This update focuses on Agent conversation continuity, reasoning process display and orchestration asset reuse. It adds memory control, reasoning process display configuration, and Agent and workflow resource import capabilities, while continuously optimizing the agent creation, debugging and orchestration workflow to help developers build and reuse lightweight intelligent applications more conveniently.
- Improving long-term memory for Agents to make cross-session conversations more coherent
In intelligent agent applications, conversations are not always completed in one session. Users may continue the same task at different times, and follow-up questions may depend on information already exchanged. If an Agent cannot carry context across sessions, conversation continuity can be affected.
Previously, Agents under the Bot system still lacked long-term memory control, making it difficult to stably carry context in cross-session scenarios. This affected use cases requiring continuous communication and increased developers’ need to configure Agent conversation behavior.
To address this, qKnow Open-Source Edition v2.4.4 adds conversation history and long-term memory switches. Developers can independently control whether to enable memory capabilities according to actual needs. When enabled, Agents can continuously carry context across sessions, supporting long conversation continuity. Developers can also decide whether to enable memory based on application requirements, rather than having memory fixed as a single usage mode.
This update mainly addresses two issues:
🔹 Continuity of conversation context: supporting Agents in carrying previous context for scenarios requiring multi-turn communication and cross-session continuation.
🔹 Autonomous control of memory capabilities: developers can decide whether to enable conversation history and long-term memory based on application requirements.
For intelligent applications requiring continuous interaction, conversation continuity is an important factor affecting user experience. This update provides developers with clearer memory control, though specific effects still depend on actual application scenarios and memory configuration.
- Adding reasoning process display configuration for debugging and demonstration needs
During agent development, debugging and external demonstrations often have different display requirements. When troubleshooting, developers need to understand the Agent’s reasoning process to analyze execution and locate issues. When demonstrating features to external users, the focus is usually on final conversation results, and detailed reasoning processes may not need to be displayed.
If display methods lack flexible control, it is difficult to adapt to both development debugging and external demonstration scenarios. To meet this need, qKnow Open-Source Edition v2.4.4 adds Agent reasoning process display configuration, supporting on-demand enabling or disabling of Agent reasoning process display.
During debugging, developers can enable the display to view the complete reasoning chain and assist in locating issues. During external demonstrations, reasoning process display can be disabled to hide related details and make conversation results more concise.
The focus of this capability is not to change the Agent’s reasoning ability, but to provide a configurable display method for the reasoning process, allowing the same Agent to adjust displayed content according to different usage scenarios. For developers who need to repeatedly debug, validate and demonstrate Agents, this configuration allows debugging analysis and result presentation to adapt to different needs, providing more flexible options for troubleshooting and application demonstrations.
- Supporting Agent and workflow resource import to reduce repetitive orchestration
Intelligent agent application development includes not only single Agent configuration, but may also involve workflows and their orchestration relationships. After completing local debugging, if developers need to migrate existing orchestration results to the platform for continued use, repeated configuration and rebuilding increase workload.
This issue can be summarized as: Agent and workflow orchestration assets completed through local debugging cannot be quickly migrated and reused, with many repetitive configuration processes, increasing the onboarding cost for developers building lightweight intelligent applications. As orchestration assets accumulate during development, making existing results easier to bring into the platform and continue using has become a practical need in open-source development scenarios.
To this end, qKnow Open-Source Edition v2.4.4 adds Agent and workflow resource import capabilities, supporting the import of locally debugged orchestration process assets into the platform to reduce repetitive configuration and rebuilding.
This update focuses on the flow and reuse of existing orchestration assets. Developers can import already debugged Agent or workflow resources into the platform for continued use, without having to rebuild the same orchestration content from scratch every time. Its practical value is mainly reflected in three aspects:
🔹 Reducing repetitive configuration: existing orchestration resources can be imported into the platform, lowering the workload of repetitive building.
🔹 Promoting asset reuse: turning locally debugged orchestration results into resources that can continue to be used.
🔹 Improving development continuity: making resource flow between local debugging and platform use more convenient.
- Continuously optimizing the full-chain agent experience to lower the developer onboarding threshold
In addition to the three specific capabilities above, qKnow Open-Source Edition v2.4.4 also continuously improves the basic experience of the full agent creation, debugging and orchestration process. From a development perspective, building intelligent agent applications is not a single operation but involves creation and configuration, debugging and validation, orchestration and organization, and subsequent adjustments. When the experience in these stages is not sufficiently polished, developers need to spend more time familiarizing themselves with operation processes, which is also not conducive to accumulating and reusing existing orchestration results.
This update optimizes the experience around the full Agent creation, debugging and orchestration process, and combines memory switches, reasoning process display configuration and resource import capabilities to further improve the development experience for lightweight intelligent applications.
These capabilities act on different stages:
🔹 Creation and conversation stage: on-demand control of memory capabilities through conversation history and long-term memory switches.
🔹 Debugging and display stage: adapting to different needs of troubleshooting and external demonstrations through reasoning process display configuration.
🔹 Orchestration and reuse stage: reducing repetitive building of existing orchestration assets through Agent and workflow resource import.
Viewed within the entire development process, these updates are not isolated feature points but respectively respond to practical needs such as conversation continuity, debugging display and orchestration resource reuse. For open-source developers hoping to quickly build lightweight intelligent applications, continuously improving these basic stages helps reduce repetitive operations during development and makes existing capabilities and orchestration results easier to put into subsequent use.
Summary: Continuously improving the agent building experience around actual development needs
The updates in qKnow Intelligent Agent Building Platform Open-Source Edition v2.4.4 mainly focus on three specific issues in agent development and use:
🔹 Lack of memory control for cross-session conversations;
🔹 Different display needs for reasoning processes in debugging and demonstrations;
🔹 Difficulty in quickly reusing locally completed Agent and workflow orchestration assets.
To address these issues, this version adds conversation history and long-term memory switches, Agent reasoning process display configuration, and Agent and workflow resource import capabilities, while continuously optimizing the full-chain experience of agent creation, debugging and orchestration.
From the product iteration direction, these updates focus on supplementing basic capabilities and improving usage processes rather than simply increasing the number of features. For enterprises and developers, building agents requires not only capability configuration but also consideration of whether conversations can continue, whether debugging is convenient, and whether existing orchestration results can be reused.
Going forward, qKnow will continue to focus on the actual usage needs of open-source developers, continuously refine the basic capabilities and user experience in the agent building process, and support the development and iteration of lightweight intelligent applications.
Top comments (0)