This log shows the most important fixes or features added to the platform in 2025.
- Flows
- Improved single sign-on to use custom IdPs and custom subdomains, enabling secure access to Flows.
- Configured the organization UIConfiguration parameter to reference the client IdP for Flows.
- Set up a dedicated OpenID Connect IdP in Keycloak and mapped roles and session attributes.
- Ensured the GAM application includes the OpenID scope.
- Improved diagnostics to enable optional endpoint interaction logging and clearer execution logs in the Flows Editor.
- Set DEBUG_FLOWS_API=1 to increase request and response verbosity.
- Set FLOWS_FORCE_SYNC_EXECUTION=1 to force synchronous execution for tenant API interactions when needed.
- Fixed error handling to prevent error messages from being sent to end users when Handle error is enabled in an Agent.
- Fixed missing items in the Flows Editor console panel so complete execution logs are displayed.
- Console
- Improved single sign-on to allow the use of custom IdPs and custom subdomains across the Console.
- Updated the organization UIConfiguration parameter to ensure SSO works with client IdPs and branded URLs.
- Ensured the FluentLab GAM module includes the OpenID scope by default for new environments.
- Fixed quota tracking to ensure the correct number of active sessions is shown when contexts are deleted.
- Fixed Control Panel taxonomy migrations so they can be run again from the Console.
- LLMs
- Fixed OpenAI rate limit disruptions to reduce 429 errors by enabling provider fallback for SaaS usage.
- Security
- Fixed sensitive OAuth token exposure on logout to reduce token data recorded in logs.
- Fixed a denial-of-service risk in Scrapy, ensuring protection against decompression-based DoS attacks.
- Fixed information disclosure in AI error responses to avoid exposing system details, returning generic messages instead.
- API
- Fixed GetPlugins Agent listing limits. Allows retrieving all Agents from GetPlugins even when a project has more than 100 Agents.
- Internal pagination was implemented transparently, and increasing page size as a workaround is no longer required.
- Console
- Fixed Organization Settings access control: Ensures that Organization Settings are visible only if you have an Organization role for that organization.
- Being a Project member or Project administrator in another organization no longer grants access to that organization’s settings.
- Fixed 403 errors behind OWASP‑protected NGINX at EY: Prevents receiving "403 Forbidden" errors when confirming TRNs or executing Console actions that send large payloads.
- Reduced POST payload size by avoiding full posts, removing unnecessary control properties, and preventing the WorkWithPlus menu SDT from bloating requests to over 1,000 parameters.
- RAG
- Improved ingestion stability on Azure with container‑aware memory settings. Allows ingesting large media files (e.g., MP4) and documents more reliably without OutOfMemoryError.
- Standardized JVM memory parameters (MaxRAM and MaxRAMPercentage) to match IaC settings, and applied the container‑aware memory fix validated on AWS to Azure.
- Fixed "File rejected" errors for .doc/.docx uploads in console.saia. Enables uploading Word documents to RAG from console.saia without rejection errors.
- MIME type validation and ingestion pipeline handling are aligned with the corporate environment.
- Fixed stuck bulk uploads and blocked new uploads after interruptions. Ensures uploads can be restarted and new files queued without encountering messages such as "Requests are currently being processed …%" or "File rejected."
- Interrupted batch tasks are now identified and canceled, and pending queue items are cleared to unblock new requests.
- Security
- Fixed vulnerabilities in GEAIRAG container images: Allows benefiting from updated base images and dependency patches identified by container scans.
- Changes are available in environments built from 20251201 onward.
- Data Analyst Assistant is deprecated.
- Flows
- New partial-result streaming so you can receive progress updates while a Flow executes across Slack and Teams. A streaming API is also exposed for programmatic consumption.
- Improved Flow creation by enabling “Create from import” in The Lab.
- Fixed a blank page when opening the bot builder from The Lab.
- Tools/Agents
- New “Test” button on each Agent card in The Lab so you can open a chat with that Agent without entering edit mode.
- Improved Agent import to automatically select the first available model in your environment from the template’s list, avoiding multiple fallbacks and UI mismatches.
- Improved Agent configuration to enable file uploads only when the chosen model supports it, showing supported extensions, categories, and size limits in context.
- Fixed an authentication error that returned an empty list when using the retrieve_tools Tool from Iris.
- Fixed Agent imports that referenced existing private Agents, reusing them instead of failing or duplicating.
- Fixed error tracking in the new “Test” chat so the trace ID displays correctly.
- Fixed chat history not rendering in the “Run Test” panel and ensured messages persist on reload.
- Fixed the “More options” menu on the last row of the dashboard so you can access all actions.
- API
- New OpenAI-compatible endpoint chat/completions so you can plug in Xcode and other standard clients without changes.
- New images/edits endpoint in Image API (v1/images/edits) for image editing with OpenAI models.
- New video generation endpoint for Vertex AI Veo 2 so you can request text-to-video generation via the API.
- New Parameters API so you can retrieve Globant Enterprise AI configuration programmatically from The Lab and Station.
- Improved APIM integrations by forwarding the JWT and injecting Ocp-Apim-Subscription-Key when the provider “flow” flag is enabled.
- Improved stability by enforcing a 60-second max execution time at the MySQL driver level to abort long-running queries automatically.
- Fixed S3 uploads in “bucket owner enforced” mode by removing PutObjectAcl calls that caused 403 AccessDenied.
- RAG
- New S3Vectors vector store support so you can create, delete, and query indexes programmatically, with automatic index creation and required IAM permissions applied.
- New observability by propagating the X-Saia-Trace-Id header through RAG requests so you can correlate traces across services.
- Improved telemetry by fixing OpenTelemetry traces that did not appear in AWS X-Ray.
- Fixed a non-deterministic deadlock when deleting documents by adjusting the persistence strategy.
- Fixed an error with gpt-5-nano in Assistants by removing the n:1 default in LLMs templates and issuing a new module revision.
- Improvements are made to the omni-parser, which speeds up processing.
- LLMs
- New Anthropic models.
- Haiku 4.5-20251001
- Claude-opus-4-5-20251101
- Claude-opus-4-5-20251101-v1:0
- New AWS Bedrock models.
- awsbedrock/openai.gpt-oss-120b-1:0
- awsbedrock/openai.gpt-oss-20b-1:0
- awsbedrock/haiku 4.5-20251001
- New Open AI models.
- gpt-image-1-mini
- gpt-5-chat-latest
- gpt-5-codex
- gpt-5-pro
- gpt-5.1
- gpt-5.1-codex
- gpt-5.1-codex-min
- New Open Router model.
- moonshotai-kimi-k2-thinking
- New Vertex AI models.
- claude-opus-4-5-20251101
- gemini-3-pro-image-preview
- gemini-2.5-flash-image (Nano banana)
- openai-gpt-oss-120b-maas
- openai-gpt-oss-20b-maas
- New xAI models.
- grok-4-1-fast-reasoning
- grok-4-fast-reasoning
- Improved OpenAI routing by migrating chatgpt-4o-latest to gpt-4.1 with a safe fallback so you get function calling support.
- Fixed codex-mini requests by dropping the unsupported parameter stream_options.include_usage.
- Fixed GPT-5 requests by ignoring the unsupported reasoning_effort parameter when sent.
- Lab
- Improved Agents file upload configuration so you can enable uploads when supported by the model, with clear limits and supported types. The checkbox visibility in Agent detail was fixed.
- Improved the Agentic Processes engine with throw/catch handling in intermediate events.
- Fixed multiple input files being overwritten during process start or intermediate events so each file is preserved and processed correctly.
- Fixed 502 errors when listing Jobs in environments with large datasets.
- Station: Discover and Scale AI Across the Organization
- The Station is the new Globant Enterprise AI interface designed for consumers to easily explore, adopt, and execute AI solutions. Serving as the centralized entry point for company-wide AI enablement, it empowers end consumers to discover and interact with Agents.
- This first release includes:
- Search, filter, and discovery of available AI solutions
- Detailed solution pages to understand capabilities and use cases
- Solution sharing via public links
- Redirection to Workspace for execution
- Ratings and reviews to gather user feedback
- Help and contact support options to ensure user success
- Tenant-isolated SaaS deployment option for Station, provisioned in a client’s cloud or on-prem environment, with IdP integration (Azure AD/Entra, Okta, SAML/OIDC) and in-tenant admin moderation for approvals.
- AI Pods client environments in Station, configurable per client organization, including IdP setup and an administrator role to moderate and publish only approved solutions.
- Global Navigation header so you can switch suite (Station, Lab, Console, Workspace), Organization, and Project from the top bar. The header uses geai-ui-library, includes role-aware modals, and routes to modules via dynamic, environment-aware URLs.
- New shareable links for the Solutions Gallery so you can share the current search, filters, and ordering.
- New taxonomy model support across Station and Admin so you can manage org-scoped taxonomies with translations and use them in filters, publish flow, and solution details.
- CRUD endpoints with translation support and org scoping.
- Middleware normalization (caption + stable taxonomyKey) and recursive item handling.
- Optional embeddings for taxonomies so semantic search can include taxonomy context.
- Improved Publish to Station flow so you can select org-specific taxonomies dynamically in Step 3, with clearer notifications and a refined modal. The Subscription button is hidden to avoid confusion.
- Improved execution and sharing controls so you can set chat sharing and external execution permissions per Agent.
- The UI enforces the maximum scope configured by your organization.
- Confidential solutions disable the Run button and show a contextual tooltip; cross-project execution respects project-only permissions.
- Improved Admin experience so you can sort the Solutions list and see only the current project’s data, with new roles replacing the legacy AIStationAdministrator:
- StationOrganizationAdministrator — org-wide moderation.
- StationProjectAdministrator — project-scoped moderation.
- Improved search usability with inline warnings for fewer than 3 characters and unsupported special characters, and a clearer “Explore Solutions” CTA on empty results. The Share button now appears when searching.
- Improved UI polish to match design specs:
- Fixed filter reset/clear/apply buttons and handling of items without children.
- Fixed taxonomy display in shared links when the target org lacks those items.
- Updated filter icon, spacing between title and subtitle, avatar size (88x88), font weights, and Arabic sidebar width.
- Enabled Japanese in the language selector.
- Fixed guided walkthrough issues: corrected tooltip order, alignment with the Run button, persistent overlays, and reappearance on org changes.
- Console
- New API dimension in Dashboards so you can analyze Cost, Tokens, and Requests by API Token and model, with date filters and usage KPIs. Usage details now include User and API Token dimensions. More information: API Dashboard
- Improved Organization Settings with a redesigned General Settings card so you can control “Publish to Station” and “Share conversations,” including a maximum execution scope and org-wide warnings.
- Improved labeling by renaming “Assistant” to “Agent” across Dashboards.
- Fixed Reasoning Strategy creation so you can create strategies and see network activity as expected.
- Fixed document indexing to RAG Assistants that timed out at /upsertDoc.
- Fixed legacy objects by removing the deprecated PlaygroundChatbotWeb.
- Workspace
- New paste-from-clipboard support so you can paste screenshots directly into chat.
- New drag & drop support so you can drop files into chat when the selected model supports uploads.
- New inline safety warning in Legacy Workspace so you see a non-blocking alert when files or external links appear in chat.
- New Global Navigation header aligned with the suite-wide experience.
- Fixed Japanese localization issues so item names do not overlap and chat titles can be edited in Kanji without saving after each character.
- Fixed styling in the Workspace so the Consumer Information combo is visible.
- Fixed permissions for creating anonymous conversation links so you can share when org policies allow it.
- Security
- The controls performed in the file upload process were improved.
- Security fixed a critical jersey-client vulnerability by updating org.glassfish.jersey.core:jersey-client to a non-vulnerable version family.
- Security addressed a High vulnerability in RAG (expr-eval CVE-2025-12735) by updating/mitigating the dependency.
- Security reduced container attack surface for Console images by triaging and remediating Trivy findings.
- Security migrated system notifications to AWS SES (no-reply-geai@globant.com) to comply with email security policies.
- Flows
- Fixed Chat API variables not reaching Flows, ensuring the User Account ID and other context variables are automatically available to the Flow.
- Fixed intermittent Flow executions that would get stuck without responding in Workspace and Microsoft Teams, improving reliability for ongoing conversations.
- Fixed the inability to use context variables in the Email component fields, enabling dynamic personalization of message fields.
- Fixed permission handling so that Editor Users can export Flows, aligning export capabilities with import permissions.
- Tools/Agents
- Security resolved repository vulnerabilities in Tools to harden Tool distribution and consumption.
- API
- New Access Control API to manage and verify user memberships, roles, and permissions across Organizations and Projects.
- New System API that exposes the Active Docker Image Version for observability and support.
- New Security API enables user authentication and profile retrieval within Globant Enterprise AI. It provides endpoints to generate OAuth 2.0 access tokens and access user information for secure integration with protected resources.
- RAG
- Improved configuration of the RAG Document Service to support per‑Organization and per‑Project host URLs, enabling granular routing in multi‑tenant environments.
- Improved usage reporting to include token counts and cost when a vLLM is used.
- Honors INCLUDE_USAGE_COST_IN_RESPONSE=1 to return detailed usage.
- Uses LOG_OMNIPARSER to control response logging size (disabled by default).
- Adds RequestId to responses for traceability.
- LLMs
- New Anthropic Claude Sonnet 4.5 models.
- Anthropic Direct: claude-sonnet-4-5-20250929.
- AWS Bedrock: us.anthropic.claude-sonnet-4-5-20250929-v1:0.
- Google Vertex AI: vertex_ai/claude-sonnet-4-5 and vertex_ai/claude-sonnet-4-5@20250929.
- 200K context window, up to 64K output tokens, reasoning, function calling, prompt caching, batch predictions, and multimodal I/O.
- Deprecated Models: claude‑3‑5‑sonnet‑20240620 and claude‑3‑5‑sonnet‑20241022 with fallbacks configured for continuity. Migration path: Claude Sonnet 4.5.
- Lab
- Improved Agent execution throughput by adopting a new job pooling strategy, reducing queue delays and CPU usage in the Workflow Engine.
- Fixed Agent export to detect circular dependencies and return a clear error when Agents reference each other in loops.
- Improved error handling on chat failures by validating Status Code 200 before saving conversations and surfacing the actual Chat API error.
- Console
- Fixed new installations failing to create Projects due to missing base data so you can provision instances without manual steps.
- The installer now initializes SearchProfiles and Vector Stores.
- Fixed truncated Input (and Output) fields in Request Exports so you can download complete data for audits and analysis.
- Fixed Data Analyst assistant failures using SQL unaccent so you can run accent‑insensitive queries without errors.
- Improved multitenant provisioning so adding/updating UI config domains also updates GAM settings via API.
- Workspace
- Improved error handling when the Chat API fails by checking response status before saving messages and returning the original error to the UI.
- Improved Beta environment configuration by adding service Base URL variables for Lab, Station, Console, and Workspace to ensure consistent routing.
- Security
- Fixed exposure of API Keys in the Agents module by masking values and enforcing write‑only behavior; stored secrets can no longer be viewed from the UI.
- Fixed SSRF risks by sanitizing outbound HTTP client calls with protocol whitelisting and host/IP blacklist rules aligned with OWASP guidance.
- Fixed container image vulnerabilities in the GEAI‑LITE LLM component and introduced recurring assessments with Trivy to prevent regressions.
- Fixed API Tokens remaining valid after Project or Organization deactivation by enforcing token invalidation on deactivation.
- Improved compliance by ensuring all Globant Enterprise AI components compile with JDK 21.
- API
- New omni-parser llamaParse Provider
- New outputFormat GEAI Ingestion parameter
- New Analytics API that allows you to programmatically access the same adoption, usage, cost, and performance metrics shown in Dashboards (Dashboard, Adoption Dashboard, and Usage Dashboard). Endpoints support filtering by date, Organization, Project, Agent, and user. They enforce security via ApiToken and permissions, and return both focused and generic response formats.
- New endpoints in the Organization API to manage API Tokens (create, revoke, update) with proper authorization.
- Improved list endpoints for Agents API and Tools to include total count, paging, duration, and filters by id-or-name and by toolId.
- Fixed organization creation failures via Organization API.
- Fixed Files API upload responses and stability with multiple uploads.
- Resolved empty 200 OK responses on upload.
- Corrected base64 handling when uploading a second file in chat.
- Workspace
- New Agent personalization for Workspace so you can configure how an Agent presents itself and behaves in conversations. Define conversation starters to guide consumers and add information items that explain the Agent's capabilities. This helps consumers understand what the Agent can do. For more information, see: Configure the Agent Presentation, Agent Presentation
- Fixed chat streaming responses for Assistants across Lab and Workspace.
- Lab
- New option to configure task outputs so you can decide whether each task writes to a process variable or produces an artifact.
- Process variables make outputs immediately available to subsequent tasks.
- Artifacts store outputs for reuse, auditability, and sharing.
- Improvements in Agentic Processes testing that allow you to analyze executions more effectively.
- View process variables of a running instance.
- View artifacts generated during execution.
- Download the full execution log trace in JSON format.
- New multitenant and multi‑AuthType login in Lab to display all configured options per provider
- New logout behavior terminates all active UI sessions platform‑wide.
- Fixed Agent list to show more than 50 items and added proper pagination behavior.
- Fixed process execution blocking at parallel gateways by ensuring merges complete.
- Fixed chat failures after another user modifies an Agent with Iris.
- Tools/Agents
- Improved Iris suggestions to prevent SubAgents from being suggested during Agent creation.
- Improved export/import to handle public resources and referenced public Tools.
- Improved A2A Agent Card URLs to support agentId and names with spaces.
- Fixed Tool name parsing to support hyphens (“-”) when parameters use inputSchema.
- Console
- The Backoffice article has been renamed to Console, to maintain consistency with the official product name.
- General Dashboard has been renamed to Usage Dashboard. In addition, Usage Details now includes breakdowns by API token, User, and Agent (while retaining Model). Key indicators (Cost, Requests, Input Tokens, Output Tokens, Total Tokens) and the breakdown by Project and Organization remain unchanged. The Organization column is displayed only if you have access to multiple organizations or hold a System Administrator role.
- New Adoption Dashboard so you can track value delivery and growth.
- User: KPIs (active users, projects, agents), requests/day, users per Project/Agent, top Agents/users, and token consumption with filters.
- Platform: creation and modification trends for Projects, Agents, Processes, and Flows using system data and APIs.
- RAG
- Improved omni-parser quality and robustness.
- Fixed CSV parsing for semicolon (;) separators.
- Fixed text concatenation artifacts in specific documents.
- Clarified and improved documentation for llamaParse provider.
- Improved LLM parameter handling in Knowledge now accepts decimal values for topP.
- Fixed Reindex validation for Document URL pattern.
- Fixed Workspace errors when RAG names contain square brackets [ ].
- Fixed access to RAG source files so authorized users can open cited documents without additional login prompts.
- Security corrected IAM permissions for AWS Bedrock Knowledge Base retrievals.
- LLMs
- New models available:
- vertex_ai/deepseek-r1-0528-maas — on Vertex AI with streaming and tool calling support. Not multimodal.
- Improved embeddings coverage with gemini-embedding-001 support.
- New Images API model: azure/gpt-image-1.
- Fixed intermittent 500 errors and improved traceability for complex chat requests.
-
Flows
- Flows Export/Import Now Includes Agent Configuration
- The export and import functionality for Flows has been enhanced. When exporting or importing a Flow, the configuration of associated agents is now included as part of the process. This improvement ensures that Flows and their agent settings can be seamlessly transferred between projects or environments, making it easier to share, replicate, and maintain complete solutions.
- Slack loading message
- While you interact with Flow through Slack, a customizable loading message appears to indicate your request is being processed.
- Teams Typing Message new Feature
- A new functionality was added in the integration that Expose a Flow through Microsoft Teams. When you are to interact with the AI Solution and you are sending a message in Teams, a typing event shows announcing that the message is being processed.
-
Tools/Agents
- Jira Tool
- With a latest improvement, you can create Epics and Users Stories and link them in Jira Tool.
- Email Tools
- You can send emails using Gmail or SMTP and include a PDF file automatically generated from Markdown content. More information in Email Tools.
-
LLMs
- OpenAI's GPT-5 series
- OpenAI introduces GPT-5, its most advanced AI system to date. GPT-5 represents a significant leap over prior models, delivering state-of-the-art performance across coding, math, writing, health, visual perception, and more.
- Open models by OpenAI
- The gpt-oss family comprises two open-weight reasoning models—gpt-oss-120b, a 117 B-parameter MoE model delivering o4-mini–level performance on core reasoning benchmarks, and gpt-oss-20b, a 21 B-parameter lightweight variant optimized for on-device inference and agentic tasks. Both models are now available in the Beta environment through providers Azure AI Foundry, Groq, OpenRouter and Cerebras.
- Claude Opus 4.1
- This model is a drop-in replacement for Opus 4 that delivers superior performance and precision for real-world coding and agentic tasks. It handles complex, multi-step problems with more rigor and attention to detail. It is already available in the production environment via Anthropic, Vertex AI and AWS Bedrock.
- Vertex AI updates
- DeepSeek-R1-0528, the latest version of the DeepSeek's R1 model, is now available.
- gemini-embedding-001 embeddings support
- A2A (Agent-to-Agent) Protocol Support for Enhanced Integration and Extensibility
- Globant Enterprise AI now supports the A2A (Agent-to-Agent) protocol, enabling seamless integration of Agents defined in other frameworks. With this new feature, you can import external Agents and use them as Tools within Agents created in The Lab. This powerful capability significantly enhances the integration and extensibility of Globant Enterprise AI, allowing organizations to leverage existing investments, connect diverse Agent ecosystems, and build more sophisticated solutions by combining Agents across platforms.
- All Agents Automatically Exposed via A2A Protocol
- All Agents defined in The Lab are now automatically exposed through the A2A protocol, with no additional configuration required. Each Agent is published with an A2A-compliant API, and its capabilities and skills are described in an AgentCard format. The AgentCard is available at a dedicated endpoint, following the A2A standard:
- <GEAI_API_URL>/a2a/<agent-id-or-name>/.well-known/agent.json
- This enhancement allows third-party systems that support A2A to seamlessly discover and interact with Globant Enterprise AI Agents. For more details on the A2A protocol and AgentCard specification, see the official A2A documentation.
- More information: Importing Tools using MCP and A2A Servers
- Workspace
- Shareable Chat Links
- Universal File Upload Compatibility
- Consumers can now upload previously unsupported file formats—such as .doc, .docx, .odt, .rtf, .ppt, and .pptx—directly in the chat interface of Assistants. Even if the selected LLM does not natively support these formats, the platform will automatically convert the files (e.g., to PDF or plain text) at the server level before processing. This enhancement ensures broader file compatibility across both multimodal and non-multimodal models, streamlining interactions and improving consumer experience.
- Lab Improvements
- Agent Export and Import Options
- New features have been introduced: the export and import options in The Lab. These features allow you to easily share Agent definitions and their associated Tools with others, even across different Projects. This enables seamless collaboration and reuse of Agent configurations within Globant Enterprise AI.
- Agent Execution Trace Debugging & Download
- A new feature available in The Lab enhances the Agent testing experience. You can now view detailed execution traces of Agents in a dedicated debug tab while testing. Additionally, there is a new option to download the complete execution log for further analysis or record-keeping.
- New Agent Configuration Parameter: maxRuns
- A new configuration parameter called Max Runs is now available in the AI and Tools Tab of an Agent. This setting defines the maximum number of autonomous iterations an Agent can perform before returning control to the consumer. Each iteration corresponds to a single LLM call, and the default value is set to 5. This allows fine-tuning the level of Agent autonomy based on the complexity and nature of the task.
- Iris: Agents as Tools Support
- A new feature has been added to Iris, enabling Agents created by Iris to use other Agents as Tools. With this update, Agents built with Iris can seamlessly integrate and leverage the capabilities of additional Agents, significantly enhancing their functionality and enabling more complex workflows. This allows Iris-created Agents to delegate tasks or access specialized skills from other Agents, making them more versatile.
- Tools
- Per-User Consent for GDrive Tool Access
- A new consent mechanism will be introduced for Tools that integrate with Google Drive. Before an Agent can access or manipulate an consumer's GDrive data, the consumer must explicitly grant permission. This per-user consent model ensures secure, transparent usage of third-party Tools, aligning with data privacy best practices and organizational compliance requirements. More information: Google Drive Tools – OAuth Authorization Flow
- Expanded Model Support for the Create Image Tool
- The create_image Tool, which can be associated with Agents in The Lab, now supports a wider range of image generation models. Consumer can now generate images using the following models:
- openai/gpt-image-1
- openai/dall-e-3
- vertex_ai/imagen-3.0-generate-001
- vertex_ai/imagen-3.0-fast-generate-001
- vertex_ai/imagen-3.0-generate-002
- xai/grok-2-image-1212
- This expanded support gives consumer greater flexibility and more options for creating images tailored to their specific needs.
- New Public Tool: com.globant.geai.serpapi.google_search
- A new public Tool, com.globant.geai.serpapi.google_search, has been added to The Lab. This web search Tool allows you to query across various Google engines, including Google, Google Maps, Google News, Google Images, Google Videos, and Google Scholar. You can specify which search engine to use in the Agent guidelines or directly in the chat. By default, the standard Google engine is used. This Tool expands the information retrieval capabilities of your Agents, enabling more dynamic and context-aware responses.
- New Public Tools: Firecrawl Web Scraper and Web Search
- Two new public Tools from Firecrawl have been added:
- com.globant.geai.firecrawl.web_scraper
- This Tool allows Agents to fetch content from any web page. It returns page content in multiple formats, including markdown, HTML, links, and screenshots. You can specify one or more formats to retrieve (e.g., markdown, links, screenshot). Additionally, this Tool supports fetching PDF documents from the web.
- com.globant.geai.firecrawl.web_search
- This Tool enables Agents to search web pages and view short snippets from the results. It can be used in combination with the web scraper Tool to extract the full content of selected web pages.
- For detailed configuration steps, see Firecrawl Tools.
- These additions provide Agents with enhanced web browsing and data extraction capabilities, broadening the range of information accessible within Globant Enterprise AI.
- LLM Usage Limit Alerts and Notifications
- A new feature has been added to the Agents and Backoffice - Console to help you manage your LLM usage more effectively. You will now receive warning notifications when your LLM consumption exceeds the configurable alert threshold (soft limit), which can be set per project or as a general cap at the organization level. In addition, if a project or organization runs out of available balance to continue using LLMs, an error notification will be displayed. These alerts enhance visibility and control over LLM usage, helping consumers avoid unexpected interruptions.
- LLMs
- New OpenAI models already available through the Responses API and coming soon through the Chat API:
- o3-pro: Part of OpenAI’s “o” series, this model is trained with reinforcement learning to perform complex reasoning and deliver more accurate answers. o3-pro leverages increased compute to “think before it answers,” consistently providing higher-quality responses.
- codex-mini-latest: This is a fine-tuned version of o4-mini, specifically optimized for use in Codex CLI.
- New Anthropic – Web Search Tool: The web search Tool gives Claude direct access to real-time web content, enabling it to answer questions using up-to-date information beyond its training cutoff. Claude automatically cites sources from search results as part of its response. More details on usage and supported models: How to use LLMs with built-in web search tools via API.
- Claude 4: Anthropic’s latest generation of models, featuring Claude Opus 4 for advanced reasoning and coding, and Claude Sonnet 4 for high-performance, efficient task execution, is now available.
- New Providers Coming to Production: xAI (Grok models) and Cohere.
- Integration of Azure AI Foundry: Azure AI Foundry is being introduced as an LLM provider to leverage its unified platform for building, customizing, and deploying AI applications. This integration provides access to a diverse catalog of over 11,000 models from providers such as OpenAI, xAI, Microsoft, DeepSeek, Meta, Hugging Face, and Cohere, along with robust Tools for responsible AI development and seamless integration with the Azure ecosystem.
- Imagen 4: The Imagen 4 family of models is now available for text-to-image generation through the Images API via Vertex AI. This integration brings Google’s advanced Imagen 4 models—including Standard, Ultra, and Fast variants—for high-quality, brand-consistent image creation with support for multiple languages.
- Model Lifecycle Updates:
- GPT-4.5 Preview Deprecation: Access to GPT-4.5 Preview via the API will end on July 14, 2025. To avoid disruption, this model is being migrated to GPT-4.1.
- Vertex AI Gemini 2.5 Updates: New GA endpoints for Gemini 2.5 Flash (gemini-2.5-flash) and Gemini 2.5 Pro (gemini-2.5-pro) are now available (effective June 17, 2025). Existing preview endpoints for Gemini 2.5 Flash and Pro will be migrated to these new GA endpoints.
- For more information, please refer to Deprecated Models.
- Fixed wrong timeout of 600 seconds when calling assistants. It now uses the provider configured under the parameter HttpTimeout, which defaults to 120s.
- Cohere support for embed-v4.0 embeddings.
- Flows
- Slack Mentions Support for Flows
- Globant Enterprise AI now supports mentions for Flows within Slack. Consumers can add a Flow to a Slack channel and invoke it directly by @-tagging the Flow. This enables seamless initiation and management of conversation threads with Flows straight from Slack. This integration streamlines collaboration and enhances productivity by allowing teams to interact with and trigger Flows without leaving their Slack workspace.
- RAG Revision 9
- New RAG Integration for use from Agents.
- New ingestion properties also valid for omni-parser API.
- New password parameter for processing password-protected PDF files.
- New chunkStrategy parameter to decide how to process tables and images (enabled by default using byLayoutType).
- New chunkSize and chunkOverlap parameters to override the default assistant configuration.
- The Requests log section details the parameters used for ingestion.
- New RAG Document API to better serve associated documents.
- New Multivalued Filter Operators when ingesting with multivalued metadata.
- New Assistants defaults
- Embeddings configuration updated to use cache by default.
- LLM configuration updated from gpt-4o-mini to gpt-4.1-mini.
- ingestion vLLM usage from openai/gpt-4o to openai/gpt-4.1-mini, minor updates on the associated prompts.
- Fixed an issue when handling the threadId (conversation) from the Workspace.
- Fixed an issue where the plugins API did not return the StartPage section.
- Fixed a PayloadTooLargeError error when using a Prompt exceeding 12k tokens.
- Performance improvements when processing embeddings associated with xlsx/csv files.
- Performance improvements when querying Pinecone Vector Store Provider.
- Python SDK Updates and Enhancements
- The Python SDK has been updated with several new features, improvements, and changes to streamline development and agent management. These enhancements make the Python SDK more robust, user-friendly, and supportive of advanced agent development and management workflows.
- Added
- Save and Restore Chat Sessions: You can now save and restore chat sessions using JSON files, making it easier to maintain conversation history.
- Switch Agents in Chat GUI: The chat user interface now allows seamless switching between Agents within an active session.
- API Status Command: The GEAI CLI now includes a status command to check the health of your PyGEA instances.
- Reasoning Strategy in Agent Definition: Agent definitions now support specifying a reasoning strategy for more advanced customization.
- Changed
- Man Pages Installation: The script for installing man pages has been updated to support system-wide installation with the --system flag.
- Comprehensive Help in Man Pages: All help texts are now included in the man pages for the GEAI CLI.
- Simplified Lab Project Selection: The Lab no longer requires explicit project IDs; it now retrieves the project automatically using the provided API key and base URL.
- Support for Model Context Protocol (MCP) to integrate external Tools.
- The GEAI proxy is a Python-based component that enables dynamic integration of external Tools into Globant Enterprise AI via MCP. It acts as a bridge between Globant Enterprise AI and one or more MCP-compliant Tool servers.
- Once the MCP servers are properly configured and connected through the GEAI proxy, the Tools they expose become automatically available in The Lab > Tools of Globant Enterprise AI, ready for use by any Agent without additional setup.
- See more information about this protocol https://modelcontextprotocol.io/introduction.
- See how to import Tools using MCP Tool servers
- New /responses endpoint for AI Interactions
- /responses endpoint was introduced in Responses API, which is fully compatible with the OpenAI Responses API. This addition allows you to submit prompts as plain text, invoke functions, or pass files such as PDFs and images. The endpoint simplifies AI integration by supporting a familiar request/response structure, enabling a smoother transition for teams already using OpenAI-based workflows.
- New Images API
- A new API is available that lets you generate images from text prompts. Supported providers: OpenAI, Vertex AI and xAI.
- LLMs:
- New Gemini models:
- Gemini 2.5 Pro Preview 'I/O edition': Built on its predecessor with significantly enhanced coding abilities and improved reasoning for complex tasks. Designed for developers and advanced users, this edition refines performance across benchmarks and expands its problem-solving reach. Release date: May 6th, 2025.
- Gemini 2.5 Flash: Google's latest model built for complex problem-solving. It allows users to activate thinking and set a thinking budget (1–24k tokens). Designed to balance reasoning and speed, it delivers better performance and accuracy by reasoning before responding.
- Updates in OpenAI's "o" series:
- o3: The most powerful reasoning model in the "o" family; it pushes the frontier across coding, math, science, visual perception, and more.
- o4-mini: A smaller model optimized for fast, cost-efficient reasoning; it achieves remarkable performance for its size and cost, particularly in math, coding, and visual tasks.
- o1-pro: Available through our Responses API, offering a faster, more flexible, and easier way to create agentic experiences.
- Over the next few weeks, the o1‑preview model will be migrated to the new o3 model, while o1‑mini will move to o4‑mini. More info in Deprecated Models.
- Refer to the LLMs with Reasoning Capabilities article for step-by-step guidance on how to use reasoning-enabled models through the API.
- The new GPT-4.1 model series by OpenAI is now available in the production environment, featuring significant improvements in coding, instruction following, and long-context handling—along with their first-ever nano model.
- Grok 3 Model Family added, including two pairs of models:
- Lightweight Variants:
- grok-3-mini-beta and grok-3-mini-fast-beta support function calling and enhanced reasoning (with configurable effort levels) for tasks like meeting scheduling and basic customer support. Both variants deliver identical response quality; the difference lies in response latency, with the "fast" version optimized for quicker responses.
- Flagship Variants:
- grok-3-beta and grok-3-fast-beta are designed for enterprise use cases such as data extraction, coding, and text summarization. They bring deep domain expertise in fields like finance, healthcare, law, and science. Similar to the mini variants, these models have identical capabilities, with the "fast" version offering reduced response times at a higher cost.
- Llama 4 collection by Meta: We continue to expand our coverage of this model family. Recently added Llama 4 Scout and Maverick through Vertex AI's serverless API. Also available in Beta: Llama 4 Maverick via Groq and SambaNova, and Llama 4 Scout through the Cerebras provider, which offers this model with an inference speed of up to 2,600 tokens per second.
- Llama Nemotron Collection: The Llama Nemotron Ultra and Super models are now available in Beta as Nvidia NIM microservices. These are advanced reasoning models, post-trained to optimize performance on tasks such as RAG, tool calling, and alignment with human chat preferences. Both models support a context window of up to 128K tokens.
- Introducing the OpenRouter Provider (Beta):
- OpenRouter joins the GEAI model suite with its Auto Router meta-model, which analyzes each user query and dynamically routes it to the most suitable LLM. This workflow maximizes response quality while minimizing cost and latency, delivering the most efficient output possible.
- Qwen3 Family recently added: The latest generation in the Qwen large language model series, features both dense and mixture-of-experts (MoE) architectures to excel in reasoning, multilingual support, and advanced agent tasks. Its unique ability to switch seamlessly between a thinking mode for complex reasoning and a non-thinking mode for efficient dialogue ensures versatile, high-quality performance. Significantly outperforming prior models like QwQ and Qwen2.5, Qwen3 delivers superior mathematics, coding, commonsense reasoning, creative writing, and interactive dialogue capabilities.
- Better processing of error messages, for example on those cases where the LLMs return specific errors.
- New Python SDK - PyGEAI. It's composed of libraries, tools, code samples, and other documentation that allows developers to interact with the platform more easily with a 16K token context window.
- New omni-parser API to get the content of different file types.
- RAG
- Support for new audio and video formats.
- New endpoints to reindex documents in RAG Assistants API
- New parameters available when geai Ingestion Provider.
- startPage and endPage to selectively process what is needed.
- media parameters such as mediaPrompt, dialogue, frameSamplingRate and so on.
- fix parameter truncate is not supported when calling the cohere-rerank-3.5 model in Rerank API.
- Flows
- File support for Teams & Slack: You can now easily send documents, images, audio, and video files through Teams and Slack when you integrate a Flow into these conversational channels.
- Evaluation Module Enhancements
- New Metrics Introduced:
- Faithfulness: Assesses how factually consistent a response is with the retrieved context.
- Hallucination: Calculated as 1 - Faithfulness, indicating the level of fabricated information.
- Context Precision: Measures the proportion of relevant information within the retrieved contexts, compared against a reference answer for a given user input. (Note: Current calculation does not yet consider the position of retrieved chunks.)
- Noise Sensitivity: This would involve analyzing the relationship between Assistant Accuracy and Context Precision across successive runs of an evaluation plan, varying the number of chunks retrieved. It examines how much and in what way the quality of the generated response changes when irrelevant content is added to the retrieved context.
- The Lab Enhancements
- Flows Integration: The definition and management of Flows are now fully integrated into The Lab.
- Agentic Processes:
- New Conditional Gateway: Introduces the ability to define branching paths based on natural language prompts, enabling dynamic decision-making within processes.
- New Synchronization Gateway: Allows synchronization of multiple parallel paths. The process automatically waits at this point until all incoming paths are completed.
- Enhanced Task Flexibility: Now, it supports multiple inputs and outputs per task, significantly expanding the complexity and richness of the processes you can model.
- Meta-Agent Iris Improvements
- Enhanced LLM Selection Experience: When creating or editing an agent with Iris, users now benefit from a refined LLM selection flow, improving usability and model configuration accuracy.
- The Lab - Custom SSO not supported in this release.
- New The Lab
- The Lab is designed for defining, managing, and orchestrating autonomous Agents. It provides a standardized model for representing Agents, their capabilities, and their interactions within complex workflows. The core components of The Lab include:
- The Lab aims to meet the growing demand for intelligent, self-sufficient Agents capable of collaborating and solving complex problems with minimal human intervention. It offers a flexible and adaptable model, allowing for the creation and management of a diverse range of Agents, from co-pilots working alongside humans to fully automated Agents executing complex tasks. Implemented as a new platform of Globant Enterprise AI, The Lab supports the development of intelligent Agents that can work autonomously or in collaboration with humans and other Agents.
- New features in Flows
- Agent Integration Component: You can now directly integrate Agents created with the The Lab into a Flow. These Agents can be exposed through platforms like WhatsApp, Teams, or Slack.
- File Upload Support from WhatsApp: Flows now support receiving file attachments such as documents, images, audio, and video directly from WhatsApp interactions.
- Audio and Video Attachment in Web Chat: The Web Chat component now allows users to attach audio and video files, enhancing the interaction experience.
- New Features for the Data Analyst Agent
- Reduced Configuration Requirements: The setup needed to enable the Assistant to respond to a wide range of questions has been minimized.
- Enhanced Analysis Module: An additional analysis module has been incorporated to complement the responses with relevant business conclusions and interpretations of the obtained data.
- New metrics to track processed tokens.
- LLMs:
- New Gemini 2.5 Pro (via providers Vertex AI and Gemini): Gemini 2.5 is Google’s latest reasoning model, engineered to tackle increasingly complex challenges. This model is designed for tasks that demand advanced analytical thinking and robust problem-solving capabilities. More details at LLM API.
- Migration to Gemini 2.0 series (Vertex AI): Based on recommendations from Vertex AI, we have migrated from the legacy Gemini 1.0 and 1.5 models to the more advanced Gemini 2.0 series, offering improved performance, scalability, and integration capabilities. For comprehensive information, please refer to the Deprecated Models section.
- New Azure OpenAI models (o1, o1-mini and o3-mini): We have expanded our model availability by introducing these models via Azure, providing the same high-quality capabilities as those offered through the OpenAI provider.
- DeepSeek-R1 via AWS Bedrock: Recently added through a Serverless API, DeepSeek-R1 offers reliable inference with a substantial 128K token context window and up to 32K maximum output tokens.
- OpenAI's new models with built-in web search tool: These specialized models integrate web search capabilities directly into the Chat Completions API, enabling them to both interpret and execute search queries in real time.
- openai/gpt-4o-search-preview
- openai/gpt-4o-mini-search-preview
- New models - Beta only:
- gemini/gemma-3-27b-it: Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models. Gemma 3 has a large, 128K context window, and multilingual support in over 140 languages.
- SambaNova:
- DeepSeek-R1: This Provider offers the fastest performance for running DeepSeek, processing up to 198 tokens per second per user, with a 16K token context window. The model is hosted in US data centers; privately and securely.
- DeepSeek-V3-0324: This model significantly outperforms its predecessor with enhanced reasoning benchmarks, improved code executability and refined web aesthetics, and superior Chinese writing aligned with the R1 style. It also offers better multi-turn interactive rewriting, translation quality, detailed report analysis, and more accurate function calling.
- QwQ-32B: SambaNova provides access to QwQ-32B-Preview, the best open source test-time compute model released by Alibaba.
- Llama 4 collection by Meta:
- Llama 4 Scout: A 17B-parameter multimodal MoE model with 16 experts that excels in text and image understanding. The model is currently in beta and is available via providers Nvidia, Groq and SambaNova. Via Groq, it supports a 128k tokens context window with fast inference at 460 tokens/sec, while SambaNova Cloud runs at 697 tokens/second/user.
- Llama 4 Maverick: Available via Nvidia, this 17-billion-parameter model featuring 128 experts supports a 32k-token context window.
- New LLMs:
- GPT-4.5
- Claude 3.7 Sonnet (Providers Anthropic, Vertex AI and AWS Bedrock)
- Updates in Gemini 2.0 series:
- vertex_ai/gemini-2.0-flash-lite-preview-02-05
- vertex_ai/gemini-2.0-flash-thinking-exp-01-21
- RAG Revision #6
- Support for o3-mini, gpt-4.5-preview, claude-3-7-sonnet-20250219, new DeepSeek, Gemini2* and sambanova LLM providers.
- New pinecone provider available for embeddings and rerankers.
- The CleanUp action message has been corrected to clearly specify that it will permanently delete the RAG Assistant files and update the information in the RDS.
- Added usage element on every response.
- Improvements when changing the LLM/Embeddings settings; all models and providers are normalized to be selected from standard combo-box items; use the override mechanism if you need other options.
- Support for guardrails.
- New documentAggregation property to decide how sources are grouped and returned.
- It is possible to provide feedback on the response of the Chat with Data Assistant in the Frontend.
- The new Evaluation APIs introduce key functionalities through three interconnected APIs: DataSet API, Evaluation Plan API, and Evaluation Result API. This version is primarily designed for users with a data science profile and is mainly accessed via APIs, complemented by a series of Jupyter notebooks that demonstrate their use. For a comprehensive guide on how to use these APIs, you can refer to How to evaluate an AI Assistant and the EvaluationAPITutorial.ipynb notebook, which provide practical examples and code for working through the evaluation process.
- File attachment support in Flows (version 0.9).
- Support for Full Story integration in the Workspace to generate user access statistics in Full Story.
- In the LLM API, for models that have descriptions in the specified languages, the descriptions property is included in the Response, which contains the descriptions in the available languages, such as Spanish, English and Japanese.
- Data Analyst Assistant 2.0 version presents important improvements, simplifying the interaction with the data by reducing the main components to just two: Dispatcher and Thinker. In addition, the metadata structure is automatically generated when loading the datasets, streamlining the setup process. For more information, see How to create a Data Analyst Assistant.
- The option to consult version-specific documentation is now available.
Articles with versions show the option “Other document versions” in the header. Clicking on “Other document versions” brings up a menu that allows you to choose between the most recent version (“Latest”) or earlier versions (e.g. “2025-02 or prior”). If you select a version other than “Latest”, a message appears: “This is not the latest version of this document; to access the latest version, click here”. This message provides a direct link to the most up-to-date documentation.
Components Version Update
- New documentation with details about Supported Chart Types in Chat with Data Assistant.
- New Usage Limits API.
- Flows
- RAG Revision #5
- New endpoint GET /accessControl/apitoken/validate returns information about the organization and project associated with the provided apitoken.
- New LLMs:
- Already in production
- Already in Beta
- DeepSeek:
- deepseek/deepseek-reasoner
- deepseek/deepseek-chat
- azure/deepseek-r1
- nvidia/deepseek-ai-deepseek-r1
- groq/deepseek-r1-distill-llama-70b
- sambanova/DeepSeek-R1-Distill-Llama-70B
- Updates in Gemini 2.0 series:
- gemini-2.0-flash-thinking-exp-01-21 (Via Providers Gemini and Vertex AI)
- gemini/gemini-2.0-flash-lite-preview
- gemini/gemini-2.0-pro-exp
- vertex_ai/gemini-2.0-flash-001
- sambanova/Llama-3.1-Tulu-3-405B
Components Version Update
- Internationalization, Backoffice, and frontend support for Japanese.
- Invitations now include information about the organization and project in the subject.
- New LLMs
- Already in Production
- OpenAI: o1 (2024-12-17 version)
- Already in Beta
- Guardrails configured by assistant.
- Rerank API to semantically order a list of document chunks given a query.
- New optional RAG Retrieve and Rerank adds an extra layer of precision to ensure that only the most relevant information reaches the model used in the generation step.