of datasets to update the table correctly.
- Added new API: workspace-scoped tool registry API.
- Improved support for multifield runFields.
- Enhanced insights scheduler with backend changes.
- Added ability to navigate pages in Polly and an initial set of base evaluations.
- Added tracing enhancements to Agent Builder, including tool call tracing.
- Integrated changes to support custom model configs temporarily.
langsmith-0.13.11.tgzlangsmith-0.13.10
- Internal improvements and maintenance updates.
langsmith-0.13.10.tgzlangsmith-0.13.9
- Fixed sorting of workspaces alphabetically in the new switcher to improve user experience.
- Improved playground with new tool modal design and model config popup windows for enhanced usability.
- Fixed issue with creating tags being idempotent.
- Modified Agent Builder to cache MCP tools list, session ID, and OAuth tokens for better performance.
- Fixed updated error message for exhausted agent builder runs.
- Fixed routing configuration for the agent builder /allow-run API endpoint.
- Fixed spacing of home page tables for improved UI.
- Fixed issue with datasets repeatedly fetching if empty.
- Fixed edit access for API keys for non-admin users.
- Added cost and token columns in the experiment view for better data insights.
- Fixed an issue where the Slack trigger was dropping messages due to authentication errors.
- Fixed boolean feedback values handling in comparison table cells.
- Updated service key subject for API calls to /allow-run for accurate authentication.
- Improved agent builder to use persisted simple model config.
- Fixed error state handling for OAuth login failures.
- Enhanced agent builder by ensuring threads display errors on reconnect in agent chat.
- Fixed UI to ensure the footer menu closes on organization switch.
- Improved tagging authentication by using specific resource auth.
- Enhanced UI to prevent closing the pane from within the app selector dropdown.
- Fixed potential SQL injection risks in feedback and annotation queue listing.
- Added a 15-second timeout to the OAuth HTTP client for improved connection reliability.
langsmith-0.13.9.tgzlangsmith-0.13.7
- Internal improvements and maintenance updates.
langsmith-0.13.7.tgzlangsmith-0.13.6
- Fixed an issue with truncated large numbers affecting the user interface.
- Improved error string conversion from S3 to enhance error handling.
- Updated Filters UX to save DateTimeRange, improving user experience.
- Fixed UUID conversion to ensure consistent general agent identification.
- Fixed agent ID conversion to always use a string instead of UUID for stability.
- Enhanced the experiment comparison view by showing custom computed columns.
- Fixed chat preview for langchain-shaped output for better user experience.
- Improved caching mechanisms and authentication control planes’ RetryableHTTP.
- Fixed the issue of incorrect tenant usage when bootstrapping agents.
- Improved user interface by adjusting page padding to remove visual obstructions.
- Updated wording for trace-related inquiries to improve clarity.
- Enhanced large field uploads to S3 for single run POST/PATCH endpoints.
langsmith-0.13.6.tgzlangsmith-0.13.5
- Fixed regression on cloning prebuilt dashboards to enhance user experience.
- Updated Filters UX to match new mocks and improved the view dropdown.
- Fixed Agent Builder by converting
UUIDtostrbefore JSON encoding to prevent errors. - Added an updated insights sidebar for a more informative UI.
- Implemented bulk actions for tables to improve data management efficiency.
- Added a redesigned feedback banner for annotation queue in the frontend.
- Provided option to require feedback in the Smith frontend to enhance user communication.
- Limited Agent Builder template list to 2 rows for better usability.
- Fixed cursor jumping issue when editing agent builder skill name and description.
- Solved SSO+SCIM issue, always provisioning users to workspaces.
- Fixed Slack Auth disconnecting issue for workspace users.
- Enhanced missing tool experience in the Agent Builder by adjusting prompts.
- Prevented session fixation in self-hosted OAuth for enhanced security.
- Fixed the ‘view run’ regression in the annotation queue redesign in Smith frontend.
- Reduced gRPC streaming chunk size from 1MB to 64KB for improved performance.
- Added download zip button inside the Agent Builder Explorer.
- Enhanced tracing URLs allowed by adding them to the Datadog RUM config.
langsmith-0.13.5.tgzlangsmith-0.13.4
- Fixed the toggle functionality for all column sections when clicking the Columns header.
- Fixed failure to edit SSO settings
- Improved frontend performance by using BarSeries instead of AnimatedBarSeries for granular usage tabs.
- Added a Cmd + Enter hotkey to the new annotation queue for enhanced user interaction.
- Added an option in the playground UI to mitigate loading errors by defaulting to
use_responses_api=true. - Added support for custom Azure models.
- Updated the playground UI for improved user experience.
- Improved studio chat UX with auto-resize textarea and keyboard shortcuts.
- Fixed prompt saving issue with incorrect naming.
- Improved visibility and user experience by making search icons larger and tracking navigation of AQ, prompts, and deployments.
- Enhanced UI for the Agent Builder with the introduction of a new Diff UI with context lines.
- Enabled accurate Y-axis margins using d3 nice ticks for better frontend visualization.
- Allowed the ‘recently added to’ section to pop over in the ‘add to dataset’ UI.
- Added pagination to the dashboard select view for better usability.
- Added support to limit long-lived TTL options and expose them in the org TTL settings endpoint.
- Fixed a bug with streaming flicker using the responses API in the playground.
- Enhanced the annotation queue user experience with custom output rendering in new queues.
- Optimized dataset session comparison using example_ids filtering for better performance.
- Provided a download ZIP file button in the Agent Builder for user convenience.
- Added a streamlined loading state for the Agent Builder’s agent generator graph.
- Allowed updating and creating SSO settings specifically in self-hosted environments.
- Added support for displaying SCIM user’s
displayNameattribute.
langsmith-0.13.4.tgzlangsmith-0.13.3
- Improved streaming to accumulate streamed delta data without loss.
- Added Google Sheets/Docs tools integration.
- Enhanced UX for OAuth servers in settings.
- Made view trace links more obvious in the UI.
- Resolved multiple bugs and performance issues in Agent Builder, including writing self-hosted Agent Builder runs to a single project and collapsing subagent tool cards by default.
- Improved experiment UI, now showing experiment descriptions on the single experiment page and fixing a blank page issue when accessing experiments via public URL.
- Enhanced UX by solving bulk export flakiness and allowing updating of bulk export destination credentials.
- Improved inline filter UX with multiple bug fixes and added edit functionality.
- Added support for configurable run input/output preview paths on the backend.
- Introduced new workspace management permissions for managing members’ access.
- Optimized Agent Builder to reduce network calls.
- Added Action Menu to MyAgents Navlink in Agent Builder.
- Improved frontend performance, including solving slowness when loading authentication status.
- Enhanced Agent Builder with the API Key Needed Label/Button and collapsed action cluster features.
- Resolved playground tool modal overflow issue.
- Improved security by fixing frontend vulnerability in remix run router and clearing URLs on logout to prevent workspace misdirection.
- Updated API to use invoices for monthly burndown tracking and added support for V2 API.
- Improved frontend to handle malformed LLM outputs gracefully.
- Improved date/time selection UI by allowing bold “Last time” values for DateTimeRangePicker component.
langsmith-0.13.3.tgzlangsmith-0.13.2
- Fixed content-type validation for dataset uploads to improve data handling.
- Improved the new annotation queue page with a pretty JSON editor and message list editor.
- Enabled better handling of retriable ingest errors on QueueRunPayloads.
- Fixed Cron Trigger updates in the Agent Builder UI.
- Added support for Redis Cluster, enhancing scalability for self-hosted setups.
- Fixed OAuth account takeover vulnerability to enhance security.
- Improved rendering of raw output in the playground.
- Added new inline UX filters and a View Dropdown component for enhanced user interaction.
- Increased line-height on XS Text variant to prevent clipping issues in the frontend.
langsmith-0.13.2.tgzlangsmith-0.13.1
- Added webhook publishing feature in Agent Builder
- Added support for Anthropic models via Vertex AI in the playground
- Added “Use Responses API” toggle for OpenAI models in the playground
- Added Slack reauth banner when reconnection is needed
- Added edit file tool diffs for better visibility of changes
- Improved partition management for dropping old tables
- Fixed self-hosted issues with Agent Builder from QA
- Fixed mode switcher visibility when Agent Builder is disabled on self-hosted
- Fixed various Agent Builder navigation and UI issues
langsmith-0.13.1.tgzlangsmith-0.13.0
- Added support for Agent Builder in self-hosted deployments
- Added configurable trace TTL for long-lived traces
- Added ability to conditionally enable OAuth tools and triggers
- Added sample application creation during onboarding
- Fixed feedback pagination and auto-pagination bugs
- Fixed trace drawer skeleton not appearing immediately
langsmith-0.13.0.tgzlangsmith-0.12.37
- Added
agent_builder_enabledfeature flag for self-hosted deployments - Added ability to delete files in the memories folder
- Added secrets import flow in workspace settings
- Improved streaming tree performance
- Fixed scroll behavior in trace tree
- Fixed unicode truncation issues in trace display
- Fixed onboarding screen showing when runs exist
- Fixed various Agent Builder sidebar and navigation issues
- Fixed OAuth callback URL enforcement to require HTTPS
langsmith-0.12.37.tgzlangsmith-0.12.36
- Added support for custom MCP servers with OAuth in Agent Builder
- Added ability to view and edit memory files in Agent Builder
- Added support for attachments in the frontend
- Improved streaming tree performance in trace viewer
- Fixed scroll behavior in trace tree
- Fixed unicode truncation issues in trace display
- Fixed onboarding screen showing incorrectly when runs exist
- Increased maximum automation rules per workspace to 200
langsmith-0.12.36.tgzlangsmith-0.12.35
- Added per-bar highlighting for feedback charts in experiments
- Added Agent Builder activity feed
- Changed experiment chip to hover card for better usability
- Fixed agents appearing in random sidebar positions
- Fixed tool modal nesting issue in playground
- Fixed diff mode fallback on comparison page
- Fixed race condition in OAuth authentication requests
langsmith-0.12.35.tgzlangsmith-0.12.34
- Added Redis IAM authentication support for GCP and Azure
- Added self-serve audit logs in OCSF format
- Added hide column option to experiment outputs header
- Added message_user tool to the tool server
- Improved trace tree loading speed
- Allowed basic auth installations to disable invites via API
- Fixed tenant ID handling in navigation
- Fixed scrolling in Agent Builder templates view
- Fixed Gmail account connection limit tooltip
- Fixed user view preference persistence on page load
- Fixed tab wrapping in UI
- Made feedback charts visible by default
- Made SCIM group name matching case-insensitive
langsmith-0.12.34.tgzlangsmith-0.12.33
- Security fix: fixed Studio vulnerability to malicious
baseUrlparam by requiring user-defined allowed origins - Allow enabling invites alongside JIT provisioning for SSO (OAuth with Client Secret mode only)
- Added self-serve audit logs for administrative actions (Private Preview)
langsmith-0.12.33.tgzlangsmith-0.12.32
- Added IAM connection support for PostgreSQL (AWS only).
- Added GPT-5.2 model support to the playground.
- Added support for setting memory limits on executor pods.
langsmith-0.12.32.tgzlangsmith-0.12.31
- Improved error messages for basic authentication misconfiguration.
- Added organization operator role support.
- Fixed issues with streaming datasets endpoint.
langsmith-0.12.31.tgzlangsmith-0.12.30
- Fixed API Docs button not redirecting to the correct URL when using a sub path.
- Performance improvements and bug fixes.
langsmith-0.12.30.tgzlangsmith-0.12.29
- Added mTLS (mutual TLS) support for ClickHouse connections to enhance security for database communication.
langsmith-0.12.29.tgzlangsmith-0.12.28
- Added mTLS (mutual TLS) support for PostgreSQL connections to enhance security for database communication.
- Added mTLS support for ClickHouse clients.
- Fixed Agent Builder onboarding and side navigation visibility when disabled in self-hosted deployments.
langsmith-0.12.28.tgzlangsmith-0.12.27
- Added mTLS (mutual TLS) support for Redis connections to enhance security.
- Added support for empty trigger server configuration in self-hosted deployments.
- Improved incident banner styling and content.
langsmith-0.12.27.tgzlangsmith-0.12.25
- Enabled Agent Builder UI feature flag for self-hosted deployments.
- Added Redis Cluster support for improved scalability and high availability.
langsmith-0.12.25.tgzlangsmith-0.12.24
- Added dequeue timeouts to all SAQ (Simple Async Queue) queues to improve reliability.
- Performance improvements and bug fixes.
langsmith-0.12.24.tgzlangsmith-0.12.22
- Added Claude Opus 4.5 model support to the playground.
- Updated dataplane operator version.
- Added
LANGCHAIN_ENDPOINTenvironment variable when basePath is configured.
langsmith-0.12.22.tgzlangsmith-0.12.21
- Added explicit
revisionHistoryLimitconfiguration for operator deployment template.
langsmith-0.12.21.tgzlangsmith-0.12.20
- Added support for self-hosted customers to opt into the pairwise annotation queue feature.
- Updated operator to version 0.1.21 in LangSmith and data plane charts.
langsmith-0.12.20.tgzlangsmith-0.12.19
- Fixed playground environment configuration to use correct default settings.
langsmith-0.12.19.tgzAdditional Helm chart releases are available in the
langchain-ai/helm GitHub repository.Connect these docs to Claude, VSCode, and more via MCP for real-time answers.