We’re excited to share new updates from Lyzr Agent Studio.
Conversational Agent Builder
Agents can now be built simply by having a conversation. The conversational agent builder will probe deeply and ask relevant questions to help build the agent in just a few minutes, saving both time and effort compared to going through the manual route.
Voice Agent Updates
Voice agents have received a major upgrade. The entire voice architecture now runs on LiveKit, enabling users to choose between realtime and pipeline models. It also introduces several voice-specific capabilities, including avatars, advanced noise cancellation, custom pronunciation, and access to many new voice models and voices.
Multi-turn evaluation in Simulation Engine
Multi-turn evaluation enables the simulation and evaluation of conversations with multiple back-and-forth exchanges, instead of evaluating a single input-output pair. By analyzing the entire conversation, it delivers more contextual insights and more accurate evaluation results.
Gitagent
A GitAgent is a git repository as an agent. The agent definition lives as files in a repo - versioned, reviewable, and reproducible. You get branching, diffs, pull requests, and collaboration for agent behavior the same way you manage code. You define once, then export to your runtime of choice.
Proxy Agents
Proxy agents let you use Lyzr as a control plane for your external AI frameworks (LangChain, CrewAI, custom agents). Route your LLM calls through Lyzr as a control plane without changing your agent code.
Additional Features:
-
Schedulers and Triggers are now part of the workflow
-
Improved Home Page Screen
-
Introduction of New Image Model of Nano Banana 2
-
Addition of latest models from GPT, Gemini, Anthropic
-
Vision and OCR is now supported at file upload during time of inference
-
Introduced Cross-session memory to retain memory beyond a single user session