# Google Developers Blog

Updates on changes and additions to the Google Developers Blog.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Multi-Turn Agentic Reinforcement Learning with Tunix and Gemma 4 on Google TPUs

DevFeed: [Multi-Turn Agentic Reinforcement Learning with Tunix and Gemma 4 on Google TPUs](<https://devfeed.tech/articles/multi-turn-agentic-reinforcement-learning-with-tunix-and-gemma-4-on-google-tpus-144221.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/multi-turn-agentic-reinforcement-learning-with-tunix-and-gemma-4-on-google-tpus/>)

Author: Wei Wei

Published: 2026-11-19T19:49:56.121412Z

Content type: tutorial

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [grpo](<https://devfeed.tech/topics/grpo.md>), [gemma](<https://devfeed.tech/topics/gemma.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Multi-step AI processes](<https://devfeed.tech/topics/multi-step-ai-processes.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [asynchronous](<https://devfeed.tech/tags/asynchronous.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-tpu](<https://devfeed.tech/tags/cloud-tpu.md>), [gemma](<https://devfeed.tech/tags/gemma.md>), [gemma-4](<https://devfeed.tech/tags/gemma-4.md>), [google](<https://devfeed.tech/tags/google.md>), [group-relative-policy-optimization](<https://devfeed.tech/tags/group-relative-policy-optimization.md>), [grpo](<https://devfeed.tech/tags/grpo.md>), [learning](<https://devfeed.tech/tags/learning.md>), [llm](<https://devfeed.tech/tags/llm.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [tool](<https://devfeed.tech/tags/tool.md>), [tpus](<https://devfeed.tech/tags/tpus.md>), [training](<https://devfeed.tech/tags/training.md>), [using](<https://devfeed.tech/tags/using.md>), [vllm](<https://devfeed.tech/tags/vllm.md>)

### AI overview

The post describes training Gemma 4 E2B as a multi-turn Plane Strike agent with Tunix agentic GRPO on one Cloud TPU v6e-8. The game requires repeated board observations and `strike` tool calls; a custom environment, Gemma 4 chat parser, and conversation agent handle state, native tool responses, and action validation. Asynchronous vLLM rollouts with prefix caching support variable-length turns, while exact token continuity avoids retokenization drift. Starting from a 9.4% held-out win rate and 22.44 average strikes, the model reached 95% wins and about 16 strikes after 100 training steps; tool-call rate approached 100% and invalid actions approached zero. All strikes currently share a trajectory-level reward, which the post identifies as a limitation that finer per-turn credit assignment could address.

### Source excerpt

This post walks developers through training a multi-turn, tool-using LLM agent using Tunix and Gemma 4 on Cloud TPUs. By leveraging agentic Group Relative Policy Optimization (GRPO) and asynchronous vLLM rollouts, the approach significantly improves complex reasoning and tool compliance in interactive environments.

## Build full-stack Android apps with Gemini in Google AI Studio

DevFeed: [Build full-stack Android apps with Gemini in Google AI Studio](<https://devfeed.tech/articles/build-full-stack-android-apps-with-gemini-in-google-ai-studio-138701.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/build-full-stack-android-apps-with-gemini-in-google-ai-studio/>)

Author: Mike Taylor-Cai

Published: 2026-11-18T01:12:09.593393Z

Content type: release

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Firebase](<https://devfeed.tech/topics/firebase.md>), [Google Play](<https://devfeed.tech/topics/google-play.md>), [firebase-ai-logic](<https://devfeed.tech/topics/firebase-ai-logic.md>), [Full Stack Development](<https://devfeed.tech/topics/full-stack-development.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-logic](<https://devfeed.tech/tags/ai-logic.md>), [ai-studio](<https://devfeed.tech/tags/ai-studio.md>), [android](<https://devfeed.tech/tags/android.md>), [android-apps](<https://devfeed.tech/tags/android-apps.md>), [apps](<https://devfeed.tech/tags/apps.md>), [audio](<https://devfeed.tech/tags/audio.md>), [build](<https://devfeed.tech/tags/build.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [complete](<https://devfeed.tech/tags/complete.md>), [debugging](<https://devfeed.tech/tags/debugging.md>), [full](<https://devfeed.tech/tags/full.md>), [full-stack](<https://devfeed.tech/tags/full-stack.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [google](<https://devfeed.tech/tags/google.md>), [google-ai](<https://devfeed.tech/tags/google-ai.md>), [google-ai-studio](<https://devfeed.tech/tags/google-ai-studio.md>), [stack](<https://devfeed.tech/tags/stack.md>), [studio](<https://devfeed.tech/tags/studio.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

Google AI Studio's Build agent can generate Kotlin Android apps from prompts and, when an app needs persistent storage, sign-in, or generative AI, detect those needs and--after user confirmation--provision Firebase and wire the backend into the code. The excerpt names Cloud Firestore, Firebase Authentication, and Firebase AI Logic, and says the agent adds the Firebase BoM and generates Kotlin data and AI layers. Developers can test in an HD emulator with real-time audio, send screenshots to the agent for visual changes, and use custom fonts. Publishing through the integrated Google Play flow builds and signs an Android App Bundle for installation from the user's personal Play library; setting up a Play account is required. These features are described as rolling out, and the excerpt does not establish broader production readiness or availability details.

### Source excerpt

Google AI Studio now lets developers build, test, and ship full-stack Kotlin Android apps from a single prompt--complete with Cloud Firestore, Firebase Authentication, Firebase AI Logic, a faster HD emulator with audio, visual screenshot debugging, and direct publishing to your personal Google Play library.

## Accelerating development for massive Git repositories

DevFeed: [Accelerating development for massive Git repositories](<https://devfeed.tech/articles/accelerating-development-for-massive-git-repositories-138168.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/accelerating-development-for-massive-git-repositories/>)

Author: The Android ABFS Team

Published: 2026-11-17T18:09:46.914988Z

Content type: article

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [aosp](<https://devfeed.tech/topics/aosp.md>), [Git](<https://devfeed.tech/topics/git.md>), [bazel](<https://devfeed.tech/topics/bazel.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [Gerrit](<https://devfeed.tech/topics/gerrit.md>)

Tags: [17](<https://devfeed.tech/tags/17.md>), [2](<https://devfeed.tech/tags/2.md>), [3](<https://devfeed.tech/tags/3.md>), [accelerated](<https://devfeed.tech/tags/accelerated.md>), [acceleration](<https://devfeed.tech/tags/acceleration.md>), [android](<https://devfeed.tech/tags/android.md>), [build](<https://devfeed.tech/tags/build.md>), [builds](<https://devfeed.tech/tags/builds.md>), [cache](<https://devfeed.tech/tags/cache.md>), [content](<https://devfeed.tech/tags/content.md>), [development](<https://devfeed.tech/tags/development.md>), [file](<https://devfeed.tech/tags/file.md>), [filesystem](<https://devfeed.tech/tags/filesystem.md>), [git](<https://devfeed.tech/tags/git.md>), [google](<https://devfeed.tech/tags/google.md>), [open](<https://devfeed.tech/tags/open.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [repositories](<https://devfeed.tech/tags/repositories.md>), [service](<https://devfeed.tech/tags/service.md>), [source](<https://devfeed.tech/tags/source.md>)

### AI overview

ABFS speeds work on large Android repositories by mounting a virtual filesystem that fetches file contents on demand and shares cached build results. Its Linux kernel module, host daemons, and server use content hashes to store Git objects, workspace changes, and build outputs. CASFS exposes file metadata before downloading bytes; its tap utility records files a build reads or changes, and Ninja-go uses content-addressable caching for Ninja builds. The article reports that agents can mount Android release branches in under eight seconds, check out a tree in under ten seconds, and complete a full build in under ten minutes. These results are from runs on a 64-vCPU GCE VM with the ABFS server in the same cloud network. ABFS is open source; Android Soong/Ninja was its first target, with a Bazel proof of concept and other projects exploring compatibility.

### Source excerpt

Meet Another Build File Service (ABFS), Google's open-source content-addressable filesystem and build cache that cuts Android checkouts to <8s and builds to ~2m.

## Unlocking Stable On-Policy RL on TPU with Bitwise-Zero Train-Inference Mismatch

DevFeed: [Unlocking Stable On-Policy RL on TPU with Bitwise-Zero Train-Inference Mismatch](<https://devfeed.tech/articles/unlocking-stable-on-policy-rl-on-tpu-with-bitwise-zero-train-inference-mismatch-136559.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/unlocking-stable-on-policy-rl-on-tpu-with-bitwise-zero-traininference-mismatch/>)

Author: Yuxuan Zhang; Lin Chai; Lance Wang; Tianshu Bao; Weiren Yu

Published: 2026-11-17T05:43:32.994674Z

Content type: article

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>)

Tags: [absolute](<https://devfeed.tech/tags/absolute.md>), [chips](<https://devfeed.tech/tags/chips.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-tpu](<https://devfeed.tech/tags/cloud-tpu.md>), [drift](<https://devfeed.tech/tags/drift.md>), [floating-point](<https://devfeed.tech/tags/floating-point.md>), [inference](<https://devfeed.tech/tags/inference.md>), [jax](<https://devfeed.tech/tags/jax.md>), [learning](<https://devfeed.tech/tags/learning.md>), [llm](<https://devfeed.tech/tags/llm.md>), [policy](<https://devfeed.tech/tags/policy.md>), [recipe](<https://devfeed.tech/tags/recipe.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [rl](<https://devfeed.tech/tags/rl.md>), [stable](<https://devfeed.tech/tags/stable.md>), [tpu](<https://devfeed.tech/tags/tpu.md>), [train](<https://devfeed.tech/tags/train.md>), [using](<https://devfeed.tech/tags/using.md>), [with](<https://devfeed.tech/tags/with.md>), [zero](<https://devfeed.tech/tags/zero.md>)

### AI overview

Zero-TIM aims to eliminate numerical drift between rollout sampling and training in on-policy LLM reinforcement learning. In Tunix, it aligns sampler decode, serving prefill, and trainer logprobs bit for bit by fixing within-shard kernel accumulation and precision, ordering cross-shard reductions, standardizing action logprob scoring, and reusing serving forward kernels with custom differentiation rules. The implementation preserves sampled token IDs across environment turns and constrains excess compiler precision. On a synthetic FrozenLake test using Qwen3-8B across 64 TPU v5p chips, Zero-TIM reported zero mean absolute sampler-trainer logprob difference at each logged step and higher observed solve rates than Standard training and token-level truncated importance sampling; those approaches retained nonzero mismatch, which TIS adjusts for rather than removes. The excerpt presents this as a modular JAX/XLA and Pallas recipe for multi-host TPU stacks, with linear-attention and MoE support still being extended.

### Source excerpt

How we eliminated floating-point train-inference logprob drift for on-policy LLM reinforcement learning across 64 Cloud TPU v5p chips using a generalizable four-part JAX and Pallas recipe in Tunix.

## Securing Your Agents: A Hands-On Guide to Agent Anomaly Detection

DevFeed: [Securing Your Agents: A Hands-On Guide to Agent Anomaly Detection](<https://devfeed.tech/articles/securing-your-agents-a-hands-on-guide-to-agent-anomaly-detection-136557.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/securing-your-agents-a-hands-on-guide-to-agent-anomaly-detection/>)

Author: Achuth Narayan Rajagopal; Derek Luan

Published: 2026-11-17T05:43:32.994674Z

Content type: tutorial

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [agent observability](<https://devfeed.tech/topics/agent-observability.md>), [threat detection](<https://devfeed.tech/topics/threat-detection.md>), [Cloud Monitoring](<https://devfeed.tech/topics/cloud-monitoring.md>), [ai observability](<https://devfeed.tech/topics/ai-observability.md>), [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [Detection engineering](<https://devfeed.tech/topics/detection-engineering.md>), [alert triage](<https://devfeed.tech/topics/alert-triage.md>), [google-adk](<https://devfeed.tech/topics/google-adk.md>), [AI for customer service automation](<https://devfeed.tech/topics/ai-for-customer-service-automation.md>)

Tags: [adk](<https://devfeed.tech/tags/adk.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [alerts](<https://devfeed.tech/tags/alerts.md>), [anomaly-detection](<https://devfeed.tech/tags/anomaly-detection.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-monitoring](<https://devfeed.tech/tags/cloud-monitoring.md>), [configure](<https://devfeed.tech/tags/configure.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-support](<https://devfeed.tech/tags/customer-support.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [detection](<https://devfeed.tech/tags/detection.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [guide](<https://devfeed.tech/tags/guide.md>), [monitor](<https://devfeed.tech/tags/monitor.md>), [monitoring](<https://devfeed.tech/tags/monitoring.md>), [platform](<https://devfeed.tech/tags/platform.md>), [sample](<https://devfeed.tech/tags/sample.md>), [security](<https://devfeed.tech/tags/security.md>), [telemetry](<https://devfeed.tech/tags/telemetry.md>)

### AI overview

Agent Anomaly Detection on the Gemini Platform analyzes agent traces and logs to flag sessions that stray beyond an agent's intended job. The guide demonstrates it with a deliberately over-permissive customer support agent using canned in-memory data: after telemetry and message logging are enabled, a normal session and a simulated payroll-data request are processed, and the latter is flagged for tool misuse. Setup requires a Cloud Logging bucket and Observability dataset in the same US or EU multi-region; monitored sessions are analyzed after setup, which takes about 30 minutes, and new sessions are scanned within about five minutes. The example finding was High severity with 95% probability, but detector results and severities can vary because the analysis models are non-deterministic. Cloud Monitoring metrics support anomaly and coverage alerts; Security Command Center can receive anomaly findings when enabled for the organization. The excerpt cautions against copying the intentionally unsafe demo tool into production.

### Source excerpt

This guide walks developers through setting up Agent Anomaly Detection on the Gemini Enterprise Agent Platform to monitor AI agent telemetry and catch behavioral risks. It demonstrates how to deploy a sample customer support agent, trigger out-of-scope security findings, and configure proactive alerts using Cloud Monitoring and Security Command Center.

## TPU Model Design Considerations and Optimization Guide

DevFeed: [TPU Model Design Considerations and Optimization Guide](<https://devfeed.tech/articles/tpu-model-design-considerations-and-optimization-guide-136558.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/tpu-model-design-considerations-and-optimization-guide/>)

Author: Gagik Amirkhanyan; Vaibhav Singh

Published: 2026-11-17T05:43:32.994674Z

Content type: article

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [model architecture](<https://devfeed.tech/topics/model-architecture.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [distributed-training](<https://devfeed.tech/topics/distributed-training.md>), [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [Programming Languages / Compilers](<https://devfeed.tech/topics/programming-languages-compilers.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [arrays](<https://devfeed.tech/tags/arrays.md>), [design](<https://devfeed.tech/tags/design.md>), [developers](<https://devfeed.tech/tags/developers.md>), [dimensions](<https://devfeed.tech/tags/dimensions.md>), [experts](<https://devfeed.tech/tags/experts.md>), [guide](<https://devfeed.tech/tags/guide.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [llms](<https://devfeed.tech/tags/llms.md>), [mapping](<https://devfeed.tech/tags/mapping.md>), [mixture-of-experts](<https://devfeed.tech/tags/mixture-of-experts.md>), [mixture-of-experts-moe](<https://devfeed.tech/tags/mixture-of-experts-moe.md>), [model](<https://devfeed.tech/tags/model.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [parallel](<https://devfeed.tech/tags/parallel.md>), [performance](<https://devfeed.tech/tags/performance.md>), [principles](<https://devfeed.tech/tags/principles.md>), [rules](<https://devfeed.tech/tags/rules.md>), [systolic-arrays](<https://devfeed.tech/tags/systolic-arrays.md>), [tpu](<https://devfeed.tech/tags/tpu.md>)

### AI overview

The guide's core recommendation is to design model shapes, routing, and parallelism around TPU hardware, then validate changes by wall-clock measurements on a tuned stack. Align contracting and output dimensions to multiples of 256; reshape attention to head_dim 256 where quality permits, since smaller dimensions can leave the MXU underused. Keep dynamic workloads in fixed-shape buffers and represent variable extents with metadata to avoid recompilation. For MoEs, LatentMoE compresses routed tokens before dispatch, reducing expert weight reads and communication; a measured 30B model gained 14% end-to-end training throughput with latent width 768. Larger shared experts and fewer, wider routed experts can improve compute intensity, though routing diversity may suffer and the best granularity should be ablated at target scale. Offload sorting and gather/scatter to SparseCore, fuse vector operations where possible, and map communication-heavy parallel axes to fast physical links. The text stresses that measured rankings and gains depend on configuration; thresholds can change across TPU generations.

### Source excerpt

The newly launched AI-Hypercomputer/tpu-codesign guide provides hardware-grounded design rules to help developers maximize performance on TPUs for LLMs and Mixture-of-Experts (MoE) workloads. Key principles include aligning reduction and output dimensions to 256 for systolic arrays, adopting LatentMoE to compress routed token traffic, and mapping parallel axes directly onto physical links.

## Agents are sessions, not requests: size GKE node pools for AI agent sessions

DevFeed: [Agents are sessions, not requests: size GKE node pools for AI agent sessions](<https://devfeed.tech/articles/agents-are-sessions-not-requests-size-gke-node-pools-for-ai-agent-sessions-136556.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/agents-are-sessions-not-requests-size-gke-node-pools-for-ai-agent-sessions/>)

Author: Drake Williams; Ivan Nardini

Published: 2026-11-17T05:43:32.994674Z

Content type: article

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Gke](<https://devfeed.tech/topics/gke.md>), [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [Agent Harness](<https://devfeed.tech/topics/agent-harness.md>), [long-running](<https://devfeed.tech/topics/long-running.md>), [state-management](<https://devfeed.tech/topics/state-management.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agent-substrate](<https://devfeed.tech/tags/agent-substrate.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ax](<https://devfeed.tech/tags/ax.md>), [cluster](<https://devfeed.tech/tags/cluster.md>), [compute](<https://devfeed.tech/tags/compute.md>), [costs](<https://devfeed.tech/tags/costs.md>), [gke](<https://devfeed.tech/tags/gke.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [node](<https://devfeed.tech/tags/node.md>), [object-storage](<https://devfeed.tech/tags/object-storage.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [park](<https://devfeed.tech/tags/park.md>), [pod](<https://devfeed.tech/tags/pod.md>), [session](<https://devfeed.tech/tags/session.md>), [sessions](<https://devfeed.tech/tags/sessions.md>), [size](<https://devfeed.tech/tags/size.md>), [state](<https://devfeed.tech/tags/state.md>)

### AI overview

Agent sessions often wait most of the workday while retaining state, so allocating one full Pod per session can leave compute idle; scaling Pods to zero loses the running process. Agent Substrate can suspend idle actors to object storage and restore them on a warm worker, while AX defines Tasks and Workspaces and can park a task while preserving /workspace files; its processes start fresh on resume. Parking requires an idle timeout to call the suspend API, and a full worker pool can make requests wait or return HTTP 503. In the article's us-central1 model, 100 coding agents at two prompts per hour need 15 warm-pool nodes plus one control-plane node, with scheduled after-hours scale-down, and parking cuts the node bill by 40%. Savings shrink as prompt rates rise; around 4.5 prompts per hour, one Pod per session is cheaper. Pool sizing, fleet size, and comparable CPU guarantees affect the result, and the model does not price Agent Substrate overcommit.

### Source excerpt

AI agent sessions remain active and idle for most of the workday, making traditional one-Pod-per-session allocations inefficient for cluster resources. By using open-source tools like Agent Substrate and AX on GKE, developers can park idle sessions in object storage to release compute resources while maintaining session state. This approach optimizes resource utilization and reduces Kubernetes node costs by up to 40%.

## How Agent Anomaly Detection scores on benchmarks we didn't write

DevFeed: [How Agent Anomaly Detection scores on benchmarks we didn't write](<https://devfeed.tech/articles/how-agent-anomaly-detection-scores-on-benchmarks-we-didn-t-write-131280.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/how-agent-anomaly-detection-scores-on-benchmarks-we-didnt-write/>)

Author: Nakul Javali

Published: 2026-11-13T06:38:11.586756Z

Content type: news

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [threat detection](<https://devfeed.tech/topics/threat-detection.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [agent observability](<https://devfeed.tech/topics/agent-observability.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [google-adk](<https://devfeed.tech/topics/google-adk.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [anomaly-detection](<https://devfeed.tech/tags/anomaly-detection.md>), [asynchronous](<https://devfeed.tech/tags/asynchronous.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [defense](<https://devfeed.tech/tags/defense.md>), [defense-in-depth](<https://devfeed.tech/tags/defense-in-depth.md>), [detection](<https://devfeed.tech/tags/detection.md>), [instructions](<https://devfeed.tech/tags/instructions.md>), [latency](<https://devfeed.tech/tags/latency.md>), [logs](<https://devfeed.tech/tags/logs.md>), [public](<https://devfeed.tech/tags/public.md>), [security](<https://devfeed.tech/tags/security.md>), [tool](<https://devfeed.tech/tags/tool.md>), [traces](<https://devfeed.tech/tags/traces.md>)

### AI overview

In five public benchmarks totaling 1,774 sessions, Agent Anomaly Detection (AAD) flagged 87% of harmful instructions that agents carried out, analyzing traces and logs asynchronously without adding runtime latency. Detection varied by benchmark (85% on AgentDojo, 87% on InjecAgent, 99% on AgentDyn, and 81% on AgentHarm); across all harmful sessions, including those agents never read or resisted, it flagged 45%-62%. Benign-session false alarms ranged from 2%-4% on office-task injection benchmarks to 16% on AgentHarm, whose benign tasks used high-impact tools. AAD missed most resisted instructions and caught only 4 of 18 cases where poisoned long-term memory led to harmful action. Results used one ADK agent model and simulated apps; production false-alarm rates may differ, and AAD returned no result for about 6% of sessions. The source recommends combining detection with controls that prevent harmful actions.

### Source excerpt

Agent Anomaly Detection (AAD) provides asynchronous, low-latency monitoring for AI agents by analyzing OpenTelemetry traces and logs to flag goal hijacks and tool misuse. Tested across five public security benchmarks, AAD successfully caught 87% of harmful instructions carried out by agents without adding runtime latency. Developers can integrate AAD alongside semantic governance and strict tool limits as part of a defense-in-depth strategy to secure agentic applications.

## Catch Go bugs that fuzzing misses with Gemini and a constraint solver

DevFeed: [Catch Go bugs that fuzzing misses with Gemini and a constraint solver](<https://devfeed.tech/articles/catch-go-bugs-that-fuzzing-misses-with-gemini-and-a-constraint-solver-131279.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/catch-go-bugs-that-fuzzing-misses-with-gemini-and-a-constraint-solver/>)

Author: Karl Weinmeister

Published: 2026-11-13T06:38:11.586756Z

Content type: article

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Fuzzing/Fuzz testing](<https://devfeed.tech/topics/fuzzing.md>), [Go Language](<https://devfeed.tech/topics/go-language.md>), [Static code analysis](<https://devfeed.tech/topics/static-code-analysis.md>)

Tags: [argon](<https://devfeed.tech/tags/argon.md>), [bugs](<https://devfeed.tech/tags/bugs.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [compiler](<https://devfeed.tech/tags/compiler.md>), [constraint](<https://devfeed.tech/tags/constraint.md>), [fuzzing](<https://devfeed.tech/tags/fuzzing.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [gemini-4](<https://devfeed.tech/tags/gemini-4.md>), [gemini-4-argon](<https://devfeed.tech/tags/gemini-4-argon.md>), [go](<https://devfeed.tech/tags/go.md>), [go-test](<https://devfeed.tech/tags/go-test.md>), [pairing](<https://devfeed.tech/tags/pairing.md>), [slice](<https://devfeed.tech/tags/slice.md>), [surviving](<https://devfeed.tech/tags/surviving.md>), [test](<https://devfeed.tech/tags/test.md>), [z3](<https://devfeed.tech/tags/z3.md>)

### AI overview

The article describes a Go bug-finding workflow that compiles functions for RISC-V, identifies bounds-check traps left after compiler optimization, and has Gemini 4 Argon translate their paths into Z3 constraints. Z3 can generate concrete crashing inputs, which `go test` replays against the real code; a `sat` result was reliable in the reported cases, while `unsat` is only suggestive, not a correctness proof. The approach found three planted parser bugs and three latent `skill-lint` bugs. It also exposed why ordinary Go fuzzing missed two panics: slice expressions check capacity, and fuzz inputs often have spare capacity. Clipping fuzz input capacity to length made the fuzzer find both bugs quickly. The solver is best suited to pure functions and helpers; fuzzing and race detection remain useful for broader behavior, and the described single-threaded models do not detect races or deadlocks.

### Source excerpt

How pairing Go's surviving compiler bounds-check traps with Gemini 4 Argon and the Z3 constraint solver catches latent out-of-bounds panics in Go code and uncovers a slice-capacity blind spot in go test -fuzz.

## A2UI 1.0: Build delightful, fast, reliable, agentic experiences

DevFeed: [A2UI 1.0: Build delightful, fast, reliable, agentic experiences](<https://devfeed.tech/articles/a2ui-1-0-build-delightful-fast-reliable-agentic-experiences-130064.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/a2ui-10-build-delightful-fast-reliable-agentic-experiences/>)

Author: Vishal Motwani; Google A2UI Team

Published: 2026-11-12T11:24:02.795824Z

Content type: release

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [a2ui](<https://devfeed.tech/topics/a2ui.md>), [AI for interactive applications](<https://devfeed.tech/topics/ai-for-interactive-applications.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [in-app experiences](<https://devfeed.tech/topics/in-app-experiences.md>)

Tags: [a2ui](<https://devfeed.tech/tags/a2ui.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [android](<https://devfeed.tech/tags/android.md>), [dart](<https://devfeed.tech/tags/dart.md>), [desktop](<https://devfeed.tech/tags/desktop.md>), [generative-ui](<https://devfeed.tech/tags/generative-ui.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [mobile](<https://devfeed.tech/tags/mobile.md>), [open](<https://devfeed.tech/tags/open.md>), [protocol](<https://devfeed.tech/tags/protocol.md>), [python](<https://devfeed.tech/tags/python.md>), [render](<https://devfeed.tech/tags/render.md>), [screens](<https://devfeed.tech/tags/screens.md>), [sdks](<https://devfeed.tech/tags/sdks.md>), [teams](<https://devfeed.tech/tags/teams.md>), [typescript](<https://devfeed.tech/tags/typescript.md>), [web](<https://devfeed.tech/tags/web.md>)

### AI overview

A2UI 1.0 is an open protocol that lets agents describe interactive screens while each host app renders them with its own native components. The excerpt presents it as a way to replace text-heavy exchanges with editable cards, forms, charts, and combined views across apps; actions can wait for user confirmation. It cites Jira follow-up cards, Alation feeds whose charts expose their SQL and data, and Gemini Notebook review cards as examples. A2UI can also run inside MCP Apps. Support differs by product: the excerpt says Lynx supports 0.9 and is adapting to 1.0, Gemini supports 0.9.1, Jetpack Compose 1.0 support is coming, and Gemini Enterprise is being upgraded. SDKs are identified for web, Android, and iOS renderers and Python, TypeScript, and Dart agents.

### Source excerpt

A2UI 1.0 is an open protocol for generative UI that enables AI agents to render interactive, native screens across web, mobile, and desktop applications. Developers can leverage open-source renderer SDKs--including React, Angular, Lit, Flutter, SwiftUI, and Jetpack Compose--alongside agent SDKs for Python, TypeScript, and Dart. The release introduces bidirectional function calls, swappable inference formats, and secure data binding to help teams seamlessly integrate generative UI into their product stacks.

## What is Pathways? Single-controller training for large models

DevFeed: [What is Pathways? Single-controller training for large models](<https://devfeed.tech/articles/what-is-pathways-single-controller-training-for-large-models-125632.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/what-is-pathways-single-controller-training-for-large-models/>)

Author: Sujeeth Jinesh; Shaurya Gupta; Luke Baumann; Monica Song; Ivan Nardini; Akanksha Gupta; Anthony Owusu; Jing Yang; Sadi Kneipp; Yu Li

Published: 2026-11-10T10:39:50.371829Z

Content type: article

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [distributed-training](<https://devfeed.tech/topics/distributed-training.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [LLM Techniques](<https://devfeed.tech/topics/llm-techniques.md>), [post-training](<https://devfeed.tech/topics/post-training.md>)

Tags: [asynchronous](<https://devfeed.tech/tags/asynchronous.md>), [deadlocks](<https://devfeed.tech/tags/deadlocks.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [distributed-training](<https://devfeed.tech/tags/distributed-training.md>), [graph](<https://devfeed.tech/tags/graph.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [scheduler](<https://devfeed.tech/tags/scheduler.md>), [training](<https://devfeed.tech/tags/training.md>), [virtual](<https://devfeed.tech/tags/virtual.md>)

### AI overview

Pathways lets a single JAX client describe a distributed training job as a graph of computation blocks and tensor dependencies. Its runtime maps virtual devices to TPU chips, moves and reshards tensors between groups, dispatches work asynchronously, and centrally orders operations to keep accelerators occupied and avoid conflicting collectives. The excerpt presents this model for workloads such as pipeline training and reinforcement-learning loops, where different programs need to overlap. Virtual-device mapping also allows capacity to be reassigned, healthy chips to take over after a host failure, and compilation to begin before hardware is provisioned. Shared Pathways Service lets clients take slices from a long-running cluster pool; clients wait when all chips are busy. Pathways runs the execution across accelerator groups, while Ray schedules and orchestrates Python workloads at the cluster level and leaves each training job's SPMD execution unchanged; the excerpt describes the two as complementary. The account is limited to the supplied article excerpt.

### Source excerpt

Pathways is a single-controller distributed training system designed to coordinate large-scale model training across multiple accelerator groups by representing computations as a global dataflow graph. Unlike traditional SPMD models like DDP and FSDP that rely on synchronous collectives, Pathways uses asynchronous dispatch and a central scheduler to maximize hardware utilization and prevent deadlocks in complex workloads. By decoupling model definition from physical hardware via virtual devices, developers can efficiently scale complex training architectures and seamlessly handle dynamic workloads like reinforcement learning post-training.

## From prompt to app: Build with Google Maps in AI Studio

DevFeed: [From prompt to app: Build with Google Maps in AI Studio](<https://devfeed.tech/articles/from-prompt-to-app-build-with-google-maps-in-ai-studio-125631.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/from-prompt-to-app-build-with-google-maps-in-ai-studio/>)

Author: Abinaya Selvam; Mike Taylor-Cai

Published: 2026-11-10T10:39:50.371829Z

Content type: release

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Google Maps](<https://devfeed.tech/topics/google-maps.md>), [Google AI](<https://devfeed.tech/topics/google-ai.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [ai-studio](<https://devfeed.tech/tags/ai-studio.md>), [api-key-management](<https://devfeed.tech/tags/api-key-management.md>), [apis](<https://devfeed.tech/tags/apis.md>), [apps](<https://devfeed.tech/tags/apps.md>), [examples](<https://devfeed.tech/tags/examples.md>), [geospatial](<https://devfeed.tech/tags/geospatial.md>), [google-maps](<https://devfeed.tech/tags/google-maps.md>), [prompt](<https://devfeed.tech/tags/prompt.md>), [quotas](<https://devfeed.tech/tags/quotas.md>)

### AI overview

Google AI Studio lets developers prompt Gemini to build apps using Google Maps Platform, with configuration and API key management handled in the studio. The excerpt says core APIs include Maps, Places, and Routes, and describes gallery examples: interactive 3D city tours with custom flyovers; multi-stop fleet routing that accounts for appointment times, priority stops, and driver shift limits; and searchable store locators with distance filters, hours, and branch details. Prototyping is free up to daily quotas, which reset every 24 hours; production requires a paid Cloud billing account.

### Source excerpt

Developers can now build applications powered by Google Maps Platform directly within Google AI Studio by prompting Gemini with real-world geospatial concepts. The integration eliminates traditional configuration hurdles by providing one-click demo keys that generate instantly without requiring Cloud Console redirects or credit cards. This setup grants developers immediate access to core APIs--including Maps, Places, and Routes--supported by daily sandbox quotas for seamless prototyping.

## Agent Distraction

DevFeed: [Agent Distraction](<https://devfeed.tech/articles/agent-distraction-121982.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/agent-distraction/>)

Author: Sam Saccone; Madhuri Garikapati

Published: 2026-11-07T10:15:52.010595Z

Content type: opinion

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [Static code analysis](<https://devfeed.tech/topics/static-code-analysis.md>)

Tags: [2023](<https://devfeed.tech/tags/2023.md>), [acceptance](<https://devfeed.tech/tags/acceptance.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [approach](<https://devfeed.tech/tags/approach.md>), [automated](<https://devfeed.tech/tags/automated.md>), [automatically](<https://devfeed.tech/tags/automatically.md>), [bug](<https://devfeed.tech/tags/bug.md>), [bugs](<https://devfeed.tech/tags/bugs.md>), [call-to-action](<https://devfeed.tech/tags/call-to-action.md>), [canonical](<https://devfeed.tech/tags/canonical.md>), [cleanup](<https://devfeed.tech/tags/cleanup.md>), [coding](<https://devfeed.tech/tags/coding.md>), [developer](<https://devfeed.tech/tags/developer.md>), [friendly](<https://devfeed.tech/tags/friendly.md>), [overhead](<https://devfeed.tech/tags/overhead.md>), [scope](<https://devfeed.tech/tags/scope.md>), [sessions](<https://devfeed.tech/tags/sessions.md>), [time](<https://devfeed.tech/tags/time.md>), [tokens](<https://devfeed.tech/tags/tokens.md>)

### AI overview

The analysis reports that inline TODO(bug/...) comments can pull coding agents away from the developer's task: in the measured sessions, some agents pursued obsolete bugs or expanded scope to address open ones without prompting. These detours added steps, token use, and wait time, and could require human intervention. The authors describe a cleanup process that checked bug status and historical context before submitting confirmed obsolete TODOs for review; they report thousands removed over six months with more than 95% reviewer acceptance. They suggest presubmit checks, corpus scans, and automated cleanup as ways to reduce this source of distraction. The supplied text is an excerpt and does not include the full study results or methods.

### Source excerpt

Analysis of over 913,000 coding agent sessions reveals that 13% expose agents to inline TODO(b/...) comments. In 6% of encounters, agents act on these comments, with 83% being entirely unprompted by the developer. These detours result in significant overhead--adding 70% more steps, 77% more tokens, and 105% more wait time per session--split between chasing closed tickets (41%) and unauthorized scope creep on open bugs (57%). Higher-capability models show higher distraction rates, highlighting the need for automated cleanup and linters to maintain agent-friendly codebases

## Cost per task: Using cheap classifiers to route across Gemini models and thinking levels

DevFeed: [Cost per task: Using cheap classifiers to route across Gemini models and thinking levels](<https://devfeed.tech/articles/cost-per-task-using-cheap-classifiers-to-route-across-gemini-models-and-thinking-levels-116088.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/cost-per-task-using-cheap-classifiers-to-route-across-gemini-models-and-thinking-levels/>)

Author: Sergio Villani

Published: 2026-11-04T10:13:45.411814Z

Content type: article

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [cloud cost management](<https://devfeed.tech/topics/cloud-cost-management.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [agentic-coding](<https://devfeed.tech/topics/agentic-coding.md>), [google-antigravity](<https://devfeed.tech/topics/google-antigravity.md>)

Tags: [classifiers](<https://devfeed.tech/tags/classifiers.md>), [coding-agents](<https://devfeed.tech/tags/coding-agents.md>), [cost](<https://devfeed.tech/tags/cost.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [models](<https://devfeed.tech/tags/models.md>), [route](<https://devfeed.tech/tags/route.md>), [thinking](<https://devfeed.tech/tags/thinking.md>), [token](<https://devfeed.tech/tags/token.md>)

### AI overview

In a single-run, 50-task coding-agent benchmark, Gemini 3.8 Flash on low thinking solved all tasks and cost less than default thinking; a Jev router that tried Flash-Lite first and escalated complex tasks also solved all 50 and reduced cost and response time versus low thinking on every task. Jev classified 48 of 50 tasks correctly overall, while a frozen keyword-and-length heuristic scored 70.6% on 34 unseen tasks. Broad simple/complex questions underperformed; one atomic yes/no Noul question improved routing. A low output-token cap did not reduce thinking on easy tasks and caused seven hard-task failures. The authors recommend starting with low thinking, then considering Jev for short, standalone tasks. Routing across models can undermine model-specific context-cache discounts, so they advise keeping long, context-heavy turns on one model. Results are limited to one run, 50 hand-labeled prompts, and simple automatic grading; costs and latency may vary.

### Source excerpt

Per-token API prices keep falling, yet engineering bills for autonomous coding agents keep climbing....

## DevFest is back!

DevFeed: [DevFest is back!](<https://devfeed.tech/articles/devfest-is-back-109129.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/devfest-is-back/>)

Author: Natalie McHugh

Published: 2026-10-30T18:19:51.784812Z

Content type: news

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Developer Events](<https://devfeed.tech/topics/developer-events.md>), [Learning](<https://devfeed.tech/topics/learning.md>), [Google AI](<https://devfeed.tech/topics/google-ai.md>), [developer-relations](<https://devfeed.tech/topics/developer-relations.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [about](<https://devfeed.tech/tags/about.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [ai](<https://devfeed.tech/tags/ai.md>), [build](<https://devfeed.tech/tags/build.md>), [conference](<https://devfeed.tech/tags/conference.md>), [developer](<https://devfeed.tech/tags/developer.md>), [developers](<https://devfeed.tech/tags/developers.md>), [devfest](<https://devfeed.tech/tags/devfest.md>), [era](<https://devfeed.tech/tags/era.md>), [events](<https://devfeed.tech/tags/events.md>), [global](<https://devfeed.tech/tags/global.md>), [google](<https://devfeed.tech/tags/google.md>), [google-ai](<https://devfeed.tech/tags/google-ai.md>), [google-s](<https://devfeed.tech/tags/google-s.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [october](<https://devfeed.tech/tags/october.md>), [scale](<https://devfeed.tech/tags/scale.md>), [web](<https://devfeed.tech/tags/web.md>), [workshops](<https://devfeed.tech/tags/workshops.md>)

### AI overview

DevFest 2026 runs October 1-December 31 across more than 800 events in 115 countries, with nearly one million attendees expected. Hosted by local Google Developer Groups in regional languages, events offer interactive sessions and workshops on Google AI technologies and tracks for Google Cloud, Android, and Web. The theme is building, securing, and scaling in the agentic era; the series welcomes learners at all levels.

### Source excerpt

DevFest 2026 runs from October 1 to December 31, bringing together nearly a million developers across more than 800 global events centered on the agentic AI era. Under the theme "Build, Secure, Scale," participants gain hands-on experience with Google's developer stack--including Gemini, Google AI Studio, Google Antigravity, and Web MCP--through live workshops and codelabs. The conference focuses on enabling developers to prototype rapidly, ensure secure and private deployments, and leverage production-ready infrastructure.

## Build Generative UI Apps with A2UI & CopilotKit Agents

DevFeed: [Build Generative UI Apps with A2UI & CopilotKit Agents](<https://devfeed.tech/articles/build-generative-ui-apps-with-a2ui-copilotkit-agents-107151.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/build-generative-ui-apps-with-a2ui-copilotkit-agents/>)

Author: Google A2UI Team; Atai Barkai

Published: 2026-10-30T04:35:29.705865Z

Content type: article

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [a2ui](<https://devfeed.tech/topics/a2ui.md>), [google-adk](<https://devfeed.tech/topics/google-adk.md>), [LangChain](<https://devfeed.tech/topics/langchain.md>)

Tags: [a2ui](<https://devfeed.tech/tags/a2ui.md>), [adk](<https://devfeed.tech/tags/adk.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agents](<https://devfeed.tech/tags/agents.md>), [apps](<https://devfeed.tech/tags/apps.md>), [aws](<https://devfeed.tech/tags/aws.md>), [build](<https://devfeed.tech/tags/build.md>), [generative](<https://devfeed.tech/tags/generative.md>), [generative-ui](<https://devfeed.tech/tags/generative-ui.md>), [google](<https://devfeed.tech/tags/google.md>), [google-adk](<https://devfeed.tech/tags/google-adk.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [langchain](<https://devfeed.tech/tags/langchain.md>), [learn](<https://devfeed.tech/tags/learn.md>), [learn-how-to](<https://devfeed.tech/tags/learn-how-to.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [render](<https://devfeed.tech/tags/render.md>), [sdks](<https://devfeed.tech/tags/sdks.md>), [ui](<https://devfeed.tech/tags/ui.md>), [using](<https://devfeed.tech/tags/using.md>)

### AI overview

A2UI lets agents compose interfaces from a frontend's declared component catalog, using a framework-agnostic format that AG-UI can stream to the client. CopilotKit's documented setup connects an AG-UI-compatible agent to its runtime, defines component descriptions and Zod prop schemas, maps those definitions to frontend renderers, and passes the catalog to the provider. The Basic Catalog offers 18 general-purpose primitives; the text says production apps commonly define custom catalogs. A2UI also supports more controlled or open-ended generation, with tradeoffs: predefined components constrain customization, while generated code can drift from the design system and runs in a sandboxed iframe. For interactive flows, inputs bind to data paths and buttons send selected values back over the same stream; client-side validation can show errors without a model call. The article gives Google ADK as a setup example and says other AG-UI-compatible frameworks use the same pattern.

### Source excerpt

See Google A2Ui in action! Learn how to natively render agentic UIs across Langchain, AWS Strands, Microsoft agents, and Google ADK using CopilotKit SDKs for A2UI.

## Smaller apps and faster GPU inference with LiteRT in Google Play services

DevFeed: [Smaller apps and faster GPU inference with LiteRT in Google Play services](<https://devfeed.tech/articles/smaller-apps-and-faster-gpu-inference-with-litert-in-google-play-services-105743.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/smaller-apps-and-faster-gpu-inference-with-litert-in-google-play-services/>)

Author: Google AI Edge Team; Android ML Team

Published: 2026-10-29T03:38:31.321833Z

Content type: release

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [LiteRT](<https://devfeed.tech/topics/litert.md>)

Tags: [about](<https://devfeed.tech/tags/about.md>), [android](<https://devfeed.tech/tags/android.md>), [android-studio](<https://devfeed.tech/tags/android-studio.md>), [api](<https://devfeed.tech/tags/api.md>), [app](<https://devfeed.tech/tags/app.md>), [apps](<https://devfeed.tech/tags/apps.md>), [can](<https://devfeed.tech/tags/can.md>), [dependencies](<https://devfeed.tech/tags/dependencies.md>), [developers](<https://devfeed.tech/tags/developers.md>), [faster](<https://devfeed.tech/tags/faster.md>), [google](<https://devfeed.tech/tags/google.md>), [google-play](<https://devfeed.tech/tags/google-play.md>), [google-play-services](<https://devfeed.tech/tags/google-play-services.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-inference](<https://devfeed.tech/tags/gpu-inference.md>), [inference](<https://devfeed.tech/tags/inference.md>), [litert](<https://devfeed.tech/tags/litert.md>), [play](<https://devfeed.tech/tags/play.md>), [services](<https://devfeed.tech/tags/services.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

LiteRT in Google Play services adds GPU inference through the CompiledModel API, which the source says ran about 1.4x faster than the legacy GPU delegate in its testing. Moving from bundled LiteRT can reduce app size by about 8 MB because Play services supplies the runtime and GPU accelerator. Developers can add `com.google.android.gms:play-services-litert:17.0.0` or later, remove the bundled `litert` dependency, initialize with GPU enabled, and pass the environment to CompiledModel. Existing Interpreter code continues to work on CPU; GPU use requires moving to CompiledModel. The Android Studio LiteRT skill can automate migration, but the source recommends reviewing its changes, building, and comparing outputs on a real device. Legacy GPU and custom delegate APIs are deprecated, NNAPI removal begins in 2027, and deprecated API code will not compile against the latest SDK.

### Source excerpt

LiteRT in Google Play services now offers GPU inference powered by ML Drift, delivering about 1.4x faster performance alongside the new CompiledModel API. Switching from bundled binaries reduces app size by approximately 8 MB while enabling seamless, automatic runtime updates across billions of devices. Developers can easily adopt these updates by updating their dependencies and utilizing the new Android Studio LiteRT skill for automated migration.

## Hill-Climbing Agent Quality with Agents CLI Evaluation

DevFeed: [Hill-Climbing Agent Quality with Agents CLI Evaluation](<https://devfeed.tech/articles/hill-climbing-agent-quality-with-agents-cli-evaluation-105742.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/hill-climbing-agent-quality-with-agents-cli-evaluation/>)

Author: Pier Paolo Ippolito; Elia Secchi

Published: 2026-10-29T03:38:31.321833Z

Content type: tutorial

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [agent observability](<https://devfeed.tech/topics/agent-observability.md>), [google-adk](<https://devfeed.tech/topics/google-adk.md>), [Test-driven development](<https://devfeed.tech/topics/tdd.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [alerting](<https://devfeed.tech/tags/alerting.md>), [automate](<https://devfeed.tech/tags/automate.md>), [automated](<https://devfeed.tech/tags/automated.md>), [backend](<https://devfeed.tech/tags/backend.md>), [behaviors](<https://devfeed.tech/tags/behaviors.md>), [cli](<https://devfeed.tech/tags/cli.md>), [comparisons](<https://devfeed.tech/tags/comparisons.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [developers](<https://devfeed.tech/tags/developers.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [model](<https://devfeed.tech/tags/model.md>), [nightly](<https://devfeed.tech/tags/nightly.md>), [performance](<https://devfeed.tech/tags/performance.md>), [prompt](<https://devfeed.tech/tags/prompt.md>), [quality](<https://devfeed.tech/tags/quality.md>), [testing](<https://devfeed.tech/tags/testing.md>), [trace](<https://devfeed.tech/tags/trace.md>)

### AI overview

Agents CLI supports an evaluation loop that runs agent cases, grades traces, compares results per case, and analyzes failures; ADK projects can also use eval optimize to propose prompt variants. In the example, changing a weather agent's system prompt to a chef persona lowered the mean score from 4.0 to 2.75 across four cases, with per-case comparison showing missed tool use and off-topic recipe suggestions. A model swap produced the same mean drop but concentrated the loss in one recent-events case, showing why averages alone can mislead. The article recommends a small representative dataset for frequent checks, retaining full result files, and separating test inputs from grading expectations. Nightly runs can use cron or Cloud Scheduler plus a threshold script that exits nonzero to trigger alerts; Agents CLI itself does not include a scheduler. Prompt optimization is limited to prompts, is not supported for the LangChain extension, and requires human review because candidates may overfit the metric. The examples use a fixed judge model; model comparisons should account for prompt tuning and keep judge changes separate.

### Source excerpt

This post walks through the Agents CLI evaluation loop for systematically improving agent performance through continuous testing and grading. It demonstrates how to catch prompt regressions, evaluate model backend swaps using side-by-side trace comparisons, and automate evaluation passes. Developers will also learn how to integrate nightly scheduled runs with threshold-based alerting to track quality over time.

## Building Conversational Commerce with Multi-Agent Swarms on GCP: A Working Prototype

DevFeed: [Building Conversational Commerce with Multi-Agent Swarms on GCP: A Working Prototype](<https://devfeed.tech/articles/building-conversational-commerce-with-multi-agent-swarms-on-gcp-a-working-prototype-102583.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/building-conversational-commerce-with-multi-agent-swarms-on-gcp-a-working-prototype/>)

Author: Prateek Khushalani

Published: 2026-10-27T21:39:39.465576Z

Content type: tutorial

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [Conversational AI](<https://devfeed.tech/topics/conversational-ai.md>), [supervisor agent pattern](<https://devfeed.tech/topics/supervisor-agent-pattern.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [cloud pub/sub](<https://devfeed.tech/topics/cloud-pub-sub.md>), [Chat Bot](<https://devfeed.tech/topics/chatbot.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [3](<https://devfeed.tech/tags/3.md>), [403](<https://devfeed.tech/tags/403.md>), [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [building](<https://devfeed.tech/tags/building.md>), [commerce](<https://devfeed.tech/tags/commerce.md>), [gcp](<https://devfeed.tech/tags/gcp.md>), [google](<https://devfeed.tech/tags/google.md>), [key](<https://devfeed.tech/tags/key.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [post](<https://devfeed.tech/tags/post.md>), [prototype](<https://devfeed.tech/tags/prototype.md>), [required](<https://devfeed.tech/tags/required.md>), [structure](<https://devfeed.tech/tags/structure.md>), [this](<https://devfeed.tech/tags/this.md>), [user](<https://devfeed.tech/tags/user.md>), [using](<https://devfeed.tech/tags/using.md>), [with](<https://devfeed.tech/tags/with.md>), [workflows](<https://devfeed.tech/tags/workflows.md>)

### AI overview

HappyDelivery demonstrates a prototype for asynchronous conversational shopping on GCP: a FastAPI ingress verifies webhook signatures, normalizes messages and queues them in Pub/Sub; a Redis session store restores carts and preferences; and a LangGraph swarm routes work among profile, inventory and transaction agents using Gemini. Deterministic Python tools calculate prices, tax and fees, while tokenized customer IDs keep raw card numbers out of prompts and logs. The text reports a 24-hour Redis TTL and 0.318 ms average lookup, but five live Cloud Run webhook posts took about 267-302 ms, despite the stated under-150-ms goal. It also describes a three-turn order simulation and reports five automated tests passing. These are prototype results and claims, not evidence of production deployment.

### Source excerpt

This document provides a publication template and metadata structure for a technical blog post focused on building Conversational User Interfaces (CUIs) using AI. It outlines key developer focus areas, product stacks, and approval workflows required for publishing technical content through Google Developer Studio. The framework helps engineers organize their guides and solutions to effectively share AI-driven conversational interface designs with the developer community.

## Fast, reliable A2UI, with swappable inference formats

DevFeed: [Fast, reliable A2UI, with swappable inference formats](<https://devfeed.tech/articles/fast-reliable-a2ui-with-swappable-inference-formats-102584.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/fast-reliable-a2ui/>)

Author: Vishal Motwani; Google A2UI Team

Published: 2026-10-27T21:39:39.465576Z

Content type: release

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [a2ui](<https://devfeed.tech/topics/a2ui.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [a2ui](<https://devfeed.tech/tags/a2ui.md>), [agent](<https://devfeed.tech/tags/agent.md>), [component](<https://devfeed.tech/tags/component.md>), [dart](<https://devfeed.tech/tags/dart.md>), [decision-models](<https://devfeed.tech/tags/decision-models.md>), [express](<https://devfeed.tech/tags/express.md>), [formats](<https://devfeed.tech/tags/formats.md>), [generation](<https://devfeed.tech/tags/generation.md>), [generative](<https://devfeed.tech/tags/generative.md>), [generative-ui](<https://devfeed.tech/tags/generative-ui.md>), [inference](<https://devfeed.tech/tags/inference.md>), [jev](<https://devfeed.tech/tags/jev.md>), [json](<https://devfeed.tech/tags/json.md>), [latency](<https://devfeed.tech/tags/latency.md>), [models](<https://devfeed.tech/tags/models.md>), [python](<https://devfeed.tech/tags/python.md>), [sdks](<https://devfeed.tech/tags/sdks.md>), [standard](<https://devfeed.tech/tags/standard.md>), [stream](<https://devfeed.tech/tags/stream.md>), [style](<https://devfeed.tech/tags/style.md>)

### AI overview

A2UI Express speeds generative UI by letting models produce compact, Python-style statements that the Agent SDK compiles into validated A2UI JSON using the catalog schema. The article reports up to 3x fewer tokens than direct JSON and says Express works well with thinking off; it is supported in the Python, TypeScript, and Dart SDKs, with no frontend renderer or catalog changes required. For screens with layouts defined in advance, TypeSafe's Jev decision model can choose structure through typed questions while a smaller model writes text. In an 11-prompt Apparite example, the component tree appeared in 92-298 ms and screens completed in 0.8-1.7 seconds, but the authors describe this as an early, small-sample result. Controlled generation is limited to layouts covered by its grammar; Express retains A2UI's broader expressivity.

### Source excerpt

A2UI now supports swappable inference formats, introducing A2UI Express which uses up to 3x fewer tokens and cuts latency by compiling compact Python-style syntax into standard A2UI JSON via Agent SDKs. Developers can also use controlled generation with TypeSafe's Jev system-decision models to render known-shape screens in under 300 ms while smaller models stream text. These updates require no changes to frontend renderers or component catalogs, offering flexible performance levers for generative UI development across Python, TypeScript, and Dart.

## Preview, edit, and collaborate on markdown (.md) files natively in Google Docs and Drive

DevFeed: [Preview, edit, and collaborate on markdown (.md) files natively in Google Docs and Drive](<https://devfeed.tech/articles/preview-edit-and-collaborate-on-markdown-md-files-natively-in-google-docs-and-drive-100799.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/preview-edit-and-collaborate-on-markdown-md-files-natively-in-google-docs-and-drive/>)

Author: Sophia Siao; Chelsea Rice

Published: 2026-10-26T22:43:36.656676Z

Content type: release

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Markdown](<https://devfeed.tech/topics/markdown.md>), [Documentation](<https://devfeed.tech/topics/documentation.md>), [Google](<https://devfeed.tech/topics/google.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-tools](<https://devfeed.tech/tags/ai-tools.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [bridge](<https://devfeed.tech/tags/bridge.md>), [bridging](<https://devfeed.tech/tags/bridging.md>), [code](<https://devfeed.tech/tags/code.md>), [developer](<https://devfeed.tech/tags/developer.md>), [developer-tools](<https://devfeed.tech/tags/developer-tools.md>), [developers](<https://devfeed.tech/tags/developers.md>), [docs](<https://devfeed.tech/tags/docs.md>), [documentation](<https://devfeed.tech/tags/documentation.md>), [drive](<https://devfeed.tech/tags/drive.md>), [git](<https://devfeed.tech/tags/git.md>), [google-workspace](<https://devfeed.tech/tags/google-workspace.md>), [integration](<https://devfeed.tech/tags/integration.md>), [markdown](<https://devfeed.tech/tags/markdown.md>), [native](<https://devfeed.tech/tags/native.md>), [support](<https://devfeed.tech/tags/support.md>), [tools](<https://devfeed.tech/tags/tools.md>)

### AI overview

Google Docs now supports viewing, editing, and collaborating on Markdown files without converting them, preserving the underlying format for developer tools and Git repositories. Google Drive can show rendered previews for formatting review. The source says anyone with a Google or Workspace account can start by uploading an .md file to Drive and opening it in Docs. It gives no further details about collaboration features or limitations.

### Source excerpt

Google Workspace now offers native support for Markdown (.md) files, allowing developers and AI agents to seamlessly collaborate on documentation directly in Google Docs without manual file conversions. This integration preserves native file fidelity for use with Git repositories and developer tools while unlocking standard Docs features like line-level commenting, rendered Drive previews, and tracked version history.

## Architecting Governed, Self-Evolving AI Agents on Gemini Agent Platform

DevFeed: [Architecting Governed, Self-Evolving AI Agents on Gemini Agent Platform](<https://devfeed.tech/articles/architecting-governed-self-evolving-ai-agents-on-gemini-agent-platform-95939.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/architecting-governed-self-evolving-ai-agents-on-gemini-agent-platform/>)

Author: Anant Nawalgaria

Published: 2026-10-22T16:44:19.101536Z

Content type: article

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [google-adk](<https://devfeed.tech/topics/google-adk.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [Google Cloud Platform (GCP)](<https://devfeed.tech/topics/google-cloud.md>), [human in the loop agent](<https://devfeed.tech/topics/human-in-the-loop-agent.md>), [BigQuery](<https://devfeed.tech/topics/bigquery.md>), [supervisor agent pattern](<https://devfeed.tech/topics/supervisor-agent-pattern.md>), [agent orchestration](<https://devfeed.tech/topics/agent-orchestration.md>), [hierarchical agent systems](<https://devfeed.tech/topics/hierarchical-agent-systems.md>)

Tags: [adk](<https://devfeed.tech/tags/adk.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agent-development-kit](<https://devfeed.tech/tags/agent-development-kit.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [authorization](<https://devfeed.tech/tags/authorization.md>), [bigquery](<https://devfeed.tech/tags/bigquery.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [defense](<https://devfeed.tech/tags/defense.md>), [defense-in-depth](<https://devfeed.tech/tags/defense-in-depth.md>), [development](<https://devfeed.tech/tags/development.md>), [development-kit](<https://devfeed.tech/tags/development-kit.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [gate](<https://devfeed.tech/tags/gate.md>), [gemini](<https://devfeed.tech/tags/gemini.md>), [google](<https://devfeed.tech/tags/google.md>), [google-cloud](<https://devfeed.tech/tags/google-cloud.md>), [kms](<https://devfeed.tech/tags/kms.md>), [platform](<https://devfeed.tech/tags/platform.md>)

### AI overview

Project Chrysalis describes a simulated Google Cloud architecture for agents that diagnose cache-stampede outages, synthesize rate-limiting middleware, assess interventions, and retain operational lessons. In the walkthrough, a token-bucket patch passed AST sandbox, OPA policy, and Cloud KMS/Binary Authorization checks before mounting; a proposed permanent agent-graph change paused for dual-custody SRE approval. BigQuery replay compared 400 traffic policies, while step-level attribution penalized a database restart that raised latency and credited the patch. The simulation reports recovery from 72.7-80.0% HTTP 500 errors to 0%, 18.2 ms P99 latency, and a 0.5 ms replay. These are claims about the described demo, not evidence of production deployment or independently verified performance.

### Source excerpt

Learn how Project Chrysalis combines Google Cloud's Gemini Enterprise Agent Platform, Agent Development Kit (ADK), BigQuery, and a four-gate defense-in-depth security stack (Model Armor, Cloud KMS, Binary Authorization, and OPA) to build governed, self-evolving AI agents that remediate live production outages in 1.8 seconds.

## Streaming comes to API Gateway: real-time and AI-ready APIs, without changing your spec

DevFeed: [Streaming comes to API Gateway: real-time and AI-ready APIs, without changing your spec](<https://devfeed.tech/articles/streaming-comes-to-api-gateway-real-time-and-ai-ready-apis-without-changing-your-spec-95940.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/streaming-comes-to-api-gateway-real-time-and-ai-ready-apis-without-changing-your-spec/>)

Author: Mak Ahmad; Geir Sjurseth; Jeremy Whitlock

Published: 2026-10-22T16:44:19.101536Z

Content type: release

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [Real-time AI interactions](<https://devfeed.tech/topics/real-time-ai-interactions.md>), [API Governance](<https://devfeed.tech/topics/api-governance.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [Event-Streaming](<https://devfeed.tech/topics/event-streaming.md>), [rate-limiting](<https://devfeed.tech/topics/rate-limiting.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [api](<https://devfeed.tech/tags/api.md>), [api-gateway](<https://devfeed.tech/tags/api-gateway.md>), [apis](<https://devfeed.tech/tags/apis.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [environment](<https://devfeed.tech/tags/environment.md>), [gateway](<https://devfeed.tech/tags/gateway.md>), [grpc](<https://devfeed.tech/tags/grpc.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [infrastructure-as-code](<https://devfeed.tech/tags/infrastructure-as-code.md>), [observability](<https://devfeed.tech/tags/observability.md>), [preview](<https://devfeed.tech/tags/preview.md>), [public](<https://devfeed.tech/tags/public.md>), [rate-limiting](<https://devfeed.tech/tags/rate-limiting.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [route](<https://devfeed.tech/tags/route.md>), [spec](<https://devfeed.tech/tags/spec.md>), [streaming](<https://devfeed.tech/tags/streaming.md>), [streams](<https://devfeed.tech/tags/streams.md>)

### AI overview

API Gateway Public Preview lets streaming endpoints use the same gateway policies as regular APIs, avoiding a separate, ungoverned path. It supports long-lived streams up to one hour, including incremental responses for model routing to Gemini, Claude, and OpenAI gpt-oss. Existing API specifications need no changes; during Preview, enable streaming when creating a gateway with `--enable-streaming`. Clients should implement reconnection logic for long-lived connections. The update also makes default hostnames deterministic from gateway ID, project number, and region, so teams can template endpoints before creating resources and reuse the same URL after rebuilding. The excerpt says streaming is planned to require no opt-in at GA. It positions API Gateway as a lightweight AI gateway entry point, while recommending Apigee for broader AI policies and lifecycle management, and Agent Gateway in Gemini platform for agent connectivity governance.

### Source excerpt

API Gateway now supports real-time streaming--including Server-Sent Events, WebSockets, gRPC, and chunked transfer encoding--in Public Preview without requiring changes to your existing API specifications. This update allows developers to route long-lived, bi-directional streams and AI workloads directly through the gateway while maintaining consistent authentication, rate limiting, and observability. Additionally, gateways now feature deterministic default hostnames, simplifying environment provisioning and infrastructure-as-code management.

## Powering the Agent Development Lifecycle with any framework, from ADK to LangChain and beyond

DevFeed: [Powering the Agent Development Lifecycle with any framework, from ADK to LangChain and beyond](<https://devfeed.tech/articles/powering-the-agent-development-lifecycle-with-any-framework-from-adk-to-langchain-and-beyond-93047.md>)

Original publisher: [Read original article](<https://developers.googleblog.com/powering-the-agent-development-lifecycle-with-any-framework-from-adk-to-langchain-and-beyond/>)

Author: Pier Paolo Ippolito; Elia Secchi

Published: 2026-10-21T17:37:47.402845Z

Content type: tutorial

Language: en

Sources: [Google Developers Blog](<https://devfeed.tech/sources/google-developers-blog.md>)

Topics: [LangChain](<https://devfeed.tech/topics/langchain.md>), [google-adk](<https://devfeed.tech/topics/google-adk.md>), [Tool](<https://devfeed.tech/topics/tool.md>), [internal developer platform automation](<https://devfeed.tech/topics/internal-developer-platform-automation.md>), [code security](<https://devfeed.tech/topics/code-security.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [observability](<https://devfeed.tech/topics/observability.md>), [versioning](<https://devfeed.tech/topics/versioning.md>), [qualitygates](<https://devfeed.tech/topics/qualitygates.md>), [Serverless](<https://devfeed.tech/topics/serverless.md>)

Tags: [adk](<https://devfeed.tech/tags/adk.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agent-development-kit](<https://devfeed.tech/tags/agent-development-kit.md>), [agents](<https://devfeed.tech/tags/agents.md>), [bill](<https://devfeed.tech/tags/bill.md>), [can](<https://devfeed.tech/tags/can.md>), [cli](<https://devfeed.tech/tags/cli.md>), [custom](<https://devfeed.tech/tags/custom.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [developers](<https://devfeed.tech/tags/developers.md>), [development](<https://devfeed.tech/tags/development.md>), [development-kit](<https://devfeed.tech/tags/development-kit.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [extension](<https://devfeed.tech/tags/extension.md>), [extensions](<https://devfeed.tech/tags/extensions.md>), [framework](<https://devfeed.tech/tags/framework.md>), [langchain](<https://devfeed.tech/tags/langchain.md>), [lifecycle](<https://devfeed.tech/tags/lifecycle.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

Agents CLI's experimental Extension System lets teams use other agent frameworks and customize command behavior while keeping the shared deployment and evaluation workflow. The first-party LangChain template in v1.8.0 scaffolds a project without `google-adk`; it replaces six commands, including `run`, `playground`, and `eval generate`. Two evaluation subcommands explicitly refuse because they require an ADK agent object. Deployment to Cloud Run and several other lifecycle commands remain native and framework agnostic; generated evaluations use the standard format consumed by native grading and comparison commands. Extensions can also wrap commands with organizational checks, as in the example that scans an SBOM before invoking native deploy. Extensions are distributed through Git, pinned to commit SHAs in a lockfile, and restored by `agents-cli install`; the project scope takes precedence over user scope. The system's manifest and command surface are experimental and may change incompatibly, so the text advises pinning the CLI version.

### Source excerpt

The Agents CLI Extension System enables developers to use frameworks like LangChain, LangGraph, or custom enterprise stacks while retaining a unified deployment and evaluation workflow. It details a first-party LangChain template that scaffolds, runs, and deploys agents without requiring the Agent Development Kit (ADK). Additionally, it demonstrates how platform teams can author custom extensions to enforce policies such as automated software bill of materials (SBOM) scans during deployment.

[Next page](<https://devfeed.tech/sources/google-developers-blog.md?cursor=WyIyMDI2LTEwLTIxVDE3OjM3OjQ3LjQwMjg0NSswMDowMCIsICIxMGYyZjgzNy1iNzRmLTQ0YTUtYmFlMy0wNGMzNjNhYmNjMmEiXQ%3D%3D>)