OCP Release 3.4.0 Overview

OCP Release 3.4.0

.

What's New in OCP 3.4.0

The OCP 3.4.0 release advances OCP across agentic AI, natural language understanding, fraud prevention, analytics, and developer experience. Below are the highlights customers can take advantage of this release.

Agentic AI & Orchestration

• A/B Model Testing for Agents (Orchestrator 1.35.1) - Agents can now split traffic across multiple model backends in production. Each agent carries a list of model configurations with traffic weights that must add up to 100%, each with its own model, backend, and generation settings. When configurations change, the agent is redeployed automatically and rolled back safely if redeployment fails. Agents not using the feature continue to work exactly as before.

• Agent A/B Models Admin Panel (Environments Manager 1.27.1) - A new Agent Models page in the OCP Console® admin area lets administrators view, edit, and clear A/B model configurations for every agent in the environment, with built-in validation that traffic shares total 100% before saving.

• NLU Specialist Agent in OCP CoPilot® 1.9.0 - CoPilot now hands off NLU-related work to a dedicated specialist agent that manages models, runs training, evaluation, and intent recognition tests, then returns control without context-switching for the user.

• Onboarding Agent for New Users (CoPilot 1.9.0) - New users on the US1 landing page can now create their first Conversational AI application through a guided chat session inside CoPilot. The agent gathers requirements, builds and configures the application, and opens a live test panel in a single flow. Sessions are saved and can be resumed at any time.

• Conversation History & Session Resume (CoPilot 1.9.0) - CoPilot now shows a list of past chat sessions so users can scroll back through previous conversations and pick up where they left off.

• Agentic Step Control in Test Runs (Testing Studio+™ 1.19.0) - Testing Studio now recognizes and asserts seamlessly any agentic steps that might be part of the dialog.

Smarter NLU & Broader Language Support

• K-Fold Cross-Validation for NLU Models (OCP NLU 1.15.0) - Administrators can now run k-fold cross-validation directly from OCP Conversational Natural Language® for a more statistically robust assessment of model performance before deployment. Validation runs can be tracked in real time, with a full breakdown of per-intent accuracy, confusion patterns, and downloadable detailed reports. One validation run per group runs at a time to avoid conflicts with active training or evaluation jobs.

• Model Data Validation Before Training (OCP NLU 1.15.0) - Training datasets can be validated up front to catch data issues before a model is trained on a flawed dataset.

• Belgian French (fr-BE) Support (OCP NLU 1.15.0) - OCP Conversational Natural Language® now supports Belgian French as a new language option, extending coverage for French-speaking users in Belgium.

• Consistent xPack Version Across Servers (OCP NLU 1.14.1) - After deploying a new deep NLU® xPack, all running servers automatically refresh to the latest version without a manual restart, ensuring consistent recognition behavior from the moment a new xPack goes live.

• xPack Accuracy Improvements (OXP 1.18.1) - Updated xPacks for en-US Universal, en-US Universal+, es-US, pt-BR, and pt-PT deliver targeted improvements for DIALOGACT, CONVENTIONAL, Numeric, Hour, CardNumber, dates spoken digit-by-digit, Spanish Year extraction, and several other entities, reducing misrecognitions in production environments.

• Snapshot and UX Improvements (OCP NLU 1.15.0) - The language dropdown now displays readable names in alphabetical order, the snapshot list shows a visual indicator for the deployed snapshot, mixed-case entity references now train correctly, and annotation requests remain available during service restarts. Administrators can also configure CPU limits, reservations, replicas, and workflow maximums per xPack domain and language.

miniApps Enhancements

• Expanded en-GB Coverage (miniApps 2.42.0) - Alpha, Amount, DateRange, and Text miniApp types now support en-GB, giving UK-English deployments consistent language handling across more conversation flows.

• Stronger JavaScript Isolation (miniApps 2.42.0) - All JavaScript snippets in miniApps, including those that use cryptographic operations, now run in a fully isolated execution environment by default. Custom script logic no longer interferes with the broader platform, providing a more reliable and secure execution experience.

• Longer DTMF Input for Numeric miniApps (miniApps 2.42.0) - The maximum DTMF numeric input length has been increased from 20 to 21 digits, so callers can enter longer numeric sequences such as IBAN numbers without being cut off.

• Improved Numeric Recognition Accuracy (miniApps 2.42.0) - The number of speech recognition candidates evaluated during numeric input validation has been increased, reducing recognition failures when the top candidate is ambiguous.

Fraud Prevention & Data Protection

• Automatic Language Detection for Redaction (Redaction Service 0.3.0) - The redaction service now identifies the language of a transcription automatically before processing, enabling accurate redaction across multilingual call centers without manual configuration. The detected language is returned in the API response so downstream systems always know which language was recognized.

• Direct Redaction Configuration per Request (Redaction Service 0.3.0) - Authorized users can pass redaction settings directly with each request, bypassing the need for a pre-configured group profile and accelerating testing, integration, and ad-hoc processing workflows.

• Spoken Number Normalization (Redaction Service 0.3.0) - Spoken numbers in transcriptions (for example, "one two three four") are converted to digit form before sensitive-term scanning, improving detection of account numbers, card numbers, and similar PII spoken aloud.

• Cross-Group Voice Biometrics Operations (Voice Biometrics 1.17.0) - Enrollment, verification, and opt-out operations can now be triggered across multiple groups within the same organization. New versioned endpoints (/api/v2/enrolments, /api/v3/users/optout, /api/v3/users/info/search) support group-specific audio download and profile-based group resolution. All existing endpoints remain backward compatible.

• Voice Biometrics Resilience and Accessibility (Voice Biometrics 1.17.0 / 1.17.1) - Verification failures at high concurrency have been resolved, voiceprints with missing vector data are now skipped gracefully, the Voice Biometrics Console now supports full keyboard navigation (ARIA/WCAG compliant), and dependency updates address high- and medium-severity vulnerabilities. Version 1.17.1 also improves enrollment handling for large audio files.

• Liveness Detection Hardening (Liveness Detection 6.0.1) - Third-party library dependencies have been updated to remediate high-severity vulnerabilities and reduce exposure to potential security risks.

Reporting & Analytics

• Liveness Detection and Blocklisting Reports (Insights 1.25.0) - OCP Conversational Insights® now includes dedicated reports for Liveness Detection and Blocklisting in the Voice Biometrics section. Fraud-operations teams can track spoofed or blocklisted calls, view heatmap breakdowns by day or week, and filter by origin and region for a clear, at-a-glance view of anti-fraud activity.

• Timezone-Aware Export Scheduling (Insights 1.25.0) - Data exports can now be scheduled with a timezone offset and hourly precision. Exported data is always delivered in UTC, with the timezone used only for date-range selection.

• Agent Assist Events in Raw Data Exports (Insights 1.25.0) - Raw data exports now support Agent Assist event types, enabling agent-side interaction data to flow into analysis and reporting pipelines.

• Duplicate Key Search in Monitor (Insights 1.25.0) - Monitor dialog search now supports the same key with multiple values simultaneously, making it easier to find conversations where a particular call attribute takes any of several values.

• Custom Headers for Export Streams (Insights 1.25.0) - Export stream configurations now support custom static HTTP headers for destinations that require specific authentication or routing values.

• Advanced Calculated Metrics for Alerting (Metrics API 3.4.4) - Alerts can now be defined using mathematical expressions such as ratios and percentages (for example, no-match rate as a percentage of completed sessions). Percentage-based thresholds and notifications are handled natively.

• Composite Metrics in Aggregations (Metrics API 3.4.4) - The aggregations API now returns raw and calculated metrics in a single, validated response, simplifying dashboards and reports. Edge cases such as division by zero or missing values are handled gracefully.

• Expanded Analytics Data Coverage (Metrics API 3.4.4) - New data sources are available for dialog-level and root-level real-time metrics, Agent Assist sentiment analysis, miniApps intent distribution, and liveness/blocklisting detection signals.

Knowledge Base & Retrieval

• Background Project Exports (Pathfinder 1.9.0) - Project exports no longer block or time out on large knowledge bases. Exports run in the background with a real progress bar, can be re-downloaded until they expire, and now include the knowledge-base metadata-field schema. Users are notified when an export is ready.

• Filename Search for Knowledge-Base Documents (Pathfinder 1.9.0) - Documents can be located quickly by searching on their filename.

• Plain-Text Document Support (Pathfinder 1.9.2) - Knowledge-base document-retrieval projects now accept .txt files in addition to JSON. Retrieval results include the text content alongside a content-type indicator, eliminating manual conversion for plain-text sources.

• More Reliable Imports and Larger Archives (Pathfinder 1.9.0 / 1.9.2) - The Import button is disabled immediately after click to prevent accidental duplicate imports, error messages on invalid or corrupt documents are clearer and safer, and the import ZIP size limit has been raised to 200 MB.

Developer Experience & APIs

• Personal Access Tokens (Dialogs API 3.2.0, Pathfinder 1.9.0) - Personal Access Tokens are now supported in the Dialogs API and Pathfinder, giving developers a manageable, long-lived authentication option alongside bearer tokens.

• Search Dialogs by Duplicate Keys (Dialogs API 3.2.0) - The /v3/dialogs/search and /v3/dialogs/count endpoints accept a new virtual_agent.multi_kvps field for searches using key-value pairs where the same key appears multiple times, enabling more precise filtering of dialog data. The previous virtual_agent.kvps field is deprecated and will be removed in a future major release; the two fields are mutually exclusive.

• More Consistent Rate Limiting (Dialogs API 3.2.0) - Rate limiting now applies request throttling more reliably, improving service stability under load.

Identity & Access Management

• Invite Users by Email (IAM 3.11.0) - Organization admins can invite new users directly by email from OCP Console®. The invited user receives a time-limited link to set their own password and is logged in immediately after. Public self-registration can now be disabled, giving administrators tighter control over who joins the organization.

• Admin-Initiated Password Reset (IAM 3.11.0) - Admins can reset a user's password directly from the user list or details view, removing the need for a public "Forgot Password" link and reducing the credential-abuse attack surface.

• User Reactivation Without Support (IAM 3.11.0) - Organization admins can reactivate a deactivated user without raising a support request, including users that belong to multiple organizations. All reactivation actions are logged for audit.

• Token & Service Account Hygiene (IAM 3.11.0) - Users now receive a reminder 30 days before a personal access token expires. Service accounts inactive for 90 days are disabled automatically, with an advance alert at 60 days. System emails now display a clean, domain-appropriate sender name, with no Omilia branding on white-label environments.

• Refreshed OCP Console Interface - Phase 1 (IAM 3.11.0) - The OCP Console® user interface has been updated as part of a wider overhaul initiative, delivering a more consistent and refined experience.

Deployment & Platform Operations

• Extended Locale Support (Deployment Service 1.5.0) - Five new locales are now supported: Dutch (Netherlands), Dutch (Belgium), Hebrew (Israel), Danish (Denmark), and French (Belgium).

• Higher Default File Limit per Application (Deployment Service 1.5.0) - The default maximum number of files per application has been raised to 300, reducing the chance of hitting limits when deploying larger or more complex applications.

• NLU Configuration Validation at Deploy Time (Deployment Service 1.5.0) - The deployment service now validates NLU configuration within application flows at deployment time, catching misconfigurations early and preventing failed deployments.

Speech Recognition

• Hybrid Turn Detection Metrics (deepASR® 8.7.2) - The deepASR® hotfix release adds in-depth metrics for hybrid turn detection and improves performance in step audio redaction. This release also includes security updates.

Connectivity & Integration

• Codec-Matched Voice for IVR TTS Streaming (AGI Connector 4.22.1) - When IVR text-to-speech streaming is enabled, the connector automatically selects the correct codec-matched voice profile for synthesis. Streamed voice prompts play with audio that matches the call's codec, preventing silent or failed audio in streaming environments where no matching default profile was configured.

Platform Component Versions - OCP 3.4.0