> ## Documentation Index > Fetch the complete documentation index at: https://explore.airia.com/llms.txt > Use this file to discover all available pages before exploring further. # Product Updates > Our changelog **New Features & Improvements**\ **Gemini Native Web Search Support** \ **What’s New**\ Airia now supports Gemini’s native web search capabilities through Google’s official Gemini API. Gemini models can now pull in live web information directly as part of their responses. \ **Why it Matters** * Helps Gemini provide more current and web-aware answers * Improves research and real-time information retrieval * Reduces reliance on custom web search workarounds * Provides a more seamless and reliable experience when working with live information **What it means for you**\ Teams using Gemini models in Airia can now access fresher, more up-to-date responses powered by built-in web search capabilities directly from Google’s official Gemini integration.\ \ **Workflow Output Property Picker** \ **What’s New**\ Building workflows is now easier with a new output property picker that lets users browse and insert data from previous workflow steps directly into node configurations.\ Instead of manually typing property paths, users can now select outputs from upstream nodes and automatically insert the correct reference syntax with a single click.\ The experience supports multiple workflow node types, including Manual, HTTP Request, and AI Model nodes. \ **Why it Matters** * Makes workflow building faster and more intuitive * Reduces errors caused by manually entering property references * Makes it easier to discover and reuse outputs from earlier workflow steps * Improves debugging and testing with live output previews and quick re-run options **What it means for you**\ Users can now create complex workflows more easily by visually selecting outputs from previous steps, helping reduce setup time and making workflow automation more accessible for both technical and non-technical users.\ \ **Bug Fixes & Improvements** * Improved reliability across data ingestion, file uploads, and third-party connectors including SharePoint, Box, and Google Drive * Resolved multiple authentication, OAuth, and credential synchronization issues across gateways and integrations * Fixed workflow, Agent Studio, and UI issues impacting execution flows, navigation, and configuration experiences * Improved Community permissions, audit logging, and overall browse experience * Addressed several AI Gateway, MCP Gateway, and guardrail stability and performance issues * Fixed multiple customer-reported issues impacting platform usability and reliability **New Features & Improvements** **Industry-Specific Graph RAG**\ **What’s New** * Build knowledge graphs from your data by extracting domain-specific entities (e.g., legal, medical, financial) * Supports customization of entity types based on your business needs * Available for both new and existing data sources (via feature flag) **Why it Matters** * Enables deeper, multi-step reasoning across your data—not just keyword or vector search * Unlocks more complex, context-aware answers for enterprise use cases \ **Vocabulary Grounding**\ **What’s New** * Define business-specific terms (acronyms, product names, internal jargon) with custom definitions * Applies automatically across semantic, hybrid, and Graph RAG search * No re-ingestion required—updates take effect on the next query * Supports bulk upload via CSV/JSON (up to 10 MB) **Why it Matters** * Ensures Airia understands your business language the way your team does * Improves retrieval accuracy and relevance for domain-specific queries \ \ **Microsoft Entra ID Authentication for Azure Foundry Models**\ **What’s New** * Added support for Microsoft Entra ID authentication for Azure AI Foundry model connections * Securely connect using service principal credentials (tenant ID, client ID, client secret) * Automatic token handling (no manual refresh needed) * API key connections continue to work as-is **Why it Matters** * Supports enterprise environments that require Entra ID * Improves security and access control without added complexity \ \ **Azure Foundry Model Selection in AI Gateway**\ **What’s New** * Select Azure AI Foundry and view all available models in your region * Automatically fetches deployed models and their endpoints * Selecting a model auto-generates routing rules (URL + version) * Admins only need to add credentials—no manual rule setup required **Why it Matters** * Speeds up gateway setup and reduces manual configuration * Minimizes errors and simplifies model onboarding \ \ **MCP Server Deploy/Connect Flow (v3)**\ **What’s New** * New two-step OAuth setup: tenant configuration + per-user connect/disconnect * Improved visibility into connection status with ability to disconnect **Why it Matters** * Simplifies credential management and improves control over connections \ \ **Model Native Search Support**\ **What’s New** * Enabled native search capabilities within supported models * Models can retrieve and ground responses using built-in search **Why it Matters** * Improves response quality with more relevant, up-to-date information * Reduces need for additional search configuration **Red Teaming Airia Key Supported** **What’s New** * You can now use library models for: * Attackers (Agentic – Campaign Creation) * Evaluations (Evaluation setup) * Target agents, including agents built with library models **Why it Matters** * Enables full library supported red-teaming across attackers, evaluations, and targets * Ensures consistent model behavior across campaign and evaluation flows * Reduces configuration mismatches * Improves reliability and flexibility in testing setups \ **Bug Fixes and Improvements** * Improved data ingestion, indexing, and search reliability across Data Sources and connectors * Resolved Agent Studio execution issues, including missing outputs, stuck agents, and tool failures * Improved MCP Gateway and OAuth reliability, including credential handling and connection flows * Fixed key customer-reported issues across Snowflake, SharePoint, on-prem, and Omni executions * Improved reporting accuracy across Insights, Feeds, budgets, and usage analytics * Addressed UI, mobile, Teams, and accessibility issues across the platform **Bug Fixes & Improvements** * Improved data ingestion and indexing reliability across Google Drive, SharePoint, Confluence, and OneDrive, fixing issues with syncing, stuck files, and incorrect file statuses. * Enhanced billing accuracy by resolving gaps in token consumption tracking and project-level budget attribution. * Increased MCP Gateway stability by fixing OAuth flows, credential handling issues, SSE disconnects, and strengthening role-based access controls. * Improved overall platform reliability by resolving agent execution failures, streaming issues, and key network-related incidents. * Enhanced user experience across Omni, Chat, and Agent Studio by fixing UI rendering issues, loading states, and interaction bugs. * Strengthened security and governance by correcting permission gaps, RBAC inconsistencies, and workflow reliability issues. * Improved Insights and analytics accuracy by fixing data loading issues, dashboards, and budget visibility gaps. **New Features & Improvements** \ **Compare Operation in SDK**\ **What’s New** * New Compare operation to evaluate two values in workflows * Supports common conditions like equals, greater than, contains, and more * Works across text, numbers, dates, and booleans * Handles edge cases like empty values and formatting automatically **Why it matters** * Makes it easier to add logic and decision-making into agents * Reduces the need for custom logic or workarounds * Helps ensure more consistent and reliable outcomes \ \ **Merge Step for Data Combination**\ **What’s New** * New Merge Step to combine data from two branches into one * Multiple merge options available (combine, match, pair, or cross-join data) * Flexible configuration depending on how you want data to align * Produces consistent, predictable outputs **Why it matters** * Simplifies combining and transforming data in workflows * Enables more advanced use cases without added complexity * Reduces errors when working with multiple data sources \ \ **Sort Node for Structured Data**\ **What’s New** * New Sort node to organize lists of data * Supports sorting by one or multiple fields * Works with both simple values and structured data * Includes options for sort direction and handling missing values **Why it matters** * Makes it easy to control how data is ordered in workflows * Improves readability and usability of outputs * Eliminates the need for manual sorting logic \ \ **Bug Fixes** * Improved reliability across **data ingestion, indexing, and knowledge graph workflows** * Fixed **integration and connector issues** (OAuth, SharePoint, MCP tools, email) * Resolved **Agent Studio and workflow execution bugs** * Enhanced **billing accuracy, permissions, and overall platform stability** **New Features & Improvements**\ **Data Sources in AI Step (Multi-Hop Retrieval)**\ Airia now allows you to attach one or more data sources directly to an AI step, making advanced retrieval simple, configurable, and accessible to any agent builder.\ **What’s New** * Ability to connect multiple knowledge sources directly within an AI (LLM) step * AI dynamically selects which sources to query, which retrieval methods to use, and how many times to search * Powered by multi-hop retrieval through the Airia Datasource MCP Server * Supports tool-based retrieval across multiple sources within a single agent execution **Why it Matters** * Transforms data retrieval into an intuitive, built-in capability * Improves response accuracy through iterative, multi-hop retrieval * Enables richer, more context-aware outputs across multiple data sources * Reduces manual orchestration within agent pipelines \ **Reranker for Hybrid Search**\ Airia now supports reranker models in hybrid search, improving the relevance and quality of returned results.\ **What’s New** * Ability to apply a reranker model on top of hybrid search results * Enhances ranking after initial keyword + semantic retrieval **Why it Matters** * Improves search accuracy and result relevance * Delivers higher-quality responses for downstream AI steps * Reduces noise in search outputs for complex queries \ **Bug Fixes & Reliability Improvements**\ This release includes key fixes across execution, integrations, and platform stability: * Improved agent execution reliability across workflows, loops, and pipelines * Resolved Agent Studio and UI issues (navigation, state, uploads, configuration) * Fixed authentication, RBAC, and SSO-related issues * Strengthened MCP/Gateway and tool integration reliability * Improved data ingestion, indexing, and search accuracy * Enhanced overall platform performance and stability * Addressed multiple high-priority and customer-reported production issues **New Features & Improvements**\ \ **Agent Discovery for Microsoft Copilot Studio**\ Airia now automatically discovers and inventories agents built in Microsoft Copilot Studio, giving security and governance teams visibility into agent proliferation across your Microsoft environment.\ **What's New** * Copilot Studio available as a new integration in the Integration Center * Automated background scanning discovers agents, their configurations, and connections after setup **Why it Matters** * Surface shadow agents and unmanaged Copilot Studio deployments that security tools miss today * Feed discovered agents into AI-SPM for risk assessment and posture management * Establish a governance baseline with a live, continuously updated agent inventory * Reduce manual audit effort with always-on discovery \ **Agent Discovery for Azure AI Foundry**\ Airia now automatically discovers and inventories agents deployed through Azure AI Foundry, extending security and governance visibility to custom-built agents running in your Azure environment. \ **What's New** * Azure AI Foundry available as a new integration in the Integration Center * Automated background scanning discovers agents and their configurations after setup **Why it Matters** * Close visibility gaps for custom agents built outside low-code platforms * Extend AI-SPM coverage to Azure-hosted agents alongside Copilot Studio and other discovered assets * Maintain a unified agent inventory across heterogeneous build platforms * Support governance workflows with continuous, automated discovery rather than point-in-time audits \ **Bug Fixes & Reliability Improvements**\ This release includes key fixes across stability, execution, connectors, and security: * Improved agent execution reliability across loops, approvals, and workflows * Resolved Agent Studio and UI issues, including uploads, navigation, and configuration errors * Fixed authentication and access control issues across SSO and permissions * Improved data ingestion, search, and connector reliability * Addressed MCP/Gateway issues and strengthened security and audit logging * Fixed multiple customer-reported production issues **New Features & Improvements**\ \ **Loop Over: Retry Logic for Failed Steps**\ We’ve enhanced the Loop Over block to give you more control over how failures are handled during execution.\ **What’s New:** * New retry option for handling failed loop steps * Ability to specify the number of retry attempts * Failed steps can retry instead of stopping or skipping * Works alongside existing loop controls **Why it Matters:** * Improves reliability for loop executions * Reduces need for manual reruns * Provides more flexibility in failure handling \ **Agent Studio: Real-Time Debug Overview**\ We’ve introduced a floating progress tracker in Agent Studio to provide real-time visibility into agent execution during testing.\ **What’s New:** * Floating Run progress panel appears during testing * Timeline view of steps with real-time status (running, succeeded, failed) * Clickable steps that navigate to the corresponding node * Responsive, scrollable panel with improved accessibility **Why it Matters:** * Makes it easier to track execution and understand agent behavior * Helps quickly identify and debug failures * Streamlines navigation between steps and the canvas \ **Tool Execution Performance Improvements**\ We’ve introduced performance optimizations for agent workflows involving multiple tools.\ **What’s New:** * Feature flag to enable enhanced tool execution performance * Optimizations across workflows with multiple tools * Supports native tools and MCP tools (legacy and Gateway) * Maintains sequential execution for canvas-configured steps **Why it Matters:** * Improves performance for tool-heavy workflows * Ensures consistent behavior across tool types * Preserves expected execution order while improving efficiency **Living disclosures**\ Create self updating point of use disclosures for AI usage governed by Airia. \ **What's new:** * Enables including AI disclosure information in a response header, body, and details callback URL for Airia agent executions tied to use cases in the monitor stage * Enables point of use AI disclosure in Airia agents deployed via a Slack interface. * Allows point of use AI disclosure for any agent proxied through the AI gateway and leveraging slack tool calls * Supports configuration of disclosed data elements based on use case **Why it matters** * Provide transparency by publishing governance intelligence at the point of use * Manage AI governance disclosures at scale with Audience-specific living disclosures of AI governance posture fed by live platform data * Replaces static governance artifacts with  continuously current governance evidence \ **Bug Fixes & Reliability Improvements**\ This release includes key fixes across stability, execution, connectors, and security: * Improved agent execution reliability across loops, approvals, and workflows * Resolved Agent Studio and UI issues, including uploads, navigation, and configuration errors * Fixed authentication and access control issues across SSO and permissions * Improved data ingestion, search, and connector reliability * Addressed MCP/Gateway issues and strengthened security and audit logging * Fixed multiple customer-reported production issues **New Features & Improvements**\ \ \ **Improved Search Relevance with Entity-Aware Retrieval**\ Search results are now enhanced through intelligent entity recognition and graph-based indexing.\ **What’s New:** * Automatic extraction of entities from ingested content * Entities stored and connected within a graph database * Search now leverages entity relationships to retrieve more relevant chunks **Why it Matters:** * Improves accuracy and relevance of search results * Enables better understanding of context across documents * Surfaces more meaningful and connected information during retrieval \ \ **Timezone-Aware Date & Time Context for Models**\ Models now automatically use the timezone defined in the user profile when Date & Time context is enabled.\ **What’s New:** * Model Date & Time context now aligns with the user’s profile timezone * Eliminates default reliance on UTC when context is toggled on * No need to manually specify timezone or country in agent prompts **Why it Matters:** * Ensures accurate, location-aware responses without additional prompt configuration * Reduces repetitive prompt engineering for timezone handling * Improves consistency between user settings and model behavior * Enhances reliability for time-sensitive workflows and automations \ \ **Custom Remote Servers: Secure Credential Management (UCS Integration)**\ Custom Remote Servers now leverage the User Credential Store (UCS) for secure handling of API keys and OAuth credentials.\ **What’s New:** * API keys for custom servers are now stored securely in UCS instead of plaintext storage * Support for both API Key and OAuth credentials using a unified Credential ID pattern * End-to-end support for Create, Update, and Delete server flows with UCS integration * OAuth Dynamic Client Registration (DCR) support for custom servers * Credentials are securely injected into request headers at runtime **Why it Matters:** * Eliminates plaintext storage of sensitive credentials, improving overall security posture * Provides a scalable foundation for migrating all credentials to UCS * Simplifies credential management across authentication types (API Key and OAuth) * Ensures secure, seamless execution of custom server integrations * Enhances reliability and consistency across server configuration workflows \ \ **Updated Navigation & Agent Favorites Enhancements**\ We’ve refreshed the navigation experience to better organize end-user features and improve discoverability across the platform.\ **What’s New:** * Updated navigation that groups end-user productivity features together * Enhanced visibility and access to thread history * Repositioned Settings for Admin users for better accessibility * Agent favorites moved into the Catalog as a new filter **Why it Matters:** * Improves ease of use with more intuitive grouping of key features * Makes it easier to find and revisit recent thread history * Streamlines navigation for both end users and admins * Enhances discoverability of favorited agents within the Catalog \ \ **Rename Agent on Import**\ You can now rename agents during import, making it easier to resolve naming conflicts without interrupting your workflow.\ **What’s New:** * Option to override the agent name during import * Prompted to rename when a duplicate name is detected **Why it Matters:** * Prevents import failures due to duplicate names * Makes it easier to move agents across projects without extra steps \ \ **Bug Fixes & Reliability Improvements**\ This release includes a broad set of fixes to improve platform stability, data handling, and overall usability: * Improved agent execution reliability across workflows, including loops, routing, and multi-step executions * Resolved issues in Agent Studio and builder, including file uploads, UI inconsistencies, and configuration errors * Fixed authentication and access control issues across SSO, service accounts, and role-based permissions * Improved data ingestion and retrieval accuracy, including entity extraction, chunking, and connector sync behavior * Addressed MCP Gateway and tool integration issues, including connection stability and execution reliability * Strengthened security and governance, including audit logging, guardrails, and credential handling * Improved connector stability across integrations like SharePoint, Google, and Slack * Enhanced UI, navigation, and accessibility across key areas of the platform * Resolved performance issues related to file processing, search, and ingestion * Fixed multiple customer-reported production issues **New Features & Improvements** **Redesigned File Viewer** We’ve redesigned the **View Files experience** to give you deeper visibility into your data and a more intuitive way to explore ingested content. **What’s New:** * View both **generated chunks** and the **original  file** * Download original files directly from the interface * Access metadata for: * Files * Chunks * SQL tables **Why it Matters:** * Gain full transparency into how your data is processed and used * Easily verify source content alongside generated chunks * Improve trust and debugging with richer metadata visibility * Faster exploration of content * Better user experience for both technical and non-technical users * Reduced friction when inspecting ingested data **Loop Over Datasource** Agents can now process multiple files from a data source in a single workflow, with flexible options for how each file is handled. **What’s New:** * Run agents across multiple files in one execution * Choose between processing file metadata only or full file content **Why it Matters:** * Simplifies handling of large datasets and document collections * Enables both lightweight workflows (faster, lower cost) and deep analysis (full document processing) * Gives teams more control over performance, cost, and level of analysis **Early Access:** This feature is currently in Early Access for select customers. Reach out to your Sales representative if you may benefit from it. **Generate Custom Attack Dataset (GROK-supported)** Easily create custom Red Teaming attack datasets aligned with OWASP guidelines **What’s New:** * Functionality to generate custom datasets for specific attack vectors **Why it Matters:** * Build targeted datasets faster and with confidence * Preview feature reduces mistakes and saves time **Bug Fixes & Reliability Improvements** * Improved platform stability and execution reliability across agent pipelines, including loops, parallel runs, and long-running executions. * Resolved Agent Studio issues impacting workflow building, publishing, and version consistency. * Fixed data handling gaps affecting datasource loops, credential usage, and message tracking. * Addressed connector and integration issues, improving reliability across external tools and APIs. * Improved export/import and configuration handling for agents and data sources. * Enhanced UI and usability across builder, debugging, and accessibility experiences. * Resolved multiple customer-reported errors and intermittent platform failures. **New Features & Improvements** \ **Improved File Move & Copy Handling for Data Source Sync**\ Data Source connectors now better handle files that are **moved or copied to new paths** in systems like OneDrive and SharePoint. \ **What’s New** * Improved detection of **files moved to new folders** * Copied files now sync correctly to the new location * Old or duplicate folder entries are no longer incorrectly shown in the UI **Why It Matters** * Keeps synced data **accurate and consistent** * Prevents stale or duplicate files after sync * Improves reliability when reorganizing files in connected systems \ \ **Agent Bias Score Evaluation**\ Governance teams can now evaluate agents for potential bias using a dataset and selected sensitive attributes.\ **What’s New** * Run bias evaluations using a **user-provided dataset**\\ * Compare agent responses across **sensitive attribute variants** (e.g., gender, age) * Uses **LLM-as-judge comparisons** to detect differences in tone, outcomes, or recommendations * Generates a **Bias Score (0–1)** with detailed evaluation results for governance review **Why It Matters** * Helps identify potential bias in agent responses * Supports responsible AI governance and oversight * Provides structured audit data for compliance and review **Cloudflare Gateway Integration for Shadow AI Detection** Security Posture Management (SPM) now integrates with Cloudflare Gateway to provide visibility and control over AI tool usage across the organization.\ **What’s New** * Monitor AI tool usage through DNS and HTTP traffic analysis * Detect unauthorized or unsanctioned AI applications * Gain real-time visibility into AI access across the network * Enforce blocking policies to control usage when needed **Bug Fixes & Reliability Improvements**\ This month we delivered fixes focused on stability, security, and core platform workflows in production, including: * **Improved platform stability and execution reliability** across agent pipelines, model execution, scheduling, and gateway services. * **Resolved authentication and access issues** affecting SSO, SCIM provisioning, OAuth integrations, and role-based permissions. * **Fixed Agent Studio workflow issues** including prompt builder segments not saving, memory step execution errors, loop execution failures, agent publishing validation, and scheduling behavior. * **Addressed tool and MCP execution issues** including authentication handling, timeout limits, and connector configuration. * **Improved connector and ingestion reliability** by resolving errors affecting file uploads, dataset processing, and integration behavior. * **Polished UI and usability issues** across chat, catalog, artifact rendering, navigation, and mobile responsiveness. * **Resolved multiple customer-reported errors and platform exceptions**, including intermittent “crab dragon” failures and API errors. **New Features & Improvements**\ \ \ **Fix with AI in Playground**\ Diagnose and resolve errors directly within the Playground using **Fix with AI**. \ **What’s New**\ Immediate visual error indicator when an issue occurs\ Clicking the indicator shows detailed error information * A **“Fix with AI”** action launches Build with AI automatically * The error is preloaded into a structured card and the agent immediately suggests a fix The button remains visible (but disabled) if Build with AI is not enabled. \ **Why It Matters** * Reduce debugging time * Keep builders focused without switching contexts * Accelerate iteration during development \ \ **Advanced Loop Execution Controls**\ The Loop node now supports configurable execution strategies, giving you greater control over performance and resource usage in complex workflows.\ **What’s New**\ Choose how iterations execute: * **Sequential** — One iteration at a time (ordered and conservative) * **Parallel** — Multiple iterations simultaneously with configurable concurrency * **Batched** — Controlled groups of iterations with adjustable batch size and timeout Additional enhancements include: * Clear progress indicators (Completed / Total + batch details) * Mode-aware error handling with partial result summaries * Memory monitoring with automatic fallback safeguards * Execution time logging for improved observability **Why It Matters** * Prevent rate limiting from excessive parallel AI calls * Reduce memory pressure in large workflows * Optimize performance for I/O-heavy operations * Improve reliability at enterprise scale \ **End User Agent Scheduling from Catalog**\ End users can now schedule agents to run automatically on a recurring basis directly from the Agent Catalog. \ **What’s New** * Create recurring schedules from the Catalog interface * Admins can view, pause, or delete schedules in Agent Studio * All schedule actions are captured in Audit Logs **Why It Matters** * Empower end users to automate repeatable workflows * Maintain full administrative visibility and control * Support governance and compliance through audit tracking * Reduce manual execution effort \ **Bug Fixes & Improvements**\ We delivered broad stability and usability enhancements across the platform, including: * Improved service reliability and access handling * Stronger autosave and draft consistency in Agent Studio * Stabilized MCP server deployments and tool configuration * Enhanced connector reliability and data ingestion stability * Improved search accuracy and execution consistency * UI refinements across Chat, navigation, and scheduling * Additional security hardening and performance optimizations * Reintroduced the Save button in Agent Studio to streamline the agent-building experience and prevent workflow interruptions. These updates strengthen platform reliability, improve governance, and enhance overall user experience. **New Features & Improvements** **Faster Agent Publishing**\ Publishing your first agent is now smoother and more intuitive. **What’s new** * A guided prompt appears after first publish to help you add an interface * Test immediately or complete setup later **Why it matters** * Get agents live and testable faster * Reduce setup friction for new deployments \ **Faster Navigation from Chat to Agent Studio**\ Admins can now jump directly from an agent’s chat experience to its configuration in Agent Studio. **Why it matters** * Review and adjust agent behavior faster * Reduce context switching * Improve operational efficiency \ **Agent Ownership Visibility**\ You can now assign an Agent Owner when deploying to the Catalog. **Why it matters** * Clearly identify who manages each agent * Improve accountability and internal coordination * Make it easier for users to find the right point of contact \ \ **Enhanced Protection Against Prompt Injection**\ We’ve strengthened safeguards to ensure agents stay within their defined roles and boundaries. **What’s new** * Responses validated against agent instructions before delivery * Misaligned outputs automatically blocked * Full audit visibility within guardrails **Why it matters** * Prevents unsafe or off-topic responses * Reduces compliance and data exposure risk * Strengthens trust in AI outputs \ **Agent Alignment Checks (Action-Level Protection)**\ Agents now validate tool calls against user intent before execution. **Why it matters** * Prevents unintended or risky actions * Reduces the impact of prompt injection attempts * Adds an additional governance layer without slowing performance \ **GitHub AI & Agent Discovery**\ Airia can now scan connected GitHub repositories to detect AI models, agents, and LLM usage across your codebase.\ **What this means for you** * Identify shadow AI and unauthorized model usage * Monitor agent proliferation across repositories * Enforce governance policies proactively No source code or secrets leave your environment — only sanitized metadata is analyzed.\ \ \ **Recurring Agent Scheduling**\ You can now schedule agents to run automatically on a recurring basis.\ **Why it matters** * Automate repeatable workflows * Reduce manual effort * Maintain full administrative oversight and audit tracking \ **LLM-Based Document Parser**\ A new LLM-powered parser improves structured data extraction from complex documents.\ **Why it matters** * Higher accuracy for legal, financial, and multi-page forms * Fewer missed fields and less manual correction * More flexible, context-aware document automation *To enable this feature, please contact your sales representative.s* **Reliability & Performance Improvements**\ We delivered broad stability and usability improvements across the platform, including: * More reliable MCP server deployments and authentication handling * Improved execution logging and scheduling accuracy * Enhanced document ingestion stability * Agent Studio UI refinements include 5 or more routes and select mode on canvas * Gateway monitoring and performance enhancements **New Features & Improvements** **SASE Integration** We've introduced SASE Integration in the Security Center, this feature generates a new endpoint that can be configured in your SASE, Zero Trust Network Access (ZTNA), or outbound network proxy to forward all LLM traffic to the Airia Gateway. All API traffic from managed, company-owned endpoints to model providers (such as OpenAI, Anthropic, etc.) can be redirected through the Airia Gateway — offering observability and runtime security such as guardrails and agent constraints. **What it means for you** * Route all LLM API traffic through the Airia Gateway automatically via your existing SASE/ZTNA infrastructure * Gain full observability into request and response details across every model provider interaction * Route traffic to approved LLMs and apply traffic routing rules for load balancing and failover * Enforce runtime security controls — including guardrails and agent constraints — without modifying individual endpoints * Maintain enterprise-grade security posture while enabling teams to use approved AI services seamlessly **Document Parsing Enhancement** We've added a new Excel row-by-row reader to our document parsing pipeline, improving how tabular data is ingested and processed. **What it means for you** * Parse Excel files row by row for more granular and accurate data extraction * Improve chunking quality for spreadsheet-heavy knowledge bases * Get more precise and context-aware responses when querying data sourced from Excel documents **Recurring Agent Scheduling from the Agent Catalog** End users can now schedule agents to run on a recurring basis directly from the Agent Catalog, with full administrative visibility and control in Agent Studio. **What this means for you** * Schedule agents to run automatically on a recurring cadence * Admins can view, pause, or delete user-created schedules from Agent Studio * Audit logs capture all schedule creation, updates, pauses, and deletions * This enhancement enables automation of routine workflows while preserving governance and oversight. **Reliability & Experience Improvements** This release focuses on making the platform faster, more stable, and easier to use across everyday workflows. * Agents and workflows now run more consistently, with improvements to scheduling, approvals, publishing, and execution handling. * We resolved several SSO and integration authentication issues to ensure users can sign in and connect tools without interruption. * Content ingestion and Enterprise Search are now more dependable across large files, media formats, and synced sources — improving answer accuracy. * Connections to external tools and MCP-based actions are more stable, reducing failed actions and reconnect loops. * We fixed navigation glitches, display issues, and interaction friction across Studio, Chat, and admin areas for a cleaner day-to-day experience. * Permissions and access behavior are now more consistent, improving safety and predictability for teams managing shared environments. **File-Based Attacks in Agentic AI Campaigns**\  Agentic AI campaigns now support attacks delivered through uploaded files, allowing scenarios to include prompts and materials packaged as documents or images. \ **What this means for you** * Run campaigns where the attack is provided as a file instead of only text input * Upload supported formats (PNG, JPG, PDF, DOCX, TXT) during setup * Validation ensures files meet format and content requirements before submission **Improved Agent Publish Experience**\ We’ve streamlined the first-time agent publish flow with a simple prompt to add an Agent Interface. \ **What’s new** * After publishing an agent for the first time, you’ll see a quick modal to add an interface * You can either set up an interface immediately to test your agent or skip and return later **Why it matters** * Faster testing after publish * Less friction for first-time agent setup **Direct Agent Studio Access from Agent Catalog Chat**\ Platform Admins, Read Only Admins, and Project Admins can now open the associated version of an agent directly in Agent Studio from the Agent Catalog chat. **What this means for you** * Quickly navigate from an agent’s chat experience to its Agent Studio configuration * Reduce context switching when reviewing or managing agent behavior * Maintain role-based access while improving operational efficiency This streamlines admin workflows by connecting agent discovery and management more seamlessly.\ \ **Bug Fixes & Reliability Improvements**\ This month we delivered broad fixes across stability, security, and core platform workflows, including: * **Improved platform stability and execution reliability** across production, on-prem, and regional environments * **Resolved authentication, SSO, and OAuth issues** affecting login, role-based access, and third-party integrations * **Fixed Agent Studio and workflow issues** including canvas interactions, imports/exports, scheduling, and execution failures * **Improved data ingestion and Enterprise Search reliability** across connectors, large files, audio/video formats, and sync jobs * **Addressed tool execution and API errors** across automation, MCP tools, and external integrations * **Strengthened security and governance controls**, including permission enforcement, credential handling, and auditability * **Polished UI and navigation issues** across chat, catalog, admin panels, and settings pages **January 2026 Release Notes**\ January focused on **platform resilience, agent building flexibility, enterprise integrations, and overall reliability**. We shipped several foundational improvements to how agents are built, tested, and executed—along with major enhancements to navigation, search, and governance.\ \ **Platform Enhancements** \ **Resilience Engine (Gateway Reliability)**\ We introduced the **Resilience Engine**, a new gateway capability designed to improve reliability for model and RAG requests. \ **What this means for you** * Automatic failover when providers, models, or endpoints experience issues * Circuit breakers to prevent repeated failures and reduce downtime * Load balancing across providers and models for improved stability * Clear visibility into routing decisions via logs and request traces \ **SQL-Based Search Service**\ A new SQL-based search service enables more structured and flexible querying across the platform.\ **What this means for you** * Use familiar SQL-style queries for precise searches * Apply consistent filtering and sorting across datasets * Support advanced and scalable enterprise search use cases \ \ **Agent Export & Import with Connected Data Sources**\ Agents can now be exported and imported as templates while safely referencing connected data sources.\ **What this means for you** * Reuse agent configurations without transferring data or credentials * Configure data sources in your own environment during import * Maintain consistent search behavior while customizing agent logic \ \ **Expanded Agent Studio Capabilities**\ Agent Studio received multiple enhancements to improve workflow design, testing, and control.\ **Highlights** * New action steps and flow control for more dynamic workflows * Step-level testing with manual inputs and output inspection * Model step retries and configurable error-handling behavior * Webhook Approval step to support human-in-the-loop workflows \ \ **Python Code Block Improvements**\ We shipped both **AI-assisted Python code generation** and a refreshed Python Code Block UI.\ **What this means for you** * Generate Python code directly within blocks using AI assistance * Edit, run, and review code without leaving the canvas * View outputs and errors in a unified, more intuitive interface \ \ **File Ingestion & Processing Enhancements**\ File ingestion was enhanced to provide richer context, improved resiliency, and broader format support.\ **What this means for you** * Improved document ingestion accuracy and performance * Support for audio and video file ingestion * Better visibility into ingestion behavior and easier recovery from failures * Enhanced file upload security and validation \ \ **Enterprise Integrations**\ We expanded support for enterprise data sources with new and improved connectors.\ **Highlights** * New on-premise SharePoint connector for self-hosted environments * Improved ingestion reliability across SharePoint, Drive, OneDrive, Outlook, and file uploads \ \ **Navigation Experience Update**\ We began migrating users from the legacy navigation to the new navigation experience.\ **What this means for you** * Automatic transition to the new navigation experience * Temporary ability to switch back to legacy navigation (\~1 week) * In-app notifications to guide users through the transition \ \ **Stability, Security & Reliability Improvements**\ Throughout January, we delivered broad improvements across the platform, including: * Increased stability across Agent Studio, execution, and workflows * Improved reliability for Enterprise Search, filtering, and pagination * Resolved authentication, SSO, permissions, and access-control issues * Fixed UI inconsistencies across canvas, chat, and tools * Addressed security findings and strengthened governance controls * General performance and reliability improvements across the platform **New Features & Improvements**\ **Resilience Engine**\ We’ve introduced the **Resilience Engine**, a new gateway capability that improves reliability for model and RAG requests through intelligent failover and load balancing. \ **What it means for you** * Automatically reroute requests when providers, models, or endpoints experience failures * Use circuit breakers to prevent repeated errors and reduce downtime * Load balance traffic across providers and models for improved stability * Get clear visibility into routing decisions through logs and request traces \ **Agent Canvas Preview on Community Agent Pages**\ You can now preview an agent’s canvas directly on Community Agent detail pages. \ **What it means for you** * Quickly understand an agent’s structure and workflow before using it * Preview how the agent is built without opening the editor * Make faster, more informed decisions when exploring community agents \ \ **AI-Assisted Python Block Code Generation**\ You can now generate Python block code with built-in AI assistance. \ **What it means for you** * Quickly generate Python code directly within blocks Reduce manual coding and setup time * Build and iterate on agent logic more efficiently \ \ **Migration to New Navigation Experience**\ Starting this week, we’ll begin moving users from the legacy navigation to the new navigation experience. \ **What it means for you**\ Users on legacy navigation will be automatically moved to the new navigation * Users can **switch back to legacy for \~1 week** before the final cutover * A popup notification will inform legacy users about the change * Users who switch back to legacy will also see a reminder modal **Bug Fixes & Stability Improvements** * Improved reliability across **data ingestion, connectors, and file uploads** * Fixed **Agent Studio, execution, and workflow** stability issues * Resolved **search, filtering, and UI** inconsistencies across the platform * Addressed **permissions, security, and governance** bugs * General **performance and reliability improvements** / **New Features & Improvements**\ **SQL-Based Search Service**\ We’ve introduced a new search service that implements SQL-based search APIs, enabling more structured and flexible querying across the platform. \ **What it means for you** * Use familiar SQL-style queries for precise searching * Apply consistent filtering and sorting across datasets * Support advanced and scalable search use cases **SharePoint On-Premise Connector**\ You can now connect to on-premise SharePoint instances as a data source using a dedicated connector. \ **What it means for you** * Connect securely to self-hosted SharePoint environments * Ingest and search documents from on-premise SharePoint * Expand data source support beyond cloud-only deployments **New Action & Flow Control Steps in Agent Studio**\ We’ve added new action steps and flow control steps in Agent Studio to give you more flexibility when designing agent workflows. \ **What it means for you** * Build richer workflows with additional action steps * Add conditional logic and flow control to manage execution paths * Create more dynamic and customizable agent behavior **Bug Fixes** * Improved Agent Studio and agent execution stability * Fixed data source ingestion and connector issues (SharePoint, Drive, OneDrive, Outlook, files) * Resolved Enterprise Search reliability and pagination issues * Addressed permissions, admin access, and security-related bugs * Improved overall platform performance and reliability **New Features & Improvements** **Agent Export & Import with Connected Data Sources**\ You can now export agents that reference connected data sources and import them as templates—without transferring any underlying data or sensitive credentials. \ **What it means for you** * Export agents that reference connected data sources without transferring data or credentials * Import agents as templates and configure data sources in your own environment * Reuse existing data sources or create new ones prefilled with recommended settings * Maintain consistent search behavior while customizing other configurations **Guardrail Assignment Exceptions**\ You can now define exceptions to guardrail assignments, allowing more flexibility in how guardrails are applied across agents and workflows. \ **What it means for you** * Apply guardrails to all agents and then exclude specific projects or agents. **Filename Search Tool**\ You can now search files by filename, making it easier to locate and retrieve relevant documents during agent execution. \ **What it means for you** * Find files using filename-based search queries * Quickly surface relevant documents without relying on content search alone * Improve agent accuracy when working with large or similarly structured file sets **Bug Fixes** * Improved overall platform stability and agent execution reliability * Fixed authentication, SSO, permissions, and access-control issues * Resolved data source ingestion, connector, and file-handling issues * Addressed UI inconsistencies and security findings across the platform **New Features & Improvements** **Python Code Block UI Refresh**\ The Python Code Block step has been updated to make it easier to edit, run, and review code directly within the canvas. \ **What it means for you** * Edit and run Python code without leaving the step * Quickly rename steps, undo or redo changes, and rerun code * View results in one place, including output and error details * Switch between result views without needing to rerun code **Model Step Retries & Error Handling**\ You can now control retry and error-handling behavior directly within Model steps. \ **What it means for you** * Automatically retry model calls when temporary issues occur * Configure retry limits and delays * Define how workflows behave when retries are exhausted **Step-Level Testing & Configuration Visibility**\ New step-level testing tools make it easier to validate and understand step behavior while building workflows. \ **What it means for you** * Test individual steps directly from the canvas * Provide manual inputs when testing steps * View step configuration and see updates immediately in draft mode * Review outputs from previous step runs * Settings remain locked for published workflows **Webhook Approval Step**\ The Webhook Approval step allows workflows to pause and wait for approval before continuing. \ **What it means for you** * Add approval gates to workflows * Route executions based on approval or denial * Support human-in-the-loop or system-based approvals * Maintain better control over sensitive workflows **Ingestion Artifacts & File Processing Improvements**\ File ingestion has been enhanced to store additional context and processing details for every file. **What it means for you** * Improve agent responses with richer file context * Support audio and video file ingestion * Resume file ingestion more easily if a failure occurs * Gain better visibility into ingestion behavior **Enhanced File Upload Security**\ We’ve improved file upload validation to further strengthen platform security. **What it means for you** * Better protection against unsafe or unsupported files * More reliable file handling without changes to your workflow **Bug Fixes** * Improved overall platform stability and reliability across workflows and model execution * Fixed issues with login, SSO, and authentication flows * Resolved UI inconsistencies across the canvas, chat, Smart Scan, and tools * Improved document ingestion accuracy and performance * Fixed workflow execution, scheduling, and data-passing issues * Addressed security findings from recent testing **New Features & Improvements** **Outlook Connector for On-Premises Environments** Airia now supports connecting to Outlook in on-premises environments, enabling integration with self-hosted systems. **What it means for you** * Connect Outlook data from on-premises environments * Support deployments with stricter infrastructure or data residency requirements **Editable Smart Scan Results** You can now edit Smart Scan results directly from the right-side drawer in Airia Chat, including updating column header names. **What it means for you** * Edit results without leaving the report view * Customize column headers to match your use case * Make quick refinements after scans run **Bug Fixes** * Improved overall platform stability and reduced intermittent errors across the canvas, workflow execution, and playground experiences * Enhanced authentication and provisioning reliability * Improved search performance and consistency * Increased connector and ingestion reliability across enterprise data sources **New Features and Enhancements** **AI Security Posture Management (AI-SPM)** Airia now provides AI Security Posture Management (AI-SPM) to help organizations identify AI-specific security risks by analyzing workflow automation traffic. With integrations like n8n, AI-SPM surfaces risky usage patterns and AI sprawl, giving teams visibility into emerging security and compliance gaps. **What it means for you** * Detect AI security risks traditional tools don’t catch * Gain visibility into AI usage across automation workflows * Improve security and governance as AI adoption scales **OAuth Token-Based AI Usage Discovery for Google Workspace & Office365** You can now integrate with Google Workspace or Office365 to securely discover AI usage across your organization using OAuth-based authentication. **What it means for you** * Track AI adoption across teams * Simplify discovery with enterprise-grade OAuth * Improve governance and oversight **Bring Your Own OAuth (BYOO) Connectors** Airia now supports BYOO connectors for OneNote, Outlook, and Confluence, allowing you to connect enterprise content using your own OAuth credentials. **What it means for you** * Maintain control over authentication and access policies * Connect trusted enterprise systems * Expand data sources available to agents **Shared Data Sources Across Multiple Projects** You can now share a single data source across multiple selected projects, without exposing it to all projects. Data sources shared across projects are treated as tenant-scoped with controlled access. **What it means for you** * Reuse data sources without duplication * Control exactly which projects can access shared data * Simplify data management across teams **WhatsApp Integration** Airia now supports WhatsApp integration, allowing you to deploy and interact with AI agents directly within WhatsApp. **What it means for you** * Reach users in a familiar messaging channel * Enable real-time conversational AI * Expand agent reach beyond traditional interfaces **Retry Failed Files** Users can now re-ingest selected failed files without reprocessing the entire data store, reducing processing time and resource usage. **What it means for you** * Quickly reprocess failed data sources * Reduce data gaps caused by transient issues * Improve reliability and continuity of agent knowledge **Edit agent-specific prompts right in the AI model settings** User can quickly define the agent-specific prompt (system instructions) and jump into the full Prompt Builder for advanced editing. **What it means for you** * Simplify prompt editing experience * A multi-line text area for short, agent-specific model instructions. **Parser configuration for inline document uploads ​** Agent builders can configure parser selection and image scanning options for inline file uploads in Chat. **What it means for you** * Load document content into the context window at runtime without requiring file ingestion, embedding creation, or knowledge storage *Note: This feature is currently available to selected customers who are granted early access. Please contact your sales representative if you wish to also receive early access. Capabilities and pricing for parsers are subject to change.* **Enhanced Guardrails**\ Airia now includes additional guard models—Azure Shield and Azure Content Safety—for prompt injection detection and moderation, with Arabic language support and inference in the MENA region. **What it means for you** * Stronger protection against prompt injection * Native Arabic moderation support * Low-latency inference in the MENA region **Custom Python-Based Guardrails**\ Airia now supports custom Python-based guardrails, enabling customers to build advanced protections that can integrate with external systems. **What it means for you** * Create tailored guardrail logic * Integrate with internal or third-party systems * Enforce advanced, context-specific controls **Bug Fixes Security & Reliability** * Strengthened protections to prevent unintended exposure of sensitive data. * Improved overall platform stability by resolving causes of intermittent system errors. Authentication & Integrations * Fixed OAuth and token refresh issues that could cause repeated authentication failures. * Improved reliability and clarity when setting up and managing connectors. Agent, Tools & Canvas Experience * Resolved issues causing errors or unexpected behavior in the agent canvas and playground. * Improved execution logging and export accuracy. Search & Data Quality * Improved Enterprise Search performance and reduced query timeouts. * Addressed data quality issues impacting search relevance and embeddings. **New Features and Improvements** \ **Agent Vibe Coding**\ Introducing **Agent Vibe Coding**, a faster and more intuitive way to build AI agents in Airia. Instead of setting up prompts, tools, and logic manually, you can now describe what you want — and Airia will generate the initial agent for you.\ **What’s new** * **Create agents from natural language** * Share your use case, and Airia auto-scaffolds the agent flow. * **Smart suggestions** * Recommends the right models, tools, and prompts based on your scenario. * **Easy refinement** * Update and improve the agent through simple conversational guidance. **What it means for you** * Build agents **much faster**, with far less manual setup * Enjoy a **more guided, beginner-friendly experience** * Achieve **more consistent, high-quality agent designs** across teams **Conversation-Based Memory**\ We’ve introduced **Conversation-Based Memory**, a new way for agents to remember information only within the specific conversation where it’s shared. This gives teams far more control over how context is stored and prevents information from carrying over where it shouldn’t.\ **What’s new** * A new “Conversation-Scoped Memory” option is available when configuring memories * Each conversation now maintains its own independent memory when this option is enabled * Clear visual labels show whether a memory is Global, User-Scoped, or Conversation-Scoped **What it means for you** * **Cleaner, more accurate agent responses** — agents won’t mix details from different conversations * **Better support for parallel workflows** — the same user can run multiple conversations without crossover * **More predictable behavior** for agents used across diverse use cases or business processes **Updated UI / Navigation** We've updated our UI throughout Airia, most notably how you navigate through Airia. * **Streamlined navigation** — key menus are now consolidated so you don’t have to hunt across different areas of the interface * **Faster workflows** — easily drill into nested options without switching tabs or losing your place * **Consistent experience** - every user now sees a unified layout, making training and support simpler **Additional Enhancements** * **Expanded Audit Logging**\ More audit events are now captured, giving teams better visibility into key actions and changes across the platform. * **Pipeline Batch Execution Endpoint**\ A new API endpoint allows you to run multiple pipeline executions in a single request, making large-scale or repeated operations easier to manage. * **Project Component Cleanup**\ You can now delete unused prompts, tools, and other components from projects to keep workspaces clean and organized. \ **Bug Fixes** * **More reliable document processing** for PPT, DOCX, PDF, and RAG chunks, ensuring complete and accurate ingestion. * **Improved connector stability**, including SharePoint, Microsoft Graph MCP, Google Drive, Outlook, and email attachments. * **Stronger agent execution reliability**, fixing empty outputs, stuck runs, webhook issues, and timeout inconsistencies. * **Smoother Agent Studio experience**, with fixes for tool/model visibility, layout issues, slow agent loading, and broken searches. * **Better access & interface behavior**, resolving login/invite problems and improving Slack, embedded chat, and export consistency. **New Features and Improvements** \ **Agent Variables in the Prompt Editor**\ We’ve moved Agent Variables into a new Variables tab inside the Prompt Editor for a smoother authoring experience. \ **What’s improved** * Drag and drop variables directly into prompt fields * Cursor-aware placement for precise insertion * A more streamlined prompt-building workflow **Support for Customer-Managed Azure AI Search Vector Databases**\ You can now connect your own Azure AI Search vector DB and use it across ingestion and RAG. **What’s improved** * Query SQL tables through the Search Service * Unified retrieval across files, connectors, and databases * No pipeline setup required **Bring Your Own OAuth for Data Sources**\ You can now authenticate select connectors using your organization’s own OAuth application, giving you full control over permissions and security policies. **What’s new** * Configure custom OAuth flows directly within the platform * Define your own access scopes for increased data governance **Supported connectors** * SharePoint * OneDrive * EWS Outlook This update empowers IT and security teams to align data ingestion with internal compliance requirements—without added complexity. \ **What’s improved** * Direct API access to semantic and hybrid search * Easy integration into your apps and backend workflows * Consistent performance across all supported vector databases **Bug Fixes** * Fixed user access issues including login loops, account-disabled errors, and SCIM syncing * Improved Agent Studio reliability for publishing, looping, and execution * Corrected gateway budget behavior to ensure traffic limits are enforced * Fixed model issues including Claude image handling and Amazon Nova Light output **New Features and Improvements** **Agentic AI Attack Campaign Creation** You can now create full agentic AI attack campaigns as part of our Red Teaming capabilities, with a guided flow for defining objectives, selecting orchestrators, and configuring models and converters. The feature supports only BYO/custom models to ensure safe, authorized security testing of your agents. Once configured, campaigns can be run immediately or scheduled, enabling structured, repeatable red teaming assessments. **New Agent Studio Playground + Compare Mode** We’ve added a new test mode directly inside Agent Studio, so you can edit and test your agent in one place without switching back and forth with Playground. The original Playground mode is still available for side-by-side comparison. Both options appear in the center-left controls of the Agent Studio canvas. **Import Modification Flow** You can now configure required model and tool credentials before importing an agent template, replacing the old placeholder process and reducing post-import cleanup. **Gateway Budget Limits** Manage AI spend by configuring daily, weekly, or monthly budget limits at a Gateway or user level. Set a budget in USD and restrict usage when the configured limits are reached. **Bug Fixes** * Fixed issue causing the export button on Data Ingestion Feed to fail. * Project Admin Role permission inconsistencies resolved. * Fixed a conditional branch and router execution bug. * Website scraper data source now auto-ingests regulation PDF files correctly. * Resolved an issue where lengthy tool responses were being cut off. * Decommissioned the v1 pipeline controller. * Corrected an issue where prompt messages would not load when editing prompts. * Disabled Conversation API access when “Limit Conversation History” is set to Yes. * Addressed an issue where Claude models were not receiving uploaded image files in chat/test. **New Features and Improvements** **Secure Agent Sharing & Creator Insights**\ You can now share agents with teammates or external users through a secure, one-time link that expires after use, and view usage metrics to see how often your agents are accessed and how they’re performing over time. \ **Slack Bot Deployment**\ Easily deploy any agent as a Slack Bot in minutes, enabling teams to interact with agents directly in Slack for faster collaboration and streamlined workflows. \ **Bug Fixes** * Fixed large file upload failures in ACC for smoother, more reliable data handling. * Resolved an issue with magic links to ensure shared agents open correctly. * Improved agent execution via API so input variables are now passed correctly. * Fixed email triggers that weren’t sending responses back to user mailboxes. * Corrected branding display so logos appear consistently across Airia Chat and Studio. **Airia Platform — October 2025 Release Notes**  **Build smarter. Work faster. Stay secure** This month’s release delivers meaningful updates across Agent Studio, Community, Connectors, and Security—designed to make Airia faster, more unified, and easier to manage.  From new interface shortcuts to improved ingestion reliability and expanded Microsoft 365 integration, September brings a mix of speed, stability, and simplicity.  **New Features & Improvements**  **Agent Studio & Execution**  * **Auto-Saving Drafts:** Drafts auto-create when you edit and auto-remove once published, keeping your workspace clean and current.  * **Flow Control Labels & Notes:** Easily follow complex pipelines with labeled steps and an inline note shortcut right on the canvas.  * **Agent Execution Feed Filtering:** Filter the execution feed by status to quickly find, review, and manage recent agent runs.  * **Agent Interface Shortcut:** A direct link to the Agent Interface now appears automatically when creating new agents, streamlining testing and iteration.  * **Improved Playground:** Chat history now persists across sessions for continuous testing, and non-streaming executions are faster and more reliable.  **Welcome Page & Navigation**  * **Personalized Welcome Page:** Quickly access recent projects, updates, and key workspace resources in one place.  * **Custom Branding Refresh:** Consistent logos and design now appear across all environments for a polished, cohesive experience.  * **Chat & Platform Language Sync:** Your language preferences now stay consistent across both Chat and Platform, creating a unified feel throughout the Airia experience.  **Community & Sharing**  * **Contributor Recognition:** New profile pages for contributors allow central access to a user’s contributed content, with enhanced details for repeat contributors.  * **Localization Support:** Translated labels and improved formatting enhance global usability.  * **Transport Improvements**: Enhanced support for transporting interfaces and custom models.  **Data, Ingestion & Connectors**  * **Expanded File Support:** You can now upload and process a wider range of file types including **YAML, HML, XML, and XSD** .  * **Website Connector Enhancements:** New Crawl4AI configuration options allow deeper and more flexible website ingestion for complex domains.  * **Ingestion Reliability:** Stability improvements for PDFs, images, and large websites reduce timeouts during crawling and indexing.  * **Amazon S3 Connector Security:** Introduced IAM ARN authentication for simpler, more secure cross-account access.  * **Microsoft Graph MCP Server Update:** Expanded support for Excel Tools adds more flexibility when working with Microsoft 365 data.  * **Permission-Aware RAG Enhancements:** SQL index views are now restricted for users without file access, and public APIs require verified end-user permissions.  **Credentials & OAuth**  * **User Credential Store:** Shared accounts and searchable credentials are now available across the platform for seamless team collaboration.  * **Token Estimation Backend:** A new backend service powers faster, more accurate token usage estimation for agents and API operations.  **Security & Identity**  * **Audit Log Filtering:** Filter audit logs by resource, severity, user, and project for faster, more targeted reviews.  * **Improved Authentication:** Login and SSO flows are more resilient, minimizing redirect loops and first-login errors.  * **Security Roles:** Security and Project Admin roles have been refined for better governance and visibility across teams.    **Fixes & Stability**  * **Admin Settings Access:** Fixed redirect issues for read-only admins.  * **Claude Code Stability:** Addressed intermittent errors when routed through the AI Gateway.  * **End-User Credential Tool:** Fixed a looping issue that prevented proper execution in some agents.  * **Prompt Segments Display:** Inherited prompt segments now display correctly.  * **Website Ingestion:** Enhanced reliability for large or complex websites to prevent timeouts.  * **Agent Execution Stability:** Fixed an issue where some agents would hang during subsequent runs.  * **Scheduled Sync Triggers:** Data-source syncs now run automatically as intended.  * **Excel File Compatibility:** Resolved issues preventing XLSX uploads in Loop Over Files / CSV steps.  * **Chat and Community Fixes:** Improved project-list load times, corrected broken links, and polished appearance for small devices and dark-mode.    **In Summary**  October’s release delivers:  * **Smarter building** with new execution filters, auto-drafts, and agent shortcuts.  * **Smoother collaboration** through improved Community sharing and global language consistency.  * **Broader integration** with Microsoft 365, Amazon S3, and new file-type support.  * **Stronger security** through IAM ARN authentication, permission-aware RAG, and refined audit controls.  * **Greater reliability** in ingestion, execution, and authentication across the platform.  Airia continues to evolve with every release—helping teams move faster, collaborate seamlessly, and scale securely.  **New Features & Improvements** **Welcome Page Updates**\ The Welcome Page has been refreshed with a cleaner design and easier navigation, helping users quickly access recent activity, updates, and key workspace resources. **Flow Control Step Labels**\ Flow Control steps now include clear, descriptive labels so it's easier to understand and manage complex automation flows. **JWT Forwarding Support (MCP)**\ Airia Deployed Agents MCP now supports JWT forwarding authentication, improving security and enabling seamless connectivity for advanced use cases like the upcoming Vibe Agent Builder. **Gateway Routing Rules**\ Easily configure rules to manage and secure how AI traffic flows through your gateway—ensuring it aligns with your organization's approved model providers, models, and endpoints. **Bug Fixes** **Loop Step Output** – Resolved an issue causing incorrect display of results in loop steps.\ **Human Approval & Embedded Agent Steps** – Background color now displays correctly as white for a consistent look and feel.\ **Prompt Version Ordering** – Fixed a problem where prompt versions appeared out of order. **New Features & Improvements** **Agent Execution Feed Filtering**\ You can now filter the execution feed by status to quickly find, review, and manage recent agent runs. **Agent Interface Shortcut**\ A quick-access link to the Agent Interface is now automatically added when creating new agents, streamlining testing and iteration. **Chat & Platform Language Sync**\ Your language preferences now stay consistent across both Chat and Platform—creating a smoother, more unified experience. **Microsoft Graph MCP Server Update**\ Expanded support for Excel Tools, giving teams more flexibility when connecting to and working with Microsoft 365 data. **Website Connector Enhancements**\ The Website Connector now supports additional Crawl4AI configuration options, making it easier to capture content from a wider range of sites and use cases. **Bug Fixes** **Admin Settings Access** – Fixed redirect issues for read-only admins accessing Settings.\ **Claude Code Stability** – Addressed intermittent 500 errors when routed through the AI Gateway.\ **End-User Credential Tool** – Resolved a problem that prevented this tool from running correctly in looping agents.\ **Prompt Segments Display** – Fixed an issue where inherited prompt segments appeared as "none."\ **Website Ingestion** – Improved reliability for large or complex websites to prevent timeouts during crawling. **New Features & Improvements** **Welcome Page**\ A new personalized Welcome Page helps users quickly access recent activity, updates, and key workspace resources. **Custom Attack Library**\ Upload and save custom datasets (CSV) for automated red-team testing and replay—extending flexibility in internal evaluation workflows. **Permission-Aware RAG Enhancements**\ Security and compliance have been strengthened: * SQL index views are now restricted when a user lacks file access. * Public search APIs no longer return data unless the end-user email is provided and permission checks are enabled. **Amazon S3 Connector Security**\ Introduced IAM ARN authentication for simpler and more secure cross-account access to S3 data. **Ingestion Enhancements**\ Broadened data coverage with new support for YAML and HML file types. **Token Estimation Backend**\ A new backend service powers faster, more reliable token usage estimation for agents and API operations. **Custom Branding Refresh**\ Consistent logos and visual identity are now applied across all environments, delivering a cohesive user experience. **Bug Fixes** * **Agent Execution Stability** – Fixed an issue where certain agents failed to complete on subsequent runs and remained in a continuous loading state. * **Scheduled Sync Trigger** – Resolved a problem preventing scheduled data-source syncs from running automatically. * **Loop Over Uploaded Files / CSV** – Fixed compatibility issue preventing Excel (XLSX) files from loading correctly. * **Claude Haiku Router (AWS Bedrock)** – Routing logic now correctly respects custom paths instead of defaulting. **New Features & Improvements** **Agent Execution Feed Filtering**\ You can now filter the execution feed by status to quickly find, review, and manage recent agent runs. **Agent Interface Shortcut**\ A quick-access link to the Agent Interface is now automatically added when creating new agents, streamlining testing and iteration. **Chat & Platform Language Sync**\ Your language preferences now stay consistent across both Chat and Platform—creating a smoother, more unified experience. **Microsoft Graph MCP Server Update**\ Expanded support for Excel Tools, giving teams more flexibility when connecting to and working with Microsoft 365 data. **Website Connector Enhancements**\ The Website Connector now supports additional Crawl4AI configuration options, making it easier to capture content from a wider range of sites and use cases. **Bug Fixes** **Admin Settings Access** – Fixed redirect issues for read-only admins accessing Settings.\ **Claude Code Stability** – Addressed intermittent 500 errors when routed through the AI Gateway.\ **End-User Credential Tool** – Resolved a problem that prevented this tool from running correctly in looping agents.\ **Prompt Segments Display** – Fixed an issue where inherited prompt segments appeared as "none."\ **Website Ingestion** – Improved reliability for large or complex websites to prevent timeouts during crawling. **Airia Platform — September 2025 Release Notes** **New Features & Improvements** **Airia Chrome and Firefox Extension** Bring your agents into the browser. Automate repetitive tasks, capture insights, and interact with Airia without ever leaving the tab you're working in. **Seamless Agent Import / Export** Easily transfer agents across projects. Whether you're collaborating with teammates or reusing your best builds, moving agents is now quick and hassle-free. **Community** Publish your agents to the Airia Community with a single click. Grow adoption, showcase your ideas, and discover what others are building. **Airia Cloud Connector (ACC)** The ACC now supports routing for MCP servers. Also, our Windows package now includes an installer that runs ACC as a background service, making setup faster and ongoing management effortless. **Support Tab Refresh** A redesigned experience with updated videos and Early Access content so you can find answers quickly and preview new features before they go live. **Simplified Agent Versioning** Manage agent versions with confidence. Work in draft mode safely, then publish updates live in one streamlined step. We are continuing to improve Agent Versioning further in the coming weeks. **Smarter Ingestion Processing** Data jobs are fairly distributed across data sources and tenants to ensure improved stability and usage of resources. **Domain-Specific Web Ingestion** Choose which web domains get ingested, giving you tighter control over data sources. **Enhanced SharePoint Navigation** Improved folder browsing with infinite scroll makes exploring deep hierarchies easier than ever. **Upgraded Custom Tools** Custom tools now automatically follow redirects while preserving authentication, reducing setup issues. **Security Admin Role** A new admin role designed specifically for managing security policies, reviewing guardrails, and accessing security-related logs. **Community Enhancements** Improved import/export flows, richer agent details, and an overall smoother experience connecting with peers and best practices. **Agent Constraints** Add conditional logic with IF-THEN rules to control when agents can access specific resources. **Dedicated Agent Details Tab** A new details tab centralizes agent settings, making it easier to review and adjust configurations. **Agent-to-Agent (A2A) Client Library** Available as a NuGet package, simplifying integrations and making it easier for developers to connect agents across systems. **Project Admin Role** Now generally available—gives organizations the flexibility to assign project-level admin permissions for stronger governance. **Docling PDF Support** On-premises deployments support PDF parsing powered by Docling, delivering more reliable results. **Hybrid search** Hybrid search is now supported with our proprietary vector database allowing customers to benefit from enhanced context-rich data retrieval. **Platform Enhancements** * More intuitive **tool management** with one-click connection testing and improved credential flows. * Expanded **MCP integrations** (Microsoft, Google, ServiceNow, Atlassian, Perplexity) with steadier authentication. * Refined **Agent Studio** with cleaner layouts and safer draft handling. * Expanded **observability metrics** to surface ingestion, model usage, and constraint activity. * Core infrastructure upgrades to improve scalability, resilience, and global performance. **Bug Fixes** * **Agent Studio**: Fixed issues with saving and restoring drafts, version mismatches, and structured outputs. * **Ingestion**: Resolved failures with PDFs, CSVs, and Excel uploads. SharePoint and ServiceNow ingestion is more reliable, and hybrid search indexing is steadier. * **Community**: Corrected glitches in filters, search, and contributor display; improved submission reliability. * **Chat**: Fixed disappearing history, broken citation links, and auto-scroll problems. Voice reconnections and playback are more stable. * **Console/Admin**: Fixed SSO/SCIM sync errors, tenant switcher bugs, and inconsistencies in audit logs. * **Mobile & Extensions**: Improved file upload handling, reduced chat decoding errors, and fixed voice playback issues. * **Models & Tools**: Corrected routing for external APIs, improved non-JSON handling, and stabilized token usage reporting. **In Summary** This release is packed with **new ways to build, share, and govern your AI agents**: * **Work anywhere** with the Chrome Extension * **Share easily** with Community publishing and Import/Export * **Govern confidently** with new Security and Project Admin roles * **Process smarter** with parallel ingestion and domain-scoped controls * **Collaborate better** with refreshed Support Tab and Community enhancements Airia continues to deliver clarity, control, and collaboration—helping you move faster, work smarter, and scale securely. **New Features & Improvements** **Chrome Extension** – Take your agents anywhere you browse. Automate tasks and bring Airia into your everyday workflows—without ever leaving your tab. **Import/Export Agents** – Move agents between projects with ease. Sharing and collaborating on your best work has never been simpler. **Submit to Community** – Share your agents with the wider Airia community. Help others discover your ideas while growing our creators program and driving adoption. **Airia Cloud Connector (Windows Package Update)** – Now ships with an installer that lets ACC run as a service, making setup and management smoother. **Support Tab Update** – Refreshed video experience plus Early Access content, so you can get answers faster and stay ahead of new features. **Bug Fixes** **Context Memory** – Fixed an issue where playground and published agents weren't keeping context in MENA and demo environments. **File Uploads** – Output from looping uploads now matches single document uploads for consistent results. **SSO Settings** – SSO settings no longer appear blank—making admin setup more reliable. **Redirects** – Fixed redirect from Platform to Chat for users with the end-user role. **Agent Draft Loading** – Agents now open with the latest draft by default, so you can pick up right where you left off. **New Features & Improvements** * **Agent Version Control** – A simpler *Draft → Publish* flow makes it easier to update agents. Draft safely, then publish live with a single click. * **Smarter Data Processing** – Large ingestion jobs no longer block others thanks to parallel processing across tenants. * **Domain-Scoped Web Ingestion** – Control exactly what web content gets ingested by limiting to a specific domain. * **Better SharePoint Experience** – Infinite scroll improves navigation through deep folder hierarchies. * **CSV & Excel for Text-to-SQL** – Cleaner indexing makes data more reliable for advanced SQL queries. * **Custom Tools Upgrade** – Tools now follow redirects automatically while keeping authentication intact, reducing setup headaches. **Bug Fixes** * **PDFs in MENA** – Fixed firewall errors and restored PDF previews in chat. * **File Uploads** – Resolved an issue that blocked multi-file uploads in chat for EU tenants. * **Credentials Management** – Fixed errors when adding or editing credentials and improved the Run Statement button behavior. * **Agent Tools & Routing** – Issues with renaming agents, orphaned steps, and route deletions are now resolved. **New Features & Improvements** * **Support Tab** – Quickly find self-service resources and help articles right inside the platform. * **Security Admin Role** – New role that allows dedicated admins to manage security settings and view security-related data. * **Airia Community** – Easily import/export agents and connect with the wider community for ideas and best practices. * **Agent Constraints** – Add IF-THEN rules that check conditions before agents access resources, giving you more control. * **Agent Details Tab** – Cleaner agent settings with a new dedicated tab for details. * **A2A Client Library** – Available as a NuGet package to simplify developer integrations. * **Pre-Release Agent Versions** – In-product announcements highlight upcoming agent versions so you can test early. **Bug Fixes** * **Data Processing Tab** – Fixed an issue where no results were displayed for on-prem deployments. * **People Page** – Sorting by last activity now works as expected. * **Slider Tooltips** – Corrected display issues where tooltips extended outside the interface. * **Options Panel** – Fixed an issue where the panel didn't close after saving an agent in Agent Studio. * **Agent Execution Events** – Resolved an issue preventing events from being sent to syslog. **New Features & Improvements** * **Project Admin Role – Global Release** – New role enables project-level admin permissions, giving organizations more flexibility and governance. * **Support for Docling** – On-premise deployments now benefit from stronger, more reliable PDF parsing. **Bug Fixes** * **Agent Router** – Fixed an issue where the Bedrock Model could not be used for routing agents. * **Admin Access** – Resolved a login problem preventing some admin users from accessing the platform. * **Conversation History** – Previous threads are now consistently visible within customer tenants. * **File Downloads** – Corrected errors where certain files returned "Blob Not Found" when downloaded via ACC. * **Custom Tools** – Fixed an issue that blocked users from importing their own custom tools. * **Database Connections** – Improved Postgres connection pooling to ensure more stable and reliable integrations. **Airia Platform — August 2025 Release Notes** **New Features & Improvements** **Red Teaming** Run targeted red-team tests against your agents to proactively uncover and mitigate vulnerabilities. **Security Center + Navigation Update** A dedicated **Security** tab brings guardrails, feeds, and alerts into one place for better visibility and faster action. **Enterprise Search — Internal Release** Unified search across connected sources (Slack, GitHub, Jira, and more) is now available for internal users, bringing "one place to find it" closer to GA. **Outlook Connector: Shared Mailboxes** Ingest from shared mailboxes and choose specific folders so teams bring in only what matters. **Seeded Example Agents** New customer environments start with a sample agent to help teams get value on day one. **Docs Telemetry** Documentation now includes usage telemetry so we can continuously improve high-impact areas. **Agent Studio UI Enhancements** Cleaner navigation and details views make building and managing agents faster and more intuitive. **MongoDB via ACC** Database tool routing through the Airia Cloud Connector now supports MongoDB. **Permission-Aware RAG Enhancements** Stronger access controls ensure restricted content is never surfaced to users who shouldn't see it. **Project Admin Role** Granular, project-level permissions to delegate governance without compromising control. **JSON Formatter Step** Agents can present results with visual and raw JSON options—better readability, better control. **"Create Your Own Agent" Redesign** A refreshed experience that shortens the path from idea to working agent. **User Authentication for Custom Tools** Build tools that use each user's own credentials (Bearer, Basic, API-Key) for secure, auditable access. **Network Groups + DB Tool Support** More flexible grouping logic and broader database integrations. **Customizable Departments** Apply department structures across requests, agents, feedback, and interfaces to streamline routing and reporting. **Dark Mode** Lower eye strain and a sleek look that's now consistent across the platform. **Platform Enhancements** * **Model Catalog Updates:** Added GPT-5 and Opus 4.1; simplified dynamic listings and cleaned up legacy/deprecated options. * **Policy & Governance:** Admins can enable/disable key features (Create-Your-Own-Agent, Knowledge, Smart Scan, Spaces) at tenant, group, or user levels. * **Observability:** Expanded metrics for ingestion, model usage, and constraint activity. * **Reliability & Scale:** Core infrastructure upgrades (including AKS 1.31 across environments) improve performance and resilience. * **MCP & Tooling:** Smoother token validation and setup flows across major connectors and MCP servers; clearer SQL tool messaging and logs. **Important Bug Fixes** **Agent Reliability** * Fixed looping behaviors (incorrect token counts, extra messages, and progress display) and misapplied max-cycle settings—agent runs are now accurate and transparent. * Resolved issues that intermittently blocked agent executions in certain environments. **Credentials & Security** * Credential types are correctly tenant-scoped, eliminating cross-tenant leakage. * Registration flow hardened to prevent bypass scenarios. * HMAC secret rotation stabilized to avoid unexpected tool call failures. **Enterprise Search & Ingestion** * Search results now consistently display correct file names, IDs, and metadata; "get file/preview/chunk" operations are more reliable. * File uploads and syncs (PDF, CSV/Excel, SharePoint, ServiceNow, OneNote, Confluence) are significantly more stable. **Project Admin Experience** * Fixed errors that blocked viewing usage, deploying agents to chat, or interacting with guardrails; reduced stray unauthorized pop-ups. **Chat, Studio & Extensions** * Restored missing threads and fixed PDF preview in chat; removed misleading browser warnings. * Improved save/publish responsiveness and load indicators in Studio. * Browser extension now isolates styles, improves input behavior, and offers a cleaner entry point. **In Summary** This release sharpens three things customers care about most: * **Security & Governance:** Red Teaming, the new Security Center, and stronger permission boundaries. * **Findability:** Enterprise Search plus a wave of reliability and UX improvements. * **Builder Velocity:** A faster agent-building experience (JSON formatter, structured output, seeded agents) and smoother admin controls. Airia helps teams move faster, work smarter, and scale securely—across every environment. **New Features & Improvements** * **Outlook Connector Enhancements** – Outlook now supports Shared Mailbox ingestion, with a new Folder Selector to easily browse and manage mailbox content. * **Security Tab: Central Command for AI Security** – A new dedicated Security tab consolidates all security-related features into one streamlined hub: * **Centralized Security Hub** – All security features now live under a single tab for easier access and management. * **Guardrails & Security Feeds** – Moved from Agent Studio into the Security tab for better visibility and control. * **Red Teaming** – Simulate adversarial attacks to identify vulnerabilities in agents before real threats can exploit them. * *Coming soon*: **Role-Based Access** – Security tab visibility and permissions aligned with analyst and CISO personas. * **Seeded Example Agent** – A ready-to-use example agent is now automatically included in new tenants to help teams get started faster. * **Agent Settings Alignment** – The Agent Settings modal in Agent Studio has been updated for a more consistent platform-wide experience. * **New Agent Details Page** – Enhanced detail view makes it easier to explore and understand agents within the Community catalog. **Bug Fixes** * **SCIM Token Expiration** – Tokens now correctly expire after 180 days (previously 30 days). * **Save & Publish** – Resolved issue causing repeated reloads after saving and publishing agents. * **Create Agent Performance** – Improved loading speed and added a clear loading indicator. * **Webhooks** – Fixed issue causing HMAC secrets to regenerate unnecessarily. * **Tenant-Wide Models** – Tenant-level models now consistently appear in the connected models screen. * **Agent Executions** – Corrected issues with API keys being displayed incorrectly in execution feeds and scheduled runs. * **Datasource Routing** – Ensured datasource chunks are properly passed through to LLMs. * **Insights Graphs** – Fixed incorrect data being displayed in analytics dashboards. * **Knowledge Toggle in Chat** – Users can now reliably disable Knowledge during conversations. **Week of August 22nd Release Notes** **New Features & Improvements** * **Opus 4.1 in Library** – Latest Opus model now available for use. * **TogetherAI Older Model Deprecations** – Deprecated outdated TogetherAI models to streamline options. * **Improved Navigation Layout in Agent Studio** – Updated layout for smoother workflows and enhanced usability. * **New Professional Plan Rollout** – Can now be activated directly after registering online. * **Permission-Aware RAG Enhancement** – Platform users cannot view chunks of files if they don't have permission to the original file and the permission check option is enabled for the data source. * **SCIM Provisioning Support for Large Deployments** – Expanded provisioning to handle larger-scale environments. * **Audit Logging** – Added audit log capabilities for better visibility and compliance. * **Direct Import Agents** – Agents can now be imported directly into the platform. * **Route via ACC for MongoDB DB Tool** – ACC routing now supported for MongoDB database tool. * **Access Token Validation** – Added validation for Notion, GitHub, Google Suite, and Hubspot integrations. * **Updated UX for Word Extension Management** – Refreshed user experience for managing Word extensions. **Bug Fixes** * **Save & Publish Actions Triggered Repeated Loading Loop (3x)** – Fixed issue where actions caused multiple unnecessary reloads. * **Slow Loading Time on "Create Agent"** – Improved loading speed and added a clear loading indicator. * **Webhooks Regenerating HMAC Secrets Continuously** – Fixed issue causing webhook secrets to refresh unnecessarily. * **Tenant-Wide Models Not Appearing in Connected Models Screen** – Ensured tenant-level models now display properly. * **Extensions Not Showing When Office Add-in FF Enabled** – Resolved issue preventing extensions from appearing. * **DB Tools Connection String Info in Logs** – Removed sensitive connection string information from database tool logs. * **New / Edit Custom Model Modal Bleeding Off Screen** – Corrected UI issue with long prompts extending past modal bounds. * **Multiple UPF Data Sources in Some Tenants** – Fixed duplication of user personal file data sources. * **Group User List Doesn't Scroll** – Corrected scrolling behavior when many users are displayed in Groups. **Week of August 15th Release Notes** **New Features & Improvements** * **Jamba Dynamic Models Added to Library** – Expanded model options with flexible, on-demand Jamba dynamic models. * **Brave Search MCP Updates** – Includes image search capabilities, additional tools, and fixes for rate-limiting issues. * **Google Drive Service Account Connector** – Now available for all customers, enabling secure service-account-based integrations. * **Configurable Embedding Model** – Choose between OpenAI and Cohere embedding models, with customizable vector dimensions to fit your use case. * **New Navigation UI Component** – A refreshed navigation element for faster, more intuitive platform navigation. * **Separate MCP & Tools Component Views** – Clearer separation of MCPs and Tools for streamlined access and management. * **Improved Navigation Layout in Agent Studio** – Updated layout for smoother workflows and enhanced usability. * **Database Tools Testing** – Improved stability and expanded capabilities for database-related tools. * **Agent Video Support for Vimeo** – Agents can now display and use videos hosted on Vimeo. * **New UX for Editing & Syncing Data Sources** – A cleaner, more flexible interface for managing data sources and sync preferences. * **Deprecated Model:** `Jamba 1.6 Models` **Bug Fixes** * **OP / Dedicated SaaS ACC Download Failing** – Resolved an issue preventing successful downloads. * **Insights Graphs Missing Current-Day Data** – Fixed graphs to include the most recent data and ensured graphs refresh with the latest information. * **Interface Tags Not Rendering** – Corrected rendering issues for interface tags. * **Folder Upload Support** – Added the ability to upload entire folders in File Upload. * **User Personal Files Created as Non–User-Specific** – Fixed file ownership assignment to ensure correct user attribution. * **OneDrive On-Prem Files** – Resolved an issue where files added via OneDrive remained stuck in "Pending" status. * **MCP Connection Issues** – Fixed ACC/MCP connection problems and baremetal environment server info calls. * **UI/UX Fixes** – Resolved tool action purpose display, secure tag misplacement on native tools, custom credential dropdown highlighting, and chat credential type display. * **Custom MCP Auto-Detection Bug** – Fixed URL detection for SSE vs Streamable HTTP. * **GitHub MCP Authentication Bypass** – Resolved issue where user-auth servers connected without credentials. **Week of August 8th Release Notes** **New Features & Improvements** * **"Create Your Own Agent" Redesign** – A refreshed, more intuitive experience for building and editing agents. * **User Authentication for Custom Tools** – Build tools for external APIs that users can securely access with their own credentials (Bearer, Basic, API-Key). * **Network Groups + DB Tool Support** – More flexible grouping logic and expanded database integration. * **Customizable Departments** – Apply department-level structure across requests, agents, feedback, and interfaces—making it easier to route, organize, and track initiatives. Department settings are now available in the Settings panel. Agent lifecycle feedback can now also be tracked and searched by department. * **Dark Mode** – Now available platform-wide for improved accessibility and a sleek user experience. * **Model Grouping by ACC** – Models now support grouping by account configuration context (ACC). * **Refresh Button in Model Cards** – A new refresh button has been added to the Model Cards section in the Configuration Drawer. * **New System Variable:** `DataStoreID` – Available for use in "Loop over Files". * **Unified Prompt Structure** – All prompts are now managed as a single message to streamline handling. The prompt management interface has been updated accordingly. * **Configurable Tool Timeouts** – Tools now support timeouts of up to 5 minutes. * **Google Connector with Service Account** – Added support for service-account-based connections. * **Updated UX for Data Source Syncing** – Users can now choose whether to save changes with or without syncing the data source, offering more flexibility based on use case. * **New JSON Formatter Step in Agent Studio** – Lets you control how your JSON is structured: * **Visual Builder** – Use agent variables as key/value pairs and preview the structure live. * **Raw JSON Editor** – Paste in any valid JSON and get immediate validation and preview. * **Deprecated Model:** `o1-preview` * **Model Library now includes GPT-5 Model Family** **Bug Fixes** * Resolved pagination issues in the MCP Library. * Addressed loading failures in OP Settings. * Improved visual consistency across layout elements (e.g., oversized model icons, small-screen responsiveness, header alignment, dark mode descriptions). * Fixed scrolling behavior in the Embedded Chat input box. * Improved error handling for application load failures and clarified error messages. * Enhanced delete flow in the Applications UI. * Corrected missing or improperly loaded translations in Smart Scan. * Fixed incorrect URL in Smart Scan failure notifications. * Resolved Excel parser issue affecting on-demand processing. * Fixed folder selector behavior for SharePoint libraries. **Airia Platform Release Notes – July 2025** **Platform Enhancements** **Looping Skills for Agents** Agents can now automatically loop through structured content—whether it's rows in a CSV or multiple files—allowing them to process large sets of information in a single workflow. Great for batch tasks, summaries, and reporting. **Testing Suites** New testing tools let you simulate inputs and verify agent responses before deploying—making it easier to build with confidence and avoid surprises. **Voice Agent Behavior** Voice-enabled behavior supports agents to have real-time, voice-to-voice interactions directly in chat, ideal for hands-free use cases. **Scheduled Sync for Data Sources** You can now schedule when your connected data sources sync, ensuring your agents always have up-to-date information without manual effort. **AI BOM (Bill of Materials)** Gain visibility into what powers each agent—models, datasets, tools, and versions. This update makes it easier to audit, maintain, and scale responsibly. **Enhanced Lifecycle Management** Administrators now have full transparency into all agents with a unified management view. Agent requests and user feedback are more actionable with built-in notifications and visibility directly within the builder experience. Departments can now be customized, and apply from request through agent, feedback, and interfaces - making it easier to route, organize, and track initiatives across teams. **Updated Interfaces Experience** We've made it easier to publish triggers, deployments, and extensions for agents — all in one streamlined interface accessible while building and administering alike. **Library MCP Servers** Run your agent libraries in multi-customer environments, securely and in isolation—ideal for enterprise-scale needs. **URL Sanitizer Guardrail** Airia now detects and removes unsafe or malformed links from agent outputs—adding an extra layer of safety to every interaction. **Platform Fixes & Stability Improvements** * More reliable file ingestion for CSVs, PDFs, and large uploads * Improved error handling in agents connected to SQL and external APIs * Enhanced output formatting for exports to Word, PDF, and slides * Fixed issues with complex prompt variables and fuzzy query parsing * Admins can now reliably delete projects with global data sources * Expanded model routing support for Azure-hosted Claude models **Week of August 1st Release Notes** Airia Platform – Recent Updates & Enhancements\ We've been busy making improvements across the Airia platform! Here's a roundup of the latest features, enhancements, and fixes: **New Features & Improvements** * **Voice is now live** in chat experiences – enabling seamless, real-time conversations. * **Smart Scan** released for enhanced document understanding. * **Loop Step on Canvas** now available for advanced agent workflows. * **Execute agent versions** using our new orchestration service for better performance and flexibility. * **Data Source Sync Scheduler** is now available for all customers. * **Native Semantic Search Tool** released for dynamic, LLM-powered search experiences. * **Draft Mode Enhancements:** In-progress edits are now clearly labeled (e.g., "3.01 Draft"). * **MCP Enhancements:** * OAuth support enabled. * Google Spreadsheets scope/tools added. * Routing via ACC now available. * **Custom Tool Improvements:** * Support for nested objects. * User-based authentication added. * **Interface & Webhook Management:** Centralized configuration rolling out across all environments. * **Audit Logging** introduced (initial release – ongoing improvements underway). * **SQL Query Results:** Column names now included in DB Tool responses. * **Agent Studio:** View and edit requested action items directly within the platform. * **Agent Search:** Users can now search agents by interface details. * **Terminology Update:** "Data" has been renamed to **"Knowledge"** for clarity in the end-user experience. * **Unified Agent Configuration:** Admins and builders can now access agent settings in a single, consistent view. * **Crawl4AI** has replaced FireCrawl for improved website data ingestion. * **Mobile App Updates:** * **Airia Chat iOS App v2.10.0** released. * UI update: "Settings" renamed to **"Profile"** for a more personalized feel. **Platform & Infrastructure** * **NAT Gateways Deployed:** Customers may notice changes in egress IPs used for firewall whitelisting. **Bug Fixes** * Fixed time mismatch between view logs, chat window, and execution ID. * Resolved issue where custom models lost support after edits. * Projects can now be deleted even when global data sources are present. * SSO users no longer need to accept Terms & Conditions twice. * Fixed duplicate user account creation issue. * Resolved missing Integrations pane in Chat. * Multiple fixes to improve Dark Mode consistency. * Addressed Entra KC and OneDrive/SharePoint sync issues. * Fixed feed filtering for project-specific views. * Project lifecycle now shows project **name** instead of ID. * Webhook edit modal has been resized for better usability. **Week of July 21st Release Notes** **Enhancements** * **Word File Exports Now Support Markdown Tables** * Tables formatted in Markdown now render properly in Word file exports for cleaner documentation. * **"Settings" Renamed to "Profile" in Airia App** * Updated to reflect a more personalized user experience. * **"Data" Renamed to "Knowledge" in End User Experience** * A clearer, more intuitive term for where your organization's structured and unstructured content lives. * **Consistent Agent Configuration View for Admins and Builders** * All agent settings are now accessible from a unified, consistent panel—improving efficiency across roles. * **Search Agents by Interface Details** * Users can now search for agents based on prompts, tools, and other interface-level configurations. * **SQL Query Results Now Include Column Names** * Query outputs from DB Tools now include column headers, improving clarity and usability. * **Data Source Sync Scheduling Released for MENA Region** * Users in the MENA region can now schedule automatic syncs with their connected data sources. * **Website Connector Upgraded to Crawl4AI** * We've upgraded from FireCrawl to Crawl4AI for enhanced website indexing and reliability. * **Infrastructure Upgrade: NAT Gateways Deployed** * Airia environments now use NAT Gateways, changing outbound IPs. Customers using IP-based firewall whitelisting should update their configurations accordingly. **Mobile Update: Airia Chat iOS App v2.10.0**\ The latest version of the **Airia Chat iOS App** is now live in the App Store, bringing several key improvements: * **Voice Mode Revamp** * A fully redesigned voice experience featuring: * Faster response times with near-instant streaming * Branded visuals using your platform's primary color * Support for markdown-formatted content (e.g., bullet points, stories) * Ability to mute/unmute your mic mid-conversation * Full conversation thread view available after voice interactions * **Canvas Mode in Chat** * Users can sketch or draw directly in chat—send visuals to the agent for intelligent parsing and response. * **Provide Feedback via Agent** * Easily submit feedback during any chat interaction. Feedback is sent to platform admins for review. * **Personalized Branding on Assistant Home Screen** * Your platform branding is now reflected in the app's home screen for a more tailored experience. **Bug Fixes** * **Resolved Time Mismatch Between View Logs, Chat Window, and Execution ID** * Logs and chats now reflect accurate and synchronized timestamps. * **Fixed Project Deletion Issue with Global Data Sources Present** * Projects can now be deleted even when tied to global data sources. * **Fixed Support Loss for Custom Models After Editing** * Edited custom models retain full support and functionality. ## **Airia Platform Release Notes – June 2025** **Platform Enhancements** **Updated Navigation** \ Made it easier for admins to easily access major modules within Airia by adding easy-to-access top navigation. Added search shortcuts to also aid in quick navigation. **Permission Aware RAG** \ With permission-aware RAG, users can retrieve only the data they have access to—nothing more. This ensures sensitive information stays protected, supporting privacy and compliance across teams and clients. **Smarter Email Agents with Graph RAG** \ Knowledge Graphs can be built inside the platform to capture and retrieve more accurate, contextual and nuanced information. **New IF/ELSE Conditions in Agent Studio** \ No-code logic just got more powerful. Use conditional blocks to create flexible agent behavior: different responses based on user input, data results, or conversation context. **Secure, User-Level Tool Permissions** \ Limit access to internal tools by user or team—ideal for managing sensitive workflows, preventing misuse, and maintaining clearer ownership across your organization. **Simplified Start Page for Airia Accounts** \ A new launch space makes it easier for Aria users to find tools, continue conversations, and get help—especially useful for first-time users or those switching between use cases. **SAML Support** \ In addition to supporting OIDC, Airia now supports SAML for Single Sign-On (SSO) ## **Airia Platform Release Notes May 2025** **Platform Enhancements** **Create Documents** Generate polished documents automatically from agent outputs—perfect for summaries, reports, or customer-facing deliverables. **AI BOM (Bill of Materials)** A new governance layer tracks which models and data power your agents, simplifying audits, security, and compliance. **ServiceNow Connector** Expanded integration enables richer data syncing and easier ticket handling within workflows. **Output Formatting Enhancements** Export agent responses directly into: * Microsoft Word and PowerPoint * Custom-branded templates * Auto-generated slide/page breaks * Documents with watermarking support Available directly through agent tools and chat interfaces. **Agent2Agent (A2A) Workflows** Full support for A2A-compatible agents to securely communicate, coordinate, and share data across applications and organizations—without heavy configuration. **Additional Model Support** * Execute OpenAI-compatible models from any client * Added Alibaba model provider * Tools/function calling for Gemini and Mistral * Library improvements for Alibaba, Cohere, Mistral, and Gemini * Support for Machine Learning models **AWS Deployment Support** Deploy Airia directly into AWS environments for customer-managed infrastructure. **Improved Debugging in Agent Studio** Easier tracing of agent behavior, quick issue identification, and development visibility. **Guardrail UX Improvements** New visual controls simplify reviewing and applying security guardrails across agents. **Status Indicators During Execution** Real-time step indicators show your agent's current execution progress. *** **Important Platform Bug Fixes** * Fixed issues where agents stalled during large file ingestions * Resolved bugs in output formatting (e.g., missing slides or broken Word exports) * Improved ingestion reliability and visibility for failed/partial uploads * Fixed prompt loading issues in complex agent projects * Updated model routing for fuzzy SQL queries and external APIs *** **Chat & Collaboration Enhancements** Conversations Feed Centralized feed of agent conversations for easy referencing, troubleshooting, and collaboration—ideal for ongoing workflows and multi-user teams. Optimized Guest Experience Streamlined guest login and interaction to improve external collaboration. *** **Important Chat Bug Fixes** * Fixed broken citation links in document responses * Resolved issues with disappearing chat history * Improved mobile chat responsiveness and fixed speech-to-text playback issues * Enhanced syncing for embedded chat widgets across devices *** **In Summary** This month's updates bring more clarity, control, and collaboration to your AI workflows: * **Smarter building** with Agent2Agent, DocGen, and new model support * **Deeper insights** through Conversations Feed and improved debugging * **Enterprise-ready flexibility** with AWS deployments and provisioning upgrades * **Enhanced stability** across Chat, Studio, and Ingestion * **Cleaner outputs** with beautiful, branded formatting options We're excited to see how these enhancements will elevate your experience with Airia and support even more powerful AI-driven solutions. hello ## **New Features & Improvements** * **Support Tab** – Quickly find self-service resources and help articles right inside the platform. * **Security Admin Role** – New role that allows dedicated admins to manage security settings and view security-related data. * **Airia Community** – Easily import/export agents and connect with the wider community for ideas and best practices. * **Agent Constraints** – Add IF-THEN rules that check conditions before agents access resources, giving you more control. * **Agent Details Tab** – Cleaner agent settings with a new dedicated tab for details. * **A2A Client Library** – Available as a NuGet package to simplify developer integrations. * **Pre-Release Agent Versions** – In-product announcements highlight upcoming agent versions so you can test early. **Bug Fixes** * **Data Processing Tab** – Fixed an issue where no results were displayed for on-prem deployments. * **People Page** – Sorting by last activity now works as expected. * **Slider Tooltips** – Corrected display issues where tooltips extended outside the interface. * **Options Panel** – Fixed an issue where the panel didn't close after saving an agent in Agent Studio. * **Agent Execution Events** – Resolved an issue preventing events from being sent to syslog. **Airia Platform Release Notes April 2025** **Major Features** * **Granular Guardrail Assignment**: Assign security guardrails at a more detailed, flexible level across users, projects, and data types. * **Tools as Actions on Canvas**: Add and manage tools directly within agent workflows to make building and customizing agents more intuitive. * **Human-in-the-Loop (HITL) Support**: Insert manual approvals into automated agent flows, giving human reviewers the ability to step in when necessary. * **IDP Provisioned Groups for Access Management**: Seamlessly manage user permissions across Airia by connecting Identity Provider (IdP) groups. * **SQL Retrieval in Agent Studio**: Easily integrate live SQL queries into agent flows, allowing agents to retrieve up-to-date data during execution. * **SQL Tool - Databricks Query Support**: Agents can now execute SQL queries on Databricks directly, unlocking more enterprise data use cases. * **Agent Router Enhancements**: Smarter routing of agent tasks to the most appropriate models and tools, improving both speed and accuracy. * **Output Formatting Enhancements**: We now support the ability to export any agent responses to Microsoft PowerPoint and Word file formats without leaving the platform. Contextually identifies logical page and slide breaks for Word and PowerPoint files respectively. * **Template Support:** You will now be able to provide a custom single slide template for PowerPoint or single page template for Word. Agent responses will be available to view in the styles of the associated template. * **Watermark Support**: Add custom watermark text to generated documents and presentations. * **Availability:** This functionality is available to be use through tool calls as well as Mobile and Web Chat experiences. * **Forecasting Capabilities**: Forecasting endpoints now available directly within Chat for easy use in conversations. Available by request, only. **Data Ingestion Enhancements** * **Folder Selection for Data Ingestion**: Choose specific folders/content when ingesting from SharePoint, OneDrive, Dropbox, OneNote and Confluence connectors. * **New connectors: Website content ingestion** * SQL ingestion of large structured data allows for text-to-SQL queries **Platform Experience and Reliability** * **Model Library Expansion**: GPT-4.1, o3, o4-mini, Llama 4, Claude 3.7 models added to the model library. * **Enhanced Data Search**: Fuzzy search is now supported in SQL data retrieval, providing more flexible results. * **Expanded Trust Center Capabilities**: Improved classification management and support for including non-Airia AI assets. * **Improved Performance and Stability**: Faster page loads, streamlined workflows across agent studio and chat environments. **Chat and Collaboration Updates** * **Spaces for Collaboration**: Launch collaborative spaces where teams and guests can chat, share artifacts, and work with agents together. * Create spaces directly from Assistants or Chat conversations. * Manage and archive spaces while maintaining full conversation history. * Invite internal users or external guests to participate securely. * **Mobile App Availability:** Our Mobile App is available across Europe, Middle East, and Asia Pacific Regions for iOS, iPadOS, and macOS * **Status Indicators During Agent Execution**: Real-time updates show when agents are processing a request. * **Optimized Guest Experience**: Easier guest login and improved experience for mobile users. * **Knowledge Management Improvements**: * Friendlier timestamps for ingested documents. * Visual status indicators for document ingestion states. * Refreshed Assistants Page: Updated styling, search, and sort of Assistants, plus a new library experience with Assistants categorization. **Key Bug Fixes** **Platform Stability** * Resolved issues where agents stalled after uploading multiple files into blob storage. * Addressed missing model options during agent routing and execution. * Improved ingestion retries and corrected unexpected ingestion charges related to upload timeouts. **Data Ingestion** * Corrected SQL ingestion failures for Excel files where only headers were processed. * Improved handling of PDF ingestion timeouts for larger documents. * Enhanced error handling and ingestion status tracking across connectors. * Enhanced OCR capability **Agent Studio and Tools** * Addressed missing prompt variables during project prompt loading. * Resolved SQL Data Search fuzzy search errors and output misalignment. * Updated external API libraries to improve connection reliability (e.g., Finance). **Chat and Collaboration** * Improved reliability for citation links opening the correct document page. * Fixed conversation history disappearing when switching between agents. * Enhanced synchronization and message reliability after agent executions in embedded chat. * Improved mobile responsiveness for speech-to-text playback and overall chat interactions **Summary** This release brings powerful upgrades to Airia's Platform and Chat experiences, including: * Collaboration through **Spaces** * Stronger security with **Granular Guardrails** and **IDP Group Management** * Smarter agents with **SQL Retrieval**, **Agent Router Improvements**, and **Databricks Integration** * Easier ingestion and file management across **SharePoint, OneDrive, Dropbox, and OneNote** * Better document generation with **Watermarks and DocGen** in Mobile and Web Chat * Over **100+ reliability and performance fixes** across the Platform We're excited to deliver even faster, more flexible, and more secure AI-powered workflows with these new capabilities!