LLM Agent Serving

We introduce Orla, a library for constructing and running LLM-based agentic systems. Modern agentic applications consist of workflows that combine multiple LLM inference steps, tool calls, and heterogeneous infrastructure. Today, developers typically build these systems by manually composing orchestration code with LLM serving engines and tool execution logic. Orla provides a general abstraction that separates request execution from workflow-level policy. It acts as a serving layer above existing LLM inference engines: developers define workflows composed of stages, while Orla manages how those stages are mapped, executed, and coordinated across models and backends. It provides agent-level control through three mechanisms: a stage mapper, which assigns each stage to an appropriate model and backend; a workflow orchestrator, which schedules stages and manages their resources and context; and a memory manager, which manages inference state such as the KV cache across workflow boundaries. We demonstrate Orla with a customer support workflow that exercises many of its capabilities. We evaluate Orla on two datasets, showing that stage mapping improves latency and cost compared to a single-model vLLM baseline, while workflow-level cache management reduces time-to-first-token.

We present a technical tutorial for building enterprise-grade realtime voice agents from first principles. While end-to-end speech-to-speech models may ultimately provide the best latency for voice agents, fully self-hosted end-to-end solutions are not yet available. We evaluate the closest candidate, Qwen3-Omni, across three configurations: its cloud-only DashScope Realtime API achieves $\sim$702ms audio-to-audio latency with streaming, but is not self-hostable; its local vLLM deployment supports only the Thinker (text generation from audio, 516ms), not the Talker (audio synthesis); and its local Transformers deployment runs the full pipeline but at $\sim$146s -- far too slow for realtime. The cascaded streaming pipeline (STT $\rightarrow$ LLM $\rightarrow$ TTS) therefore remains the practical architecture for self-hosted realtime voice agents, and the focus of this tutorial. We build a complete voice agent using Deepgram (streaming STT), vLLM-served LLMs with function calling (streaming text generation), and ElevenLabs (streaming TTS), achieving a measured time-to-first-audio of 755ms (best case 729ms) with full function calling support. We release the full codebase as a 9-chapter progressive tutorial with working, tested code for every component.

An agent meets a model as served: through an endpoint with a price card, a shared cache and other tenants, or on whatever hardware a self-hosted model runs. Benchmarks rank the weights. We introduce a Tetris benchmark that measures what agents get from models as served: every move is scored against an oracle, and every agent receives the same pieces. In five pre-specified experiments with nine open-weight models on one serverless provider, plus an open decision model self-hosted on a CPU, we find that price and size do not predict decision quality; that resending history costs almost nothing when cached input is free, would cost eleven to twelve times more if it were not, and makes play worse; that deployments keep between 4 and more than 64 agent contexts warm, in line with their throughput rather than the model's KV-cache size; that an agent's own long requests slow its slowest short decisions more than tenfold, which a simple admission rule cuts by 59% without hurting play; and that the decision model plays mid-pack but takes 27 s per move on a CPU, about 650 times longer than reported on a GPU. Architecture predicts some of what an agent sees; the deployment sets the rest.

Immersive conversational systems in production face a persistent trade-off between responsiveness and long-horizon task capability. Real-time interaction is achievable for lightweight turns, but requests involving planning and tool invocation (e.g., search and media generation) produce heavy-tail execution latency that degrades turn-taking, persona consistency, and user trust. To address this challenge, we propose DuCCAE (Conversation while Collaboration with Augmentation and Evolution), a hybrid engine for immersive conversation deployed within Baidu Search, serving millions of users. DuCCAE decouples real-time response generation from asynchronous agentic execution and synchronizes them via a shared state that maintains session context and execution traces, enabling asynchronous results to be integrated back into the ongoing dialogue. The system orchestrates five subsystems-Info, Conversation, Collaboration, Augmentation, and Evolution-to support multi-agent collaboration and continuous improvement. We evaluate DuCCAE through a comprehensive framework that combines offline benchmarking on the Du-Interact dataset and large-scale production evaluation within Baidu Search. Experimental results demonstrate that DuCCAE outperforms strong baselines in agentic execution reliability and dialogue quality while reducing latency to fit strict real-time budgets. Crucially, deployment metrics since June 2025 confirm substantial real-world effectiveness, evidenced by a tripling of Day-7 user retention to 34.2% and a surge in the complex task completion rate to 65.2%. Our hybrid architecture successfully preserves conversational continuity while enabling reliable agentic execution, offering practical guidelines for deploying scalable agentic systems in industrial settings.

In this work, we identify and address the core challenges of agentic memory management in LLM serving, where large-scale storage, frequent updates, and multiple coexisting agents jointly introduce complex and high-cost approximate nearest neighbor (ANN) searching problems. We present Pancake, a multi-tier agentic memory system that unifies three key techniques: (i) multi-level index caching for single agents, (ii) coordinated index management across multiple agents, and (iii) collaborative GPU-CPU acceleration. Pancake exposes easy-to-use interface that can be integrated into memory-based agents like Mem-GPT, and is compatible with agentic frameworks such as LangChain and LlamaIndex. Experiments on realistic agent workloads show that Pancake substantially outperforms existing frameworks, achieving more than 4.29x end-to-end throughput improvement.

Personal AI agents incur substantial cost via repeated LLM calls. We show existing caching methods fail: GPTCache achieves 37.9% accuracy on real benchmarks; APC achieves 0-12%. The root cause is optimizing for the wrong property -- cache effectiveness requires key consistency and precision, not classification accuracy. We observe cache-key evaluation reduces to clustering evaluation and apply V-measure decomposition to separate these on n=8,682 points across MASSIVE, BANKING77, CLINC150, and NyayaBench v2, our new 8,514-entry multilingual agentic dataset (528 intents, 20 W5H2 classes, 63 languages). We introduce W5H2, a structured intent decomposition framework. Using SetFit with 8 examples per class, W5H2 achieves 91.1%+/-1.7% on MASSIVE in ~2ms -- vs 37.9% for GPTCache and 68.8% for a 20B-parameter LLM at 3,447ms. On NyayaBench v2 (20 classes), SetFit achieves 55.3%, with cross-lingual transfer across 30 languages. Our five-tier cascade handles 85% of interactions locally, projecting 97.5% cost reduction. We provide risk-controlled selective prediction guarantees via RCPS with nine bound families.

LLM-driven agentic applications automate complex, multi-step tasks, but serving them efficiently remains difficult due to heterogeneous components, dynamic model-driven control flow, long-lived state, and highly variable latencies. Nalar is a serving framework for agent workflows that separates workflow specification from execution while providing the runtime visibility and control needed for robust performance. Nalar preserves ordinary Python interfaces and control flow through lightweight auto-generated stubs that turn agent and tool invocations into futures carrying dependency and execution-context metadata. A two-level control architecture combines global policy computation with local event-driven enforcement to support adaptive routing, scheduling, and resource management across evolving workflows. A workflow-aware KV-cache layer enables the runtime to manage cache placement and lifetime. Together, these mechanisms enable scalable, efficient, policy-driven serving of heterogeneous agentic applications without burdening developers with orchestration logic. Across three agentic workloads, Nalar reduces tail latency by 34-74% and achieves up to 3.38x speedups.