Release Notes - 2.0.5

ELITEA 2.0.5 focuses on making the platform easier to configure, scale, and understand in daily use. This release expands AI-assisted creation and editing, strengthens skill reuse and publishing workflows, adds usage and analytics visibility and improves settings, sidebar, indexing, toolkit, chat and artifact behavior.

• Release version: 2.0.5
• Released on: August 9, 2026
• Access: ELITEA Platform

​

New Features

​

Edit with AI

You can now use Edit with AI to update existing agents, skills and project context. Describe desired changes in plain language, review AI-generated updates and apply only the changes you want.

• Use Edit with AI to update: for agents - their details, instructions, user interaction, tools and skills; for skills - skill’s name, description and instructions; and for the Project Context - the content of the Project Background field.
• Review current vs. suggested content through side-by-side or step-based comparison flows that highlight differences.
• Keep full control over what gets applied by further editing, excluding unwanted suggestions, or canceling the flow entirely.

Benefits: Faster iteration on existing agents, skills and context with less manual rewriting. Details: Agents, Skills, Project Context.

​

Progressive Disclosure of Skills

You can now let agents apply attached skills more autonomously through progressive disclosure, reducing the need to invoke every skill manually.

• Get smarter skill-driven orchestration.
• Apply attached skills autonomously, allowing agents to use relevant skills without requiring an explicit ~skill-name in every message.
• Ensure agents only work with the skills attached to their own configuration.
• Allow multiple relevant skills to be applied in the same turn when the task calls for them.
• Surface loaded skill details more clearly during execution so users can better understand which skills were applied.

Benefits: Less manual invocation and more natural, scalable use of attached skills. Details: Skills, Agents.

​

Skill Publishing

You can now publish reusable skills into a shared catalog through a guided flow backed by deterministic and AI-powered quality validation.

• Publish skills to the Catalog through a guided self-service wizard with built-in deterministic and AI-assisted quality checks against skill name, description, instructions, icon, version name and category.
• Assign skills to curated categories so it is easier to browse and manage them in the shared catalog.
• Publish and manage multiple versions of the same skill, with version naming and validation handled within the publishing flow.
• Keep skills attached to an agent when it is published.
• Rely on platform-level controls to validate, block publishing, or unpublish skills.

Benefits: Faster skill sharing, better skill quality control and safer publishing of skill-dependent agents. Details: Skills, Catalog.

​

Skills Customization and Forking

You can now fork skills across projects, personalize skills with custom icons and write richer skill instructions.

• Maintain skills as reusable building blocks.
• Fork a skill version directly to another your project without manual export-and-import workarounds.
• Fork agents together with their attached skills to preserve skill-backed behavior.
• Write skill instructions up to 5,000 characters to ensure better guidance and include examples and output expectations.
• Upload custom icons for skills.

Benefits: Better cross-project reuse, more expressive skill definition and clearer visual identification. Details: Skills, Catalog.

​

Aha! Toolkit

You can now connect Aha! as a native toolkit to work with product-management records in agents, pipelines and chats and index supported Aha! data for AI use.

• Connect Aha! as a toolkit and validate the connection with clear success or error feedback.
• Use personal or project-scoped Aha! credentials for individual access or shared project use.
• Work with Aha! records through dedicated tools, including find projects, search, read, analyze and copy records, retrieve data, add comments, etc.

Benefits: Smoother Aha!-based agent and indexing workflows with safer authentication and error handling. Details Aha! Toolkit.

​

Bucket-Level Access to Team Artifacts

You can now control who can view or modify content in team artifact buckets through a redesigned Manage Permissions experience.

• Manage access per bucket by assigning clear bucket permission levels.
• Use the redesigned Permission Exceptions view with searchable user selection, bulk actions and clear single-user or multi-user update patterns.
• Apply bucket-level permissions to all nested folders and files through a single permission model.
• Enforce bucket permissions consistently across the ELITEA Artifacts API and S3-compatible routes, preventing bypass through direct requests.

Benefits: Better control over shared artifacts, clearer access-management workflows, consistent permission handling across interfaces. Details: Artifacts.

​

New Analytics for Costs

You can now track token usage, AI costs, sub-agent activity and chat-level insights across your projects.

• Track token consumption across supported AI activities.
• Use enhanced cost tracking to better understand the financial impact of model usage.
• Gain clearer insight into delegated and nested agent execution through improved sub-agent observability.
• Understand conversation-level activities and usage patterns better through the expanded chat analytics.

Benefits: Better usage transparency, increased cost awareness and stronger observability for complex agent workflows. Details: Analytics.

​

Usage in Settings

You can now track spending, token usage and model-level consumption from a dedicated Usage page in Settings, available for both private and team projects.

• Review current usage in a single project-focused view for private projects, team projects and per-user views within team projects.
• Monitor monthly spend against budget limits and track when the billing period resets.
• Review prompt and completion token totals together with request counts for supported shared-model usage.
• Analyze model-level usage details including cost, calls, token volume and relative share by model.
• Export usage data to Excel.

Benefits: Better AI consumption visibility, clearer budget awareness and easier model-level analysis for private and team projects. Details: Usage, Settings.

​

HTML and MDX Preview & Editing

You can now upload, preview, edit and save HTML and MDX files directly in the Artifacts and Chat canvas without leaving the platform.

• Open .html, .htm and .mdx files in Artifacts or directly from the generated chat results without downloading them first.
• Switch between the Preview (default, renders output) and Raw (syntax-highlighted source editing) modes.
• Edit, save and discard changes in HTML or MDX source in the Raw mode, writing changes back to the same file in the same artifacts bucket and path.
• Rely on sandboxed preview isolation that prevents rendered HTML or MDX content from accessing ELITEA session data, cookies, local storage, or the parent page.
• Handle unsupported or unsafe MDX constructs by blocking execution in preview and displaying a clear warning when a file contains unsupported or unsafe components.

Benefits: Faster and safer preview and editing workflows for HTML and MDX content. Details: Artifacts, Chat.

​

Personality-Specific Default User Instructions

You can now save different default user instructions for each personality type so that switching between personalities automatically switches to the matching instruction set.

• Save separate default user instructions for each personality type.
• Switch between personalities and automatically restore the instructions previously saved for the selected personality.
• Start from a blank instruction field when opening an unconfigured personality, instead of inheriting unrelated text.
• Work with a dynamic instructions field that stays aligned with the selected default personality.
• Carry forward existing default instructions into the new personality-specific model for backward compatibility.

Benefits: More predictable personalization workflow. Details: AI Personalisation.

​

Changed Features

​

Toolkits & Indexes Redesign

You can now manage the full index lifecycle — from discovery and configuration to execution, scheduling and history — through a redesigned Toolkits & Indexes landing page.

• Discover index-capable toolkits on the redesigned Toolkits & Indexes landing page with searchable toolkit cards, index counts and clearer breadcrumb-based navigation between toolkit- and index-level pages.
• Create and manage indexes from dedicated pages within the toolkit details view, with a safer reindex workflow and delete confirmation.
• Monitor index runs with clearer status indicators, indexed and skipped counts, structured run history and activity and search settings.
• Configure recurring indexing through a redesigned scheduling flow.

Benefits: Better index discoverability, clearer lifecycle management and improved visibility into run outcomes. Details: Toolkits, Indexing Overview, Create & Use Indexes.

​

Safer Reindex Workflow

You can now update indexing-critical settings for an existing index through a safer reindex workflow that keeps saved configuration in sync with actual index contents.

• Edit indexing-critical settings such as source settings, chunking parameters and filters through a workflow that recognizes when a reindex is required and applies changes through a dedicated reindex path rather than as ordinary settings updates.
• Avoid silent configuration drift where the stored index still reflects older parameters after toolkit settings have changed.
• Protect the existing index if a configuration-triggered reindex does not complete successfully.
• Get clearer signals when configuration edits affect index quality or source coverage and require rebuilding the index.

Benefits: Consistent index contents, safer maintenance, less risk during configuration changes. Details: Indexing Overview.

​

Reorganized Settings

You can now find project setup and personal preferences more easily through a reorganized Settings structure.

• Navigate through redesigned Settings with clearer grouping and improved access to settings and notifications.
• Switch projects using a project selector at the top of the sidebar with default or custom project avatars for faster recognition.
• Benefit from improved notification visibility and access.
• Match ELITEA theme to your system theme automatically.

Benefits: Clearer navigation, improved notification visibility, more flexible theme experience. Details: Overview, Settings Overview.

​

Updated Inventory Setup & Ingestion

You can now configure and operate the Inventory application more reliably with improvements across setup validation, ingestion control and bucket configuration.

• Get clearer source-related validation feedback earlier in the app creation flow.
• Stop active ingestion runs more reliably when an interruption is needed.
• Avoid overlapping ingestion runs.
• Use a fixed internal bucket for Inventory data aligned to the dedicated inventory-graphs path.
• Work with more consistent field-label sizing in Inventory and related application setup flows.
• Use a more supportable Kubernetes deployment setup.

Benefits: More predictable setup, safer ingestion control, less configuration overhead. Details: Inventory App.

​

Internal Tools Renamed to Modules

ELITEA now uses Modules instead of Internal Tools in key configuration and chat surfaces.

• See Modules as the section label on agent and pipeline detail pages where built-in capabilities are configured.
• Continue using built-in capabilities such as image creation, data analysis, planner, Python sandbox, swarm mode, smart tools selection and related options without any workflow changes.
• Distinguish built-in platform capabilities more clearly from Tools, Toolkits and MCP Tools.

Benefits: Clearer terminology. Details: Chat, Agents, Pipelines.

​

Fixed Issues

• #6057 Clicking Continue extends the previous incomplete response instead of replacing it with a separate fresh answer.
• #5996 GitLab toolkit initialization no longer fails on missing description values.
• #5994 GitLab root listing no longer requires a path value in the UI.
• #5894 Remote MCP sessions no longer remain active after logout.
• #5823 Regenerated responses show the real thinking duration for that request.
• #5821 Model-specific settings align with the selected model behavior, preventing related sub-agent execution failures.
• #5819 @Everyone does not trigger the inconsistent LLM behavior or infinite-loop pattern.
• #5784 Users no longer need to click Load tools to trigger authentication for a configured MCP.
• #5643 Nested agents with remote MCPs trigger the expected authorization request.
• #5638 SharePoint toolkits support a smart OAuth authentication flow for agents.
• #5391 Langfuse traces record all executed pipeline or agent nodes.
• #5390 Langfuse pipeline traces contain all the necessary data.
• #5267 Elitea MCP Tools are available when an agent is invoked as a sub-agent.
• #3745 Stopped indexing runs from chat, agents and pipelines have the correct status.

​

Known Issues

• #3151 PPT Files Fail to Read or Index from Artifacts or SharePoint.
• #2719 MCP client disconnects on macOS despite tray showing “connected”.
• #1567 MCP-tagged resources execute the latest version only.