• Computer Science > Distributed, Parallel, and Cluster Computing [Submitted on 26 Jan 2026] Title:Mind the Boundary: Stabilizing Gemini Enterprise A2A via a Cloud Run Hub Across Projects and Accounts View PDF HTML (experimental)Abstract:Enterprise conversational UIs increasingly need to orchestrate heterogeneous backend agents and tools across project and account boundaries in a secure and reproducible way. • Starting from Gemini Enterprise Agent-to-Agent (A2A) invocation, we implement an A2A Hub orchestrator on Cloud Run that routes queries to four paths: a public A2A agent deployed in a different project, an IAM-protected Cloud Run A2A agent in a different account, a retrieval-augmented generation path combining Discovery Engine and Vertex AI Search with direct retrieval of source text from Google Cloud Storage, and a general question answering path via Vertex AI. • We show that practical interoperability is governed not only by protocol compliance but also by Gemini Enterprise UI constraints and boundary-dependent authentication. • Real UI requests arrive as text-only inputs and include empty accepted output mode lists, so mixing structured data into JSON-RPC responses can trigger UI errors. • To address this, we enforce a text-only compatibility mode on the JSON-RPC endpoint while separating structured outputs and debugging signals into a REST tool API. • On a four-query benchmark spanning expense policy, project management assistance, general knowledge, and incident response deadli
Computer Science > Distributed, Parallel, and Cluster Computing [Submitted on 26 Jan 2026] Title:Mind the Boundary: Stabilizing Gemini Enterprise A2A via a Cloud Run Hub Across Projects and Accounts View PDF HTML (experimental)Abstract:Enterprise conversational UIs increasingly need to orchestrate heterogeneous backend agents and tools across project and account boundaries in a secure and reproducible way.
Article Summaries:
- The paper presents a Cloud Run‑based hub that stabilizes Gemini Enterprise’s agent‑to‑agent (A2A) interactions across multiple Google Cloud projects and accounts. The hub routes user queries to four distinct back‑ends: a public A2A agent in another project, an IAM‑protected agent in a separate account, a retrieval‑augmented generation path combining Discovery Engine, Vertex AI Search, and direct GCS access, and a general Vertex AI question‑answering path. To avoid UI errors caused by mixed structured data, the authors enforce a text‑only compatibility mode on the JSON‑RPC endpoint while exposing structured outputs through a separate REST API. Benchmarks on policy, project management, knowledge, and incident‑response tasks confirm deterministic routing and stable UI responses, with reproducible results available in the cited repository.
- Summary
Researchers have built a Cloud Run‑based “A2A Hub” to stabilize Gemini Enterprise’s agent‑to‑agent (A2A) interactions across multiple projects and accounts. The hub routes queries to four distinct paths: a public A2A agent in another project, an IAM‑protected agent in a separate account, a retrieval‑augmented generation route that combines Discovery Engine, Vertex AI Search, and direct GCS access, and a general Vertex AI question‑answering path. To avoid UI errors caused by mixed structured data, the team enforces a text‑only compatibility mode on the JSON‑RPC endpoint and separates structured outputs into a REST tool API. Benchmarks on expense policy, project management, general knowledge, and incident‑response queries confirm deterministic routing, stable UI responses, and reproducible results.
- Researchers have built a Cloud Run‑based “A2A Hub” to stabilize Gemini Enterprise’s agent‑to‑agent (A2A) calls across multiple Google Cloud projects and accounts. The hub routes queries to four distinct back‑ends: a public A2A agent in another project, an IAM‑protected agent in a different account, a retrieval‑augmented generation path that combines Discovery Engine, Vertex AI Search, and direct GCS retrieval, and a general Vertex AI question‑answering path. The team identified that UI requests are text‑only, so they enforced a text‑only JSON‑RPC mode while separating structured outputs into a REST API to avoid UI errors. Benchmarks on expense policy, project management, general knowledge, and incident‑response deadline extraction confirmed deterministic routing and stable UI responses. The full reproducible code is available under the a2a‑hub‑gemini‑ui‑stable‑paper tag.
Sources: