# Openlayer Gateway, customizable dashboards, and Salesforce Agentforce GA

Published: 2026-05-15

[Openlayer Gateway](https://cdn.sanity.io/images/m9qfap23/production/41296f67a05805a887e2b0f370f8dab4cf9fe4b1-2400x1532.jpg?w=1400&fit=max&auto=format)

Openlayer Gateway is a new managed control layer for every LLM call your stack makes. The admin portal handles API-key management, per-key rate limits and freeze, and team-based organization — and every request that flows through is observable end-to-end in Openlayer.

## Customizable project dashboards {#customizable-project-dashboards}

Project home pages are now fully customizable. Pin widgets for the metrics you watch most closely, see project notifications inline, and jump between alerts from a new notification center in the sidebar.

[Customizable project dashboards](https://cdn.sanity.io/images/m9qfap23/production/d20d8a2bf0eda51a72f2b5ea85dd589bb44e20c1-2400x1532.jpg?w=1400&fit=max&auto=format)

## Salesforce Agentforce Integration is live {#salesforce-agentforce-integration-is-live}

Trace Agentforce conversations end-to-end. Pick your login domain (production, sandbox, or My Domain) and let the updated Builder schema capture every intermediate step.

[Salesforce Agentforce integration](https://cdn.sanity.io/images/m9qfap23/production/bce689a569949ab55f56f11e31b171e723592666-2400x1534.jpg?w=1400&fit=max&auto=format)

## Also new this quarter {#also-new-this-quarter}

- **On-prem Helm chart. **Deploy Openlayer to Kubernetes via a Helm chart on Docker Hub OCI, with config for storage, database, replicas, image pull secrets, and Node heap.
- **Built-in governance frameworks. **EU AI Act, TRAIGA, E23 MRM, and the Openlayer Governance Framework, available out of the box.
- **Audit trails. **Full workspace activity log with export.
- **PII redaction in traces. **Workspace-wide redaction with per-project entity-type configuration.
- **New guardrails. **Llama Prompt Guard for prompt injection, English and Portuguese toxicity, and Brazilian PII recognizers (CPF, CNPJ, CEP, phone).
- **New tests. **Bias, Contains PHI, and Unauthorized tool call.
- **Universal search. **Jump across projects, sessions, users, and tests from one search bar.

Full feature, improvement, and fix lists are below.

## Features

- Salesforce Agentforce integration (GA), with configurable login domain (production, sandbox, or My Domain) and updated Builder intermediate-step schema
- Dialogflow integration
- Azure Content Understanding tracer
- PydanticAI OTel GenAI attribute parsing
- Trace OpenAI Responses API from the TypeScript SDK
- openlayer-ts now accepts .csv datasets
- Trace decorator promote parameter — surfaces nested step outputs to the parent span for easier debugging
- Prompt-injection guardrail (Llama Prompt Guard)
- Toxicity guardrails for English and Portuguese, with opt-in chunking for long inputs and category filtering
- Brazilian PII recognizers in PIIGuardrail (CPF, CNPJ, CEP, phone), with checksum validation
- Bias test
- Unauthorized tool call test
- Contains-PHI test
- Workspace-level PII redaction on trace processing
- Per-project PII redaction settings — pick which entity types to redact
- Audit trails — full activity log with export
- Domain verification on the workspace settings page
- Workspace setting to restrict workspace creation to deployment admins
- Workspace setting to prevent users from a given domain from creating workspaces
- Built-in governance frameworks — EU AI Act, TRAIGA, E23 MRM, and Openlayer Governance Framework
- Customizable widgets on project home pages
- Project notifications surfaced on the home page
- Notification center in the sidebar
- Universal search across the workspace
- More keyboard shortcuts across the app
- Batch actions on rows, rules, and projects tables (including multi-select and delete)
- Filtering and sorting on sessions, users, data sources, projects, rules, and framework progress tables
- Hide columns in the users table
- Tag multiple tests at once
- Session-level aggregated metrics — view cost, latency, and token counts rolled up per session
- LLM cost configuration — view and edit per-model pricing used for cost calculations
- View, edit, and download custom-metric parameter bundles from the UI
- LLM evaluator settings now configurable for Ragas metrics
- Monitoring-data limit notifications
- Test-connection button for LLM configuration
- Openlayer Gateway — centralized AI gateway with admin portal, API key management (including rate limits, visibility, and freeze), and team-based organization
- BigQuery data sources without a timestamp column
- Browser-based openlayer login (device code flow)
- Delete records in monitoring mode via API
- Retrieve sessions linked to a test result via API
- dateUpdated filter on the List Tests API
- Helm chart for on-prem deployments, published to Docker Hub OCI
- Set workspace admins via environment variables on deployment
- License usage and version reported from on-prem deployments to Openlayer
- Universal logout improvements
- Automatic threshold mode for custom metrics

## Improvements

- Trace viewer performance and navigation — faster resizing, better breadcrumbs, attachment rendering, and clearer tools view
- LangChain callback handler latency reduced
- Project home page redesigned
- Redesigned settings pages
- Filter and table UI polish — multi-filter flows, virtualization for large result sets, loading states, and keyboard navigation
- Faster initial app load
- MFA passcode auto-authenticates on entry
- Deep-link redirect now works after auth and with SSO
- Universal search relevance improvements
- Built-in framework refinements
- Default test thresholds set to manual mode
- Default session and user IDs assigned automatically if none supplied
- BigQuery connection now supports Service Account secrets
- Custom CA certificates flow documented

## Fixes

- LLM-as-a-Judge reliability — rate-limit retries, temperature errors, and vague skip messages resolved
- Answer-relevance test no longer breaks on OpenAI embedding usage
- Judge score false positives when set to 1 fixed
- Google ADK, Azure OpenAI, and OCI tracers handle edge cases correctly
- log_context and log_output work when called from the same function
- Tool step type now displays actual tool name
- Trace modal stability — no longer reloads when new records stream in, and data no longer overflows
- Salesforce Agentforce sync and session context fixes
- Copilot Studio sessions: failing posts to API fixed
- Slack workspace integration deletion works
- Vertex AI configuration now correctly reports as connected
- Login and auth flow fixes — SSO deep links, MFA, and WorkOS "org not found" issue
- 403 error when removing users from a workspace fixed
- Customizing the dashboard works for all roles
- Table and filter reliability — column widths, null filters, infinite scroll, and data display
- Custom metrics no longer break when adding filters or with multiple tests
- Framework sync no longer creates duplicate rules
- openlayer update no longer fails with "invalid semantic version"
- LangChain callback handler correctly attributes sessions and user IDs