关于“使用 ADK 和 Agent2Agent (A2A) SDK 連線至遠端代理”的评价

Shreetama B. · 评论about 1 month之前

I kept getting a 403 error which did not get fixed with the given instructions in the codelabs

Amartya G. · 评论about 1 month之前

things are not working

Prashanth reddy G. · 评论about 1 month之前

The lab content is good in theory, but the execution experience is very frustrating. I followed all the steps carefully and was able to successfully start the server and open the ADK UI. However, the agent did not respond to any prompts. The terminal consistently showed "403 Forbidden" errors for session-related endpoints (such as /users/user/sessions and /run_sse), which completely blocked progress. This appears to be an authentication/session issue in the lab environment rather than a mistake in following instructions. The lab seems highly sensitive to account/session configuration, but this is not clearly explained. Additionally, the troubleshooting guidance provided (e.g., modifying the URL using "authuser") was not applicable in my case because the URL did not include any "authuser" parameter. This made it even more confusing, as the suggested fix could not be applied. As a result, I spent most of the lab time debugging environment and authentication issues instead of learning the intended concepts. Suggestions: * Clearly instruct users to use a single Google account or incognito mode from the beginning * Provide accurate and applicable troubleshooting steps (especially for 403 errors) * Ensure instructions match the actual URLs and environment behavior * Improve session handling to make the lab more stable and beginner-friendly Overall, the concept is valuable, but the technical issues significantly reduce the learning experience.

Mehhuu Y. · 评论about 1 month之前

403 error coming up, authuser doesnot resolves the issue

Harsh M. · 评论about 1 month之前

shreyash w. · 评论about 1 month之前

Aditya Raj S. · 评论about 1 month之前

Sai Nireeksha M. · 评论about 1 month之前

getting Forbidden Error 403 and it is not solving so cannot complete the tasks. lost 21 credits because of that

Rupal K. · 评论about 1 month之前

Shahid G. · 评论about 1 month之前

Unable to complete task 2 and 4 due to repeated 403 error, even after trying the suggested workaround.

Lalit K. · 评论about 1 month之前

sakshi g. · 评论about 1 month之前

Vinamra J. · 评论about 1 month之前

lab not working in cloud console. Unable to get the output

DEEPAK K. · 评论about 1 month之前

Kundan K. · 评论about 1 month之前

Harshit Sai T. · 评论about 1 month之前

Anmol H. · 评论about 1 month之前

Hi Support, I would like to request a credit refund for this lab. I encountered a persistent technical issue with the ADK Web UI on Cloud Shell. Even after following the instructions, the interface failed to generate a Session ID and User ID, returning a '403 Forbidden' error on every POST /sessions request. This prevented me from completing the lab tasks despite multiple attempts (including Incognito mode and port changes). It seems to be a backend/proxy issue with the Cloud Shell Web Preview. Please check and restore my credits. Thank you.

Mutiara N. · 评论about 1 month之前

Daksh B. · 评论about 1 month之前

SOMSUBHRA N. · 评论about 1 month之前

Problem 1 — Tasks 2 & 4: ADK Web UI Broken in Cloud Shell (403 Forbidden)The lab instructs students to launch adk web and interact with agents through the Cloud Shell web preview. Every POST request (/run_sse, /apps/.../sessions) returns 403 Forbidden, while GET requests succeed. This makes the UI completely non-functional for testing agents locally.The lab itself acknowledges this failure with a buried footnote: "If you get a Forbidden Error 403, you may have multiple accounts signed into your Incognito session." This is a misleading and incorrect diagnosis. The real cause is that the Cloud Shell web preview proxy drops POST authentication for Qwiklabs-provisioned student accounts. No combination of browser switching, cookie settings, authuser parameter adjustments, ADC resets, or session clearing resolves it.Workaround attempted and confirmed working: adk run CLI mode, which bypasses the web proxy entirely. Problem 2 — Task 4: Model IDs Injected by Lab Startup Script Are Invalid The lab's startup script injects model ID environment variables (MODEL, IMAGE_MODEL) via template variables into the .env file. In this session, the injected values (gemini-1.5-flash-001, gemini-2.5-flash-image) are not available in the Qwiklabs project's Vertex AI quota, returning: 404 NOT_FOUND: Publisher Model ... was not found or your project does not have access to it. This error surfaces both locally (in the slide_content_agent) and remotely (in the deployed Cloud Run illustration_agent service). The student has no visibility into which model IDs are actually provisioned for their temporary project. Every model variant attempted — gemini-2.0-flash-001, gemini-2.0-flash, gemini-1.5-flash-001 — returned 404 in at least one environment configuration, making it impossible to complete Task 4 regardless of workarounds applied. The Cloud Run service compounds the problem because it bakes the broken model ID into the deployment, requiring gcloud run services update commands not described anywhere in the lab instructions.

MD. Mamdud H. · 评论about 1 month之前

Kept going to 403 Forbidden. Tried all the things

Palak K. · 评论about 1 month之前

Muthappa M. · 评论about 1 month之前

Yuvraj Singh S. · 评论about 1 month之前

Rupali A. · 评论about 1 month之前