v.26.6.9.3
This release introduces new components, integrations, features, and platform enhancements to improve usability, workflow management, AI orchestration, and overall experience in Robility Flow.
Limitations
1. Existing workflow nodes will continue to function. However, updating a component may cause its associated nodes to break.
2. On the Robility Flow project page, the version published to Manager is no longer directly visible. For example, previously it was displayed as sample_1_0_0.
3. Existing published workflows are not compatible with Invoke Flow and Scheduler in Manager. To resolve this, republish the workflow using the Version History feature, or export and import it as a new flow before publishing from Version History.
4. Existing Agent components must be updated to the latest supported version for compatibility with this release. Older versions are not supported in existing workflows.
5. Local Flow Creation Removed: You can no longer create local flows. All new flows, including local flows will be now created directly within a designated project and tenant folder. [Refer here.]
6. Strict Naming Conventions: Flow names are currently case-sensitive, where it allows you to create separate flows with names that differ only by capitalization (e.g., testFlow and testflow are treated as two distinct flows).
What’s New
New Components
Flow introduces new components to improve governance, workflow safety, and execution control across automation flows.
1. Policies (Beta): Establish centralized governance for your workflows with natural language–driven policy enforcement. By defining business rules in plain language, this component automatically converts them into executable guard logic, ensuring secure, compliant, and consistent behavior across your automation flows.
2. Guardrails: Add robust safety controls to your workflows with configurable runtime protections. This component enforces constraints and behavioral boundaries in real time to guide model outputs, reduce unpredictability, and ensure stable, secure execution.
New Integrations
Flow now supports the following integrations designed to enhance AI orchestration capabilities:
1. LiteLLM Proxy: Enables unified access and routing across multiple LLM providers through a single proxy layer, simplifying model management and increasing flexibility in multi-model workflows.
2. vLLM: Delivers high-performance, scalable LLM inference optimized for throughput and latency, enabling faster and more efficient execution of AI-driven workflows.
New Features
Flow introduces features to enhance workflow management, debugging, usability, and overall platform experience.
1. Version History: Acts as a built-in change management system, capturing flows published to Manager. Users can track changes, compare versions, and roll back to previous states when needed.
2. Traces: Provides step-by-step visibility into workflow execution, including inputs, outputs, and latency, helping users debug, monitor, and optimize workflows effectively.
3. Model Provider: Supports centralized model configuration, enabling models to be configured once and reused across components and workflows to reduce duplication and improve consistency.
Enhancements
1. Prompt Template with Mustache Support – It now support {{variable}} placeholders, allowing dynamic values to be inserted into prompts without manual string building or complex logic.
2. Gemini 3 Preview Models – Introduces tool calling for Gemini preview models, enabling advanced AI capabilities across Pro, Flash, and Image variants.
3. Data Operations and DataFrame Operations Renaming
Component names have been updated for better clarity and consistency, with no changes to functionality:
a. Data Operations is now JSON Operations under Processing
b. DataFrame Operations is now Table Operations under Processing
4. Component Control Panel – Redesigned as a unified panel displaying fields and parameters for the selected component. Users can choose which fields appear on the component using + and – icons, while others remain available in the panel for configuration.
Action Required
All agent’s slug names will be reset to the default slug after the agent is updated. Please ensure you take a backup of the flow before proceeding with the update and then rename the slug back to its original name afterward.
Important Note on Agent Retries & Token Usage
By default, the Agent component has retry count enabled. If an agent encounters an error or takes too long to execute, it will automatically retry the process. To prevent unexpected or excessive LLM token consumption, we highly recommend lowering the maximum retry count to (15 maximum) in your agent configurations.
Known Issues
1. The Invoke Flow component does not currently support tool mode. (Tentative fix: June 2026)
2. The Get Row component loses the configured table selection each time the workflow is launched or republished. (Tentative fix: June 2026)
3. The Loop component currently supports only linear execution. Connecting multiple downstream components from the loop output is not supported and may cause the flow to break. (Tentative fix: June 2026)
4. Deleting a project also removes all associated flows, and they cannot be recovered. (Tentative fix: June 2026)
5. If the descriptionfield is removed from the flow JSON before import, the workflow is still uploaded with an empty description. (Tentative fix: June 2026)
6. Flow names are user-specific, allowing multiple users to create flows with the same name, which may result in duplicate solution names and versions. (Tentative fix: June 2026)
7. Newly generated API keys currently apply to all workflow versions. This will be updated to restrict keys to specific versions in future releases. (Tentative fix: June 2026)
8. The Web Scrapercomponent does not support extracting data from JavaScript-based web pages. (Tentative fix: June 2026)
9. The Shareable Playground feature is not functioning as expected. (Tentative fix: June 2026)
Released Date: 09/06/2026
