cloud

[warped helm]

Hey guys! I joined here just now

I am working on creating an agent using Conversational Agents tool in GCP. I am trying to make it do aggregrations on a dataset that I have provided it. Has anyoen done something similar ?

the agent keeps giving me responses like it is not capable of doing aggregationson the dataset even though I have specifically instructed to do aggregations if required to answer the user's prompt

[errant shell]

The agent probably needs a tools for it to perform the action you want it to do. Without that, it can plan the aggregations, but not be able to execute them.

[warped helm]

I can create tools to perform aggregations on the data? what kind of tool would that be? The only options I see when I create a tool for the agent are : OpenAPI, DataStore, Function or Connector. As of now I was only using the data store tool to connect the dataset to the agent.

[errant shell]

The approach you chose here will depend on the location of your dataset. Where is your dataset located? Bigquery?

[astral linden]

Add new project in AI Studio isn't working

[pine charm]

What are you clicking on and what does it do when you try?

[astral linden]

idk why, when i add to create new project to create new API, isn't working

[pine charm]

Do you have projects in Google Cloud Platform already? And, if so, how many?

[astral linden]

none at all

[pine charm]

Did you have any keys previously in AI Studio?

[livid verge]

I've seen a few of your posts. You obviously know very much what you're talking about. If I say I had a large amount of documents I wanted to train an AI on and I mean a lot of them are personal written, some are internal, etc. I'm guessing this wouldn't cut it. I mean I'm not looking at something like Claude Code, don't get me wrong. More like a knowledge base basically. I'm More looking instead of like document retrieval actually having it in the database and being able to put cross reference etc So I guess an improved version of our AG now I will be building a home AI lab, but honestly the power won't cut it It'll be some older card like a couple k80s and accelerator and another thing So I'm gonna guess that training this would be best on a Google Cloud AI and then used locally I mean I've got plenty of storage space don't get me wrong on that and plenty of processing power just not GPU power which is kind of the killer really

Because I work in cyber security a lot of the stuff I work with frankly trips the guardrails of well everything online really and I'd rather keep a lot of our documents in house as some of them are a wee bit sensitive so what do do you think just I mean a general vague direction would be great

[pine charm]

I'm not sure there was a question in there. And I like to think I know something about this field - but I'm honestly still learning too.

It sounds like you're saying that you have sensitive information that you want to use for some internal purposes.

Remember that model tuning is good to introduce new words and patterns, but not knowledge. If knowledge is what you're looking to prode, then a RAG-like solution may be better.

And yes, you can setup training pipelines for something like Gemma 3, and then download the results of that training to use on a local model.

[livid verge]

struggling to straddle the line between what I can say and what I should say and what I can't say.

weaving together vast amounts of information into new novel information and knowledge is what I was seeking in specifically the realm of cyber security and defense but I mean that's hard I know it's possible because I've seen it I have seen this with my own two eyes but yeah that's not exactly easy

Nailed a totally novel killchainOf about 4 or 5 zero days including firmware level things in about 10 minutes of conversation with a rather interesting LLM but that there is probably going to be at my reach for a while personally.

Things are advancing rather fast and who knows what the world will look like in ten years....or 2

I will ruminate on it and explore Google's architecture when I'm not currently under active cyber attack. So, uh, not today, I guess.

[pine charm]

So that would be... never. {:
(I used to do security as well. Ages ago. I like to be able to sleep nights, now.)

[livid verge]

...heh

[livid verge]

There's active as in you know they're kind of trying and then there's active as in someone is actively at this very moment constantly trying zero day after zero day after zero day after zero day

I'm desperately trying to compile a kernel because they killed my last one and actually be able to compile my real kernel which will make attack basically impossible but unfortunately I can't even get around to that because of how often I get hit

[astral linden]

no

[pine charm]

Try going into the Google Cloud console (console.cloud.google.com) and creating a project. If that works, in AI Studio, try importing that project.

[amber gazelle]

Does anyone know if Gemini’s Google Search grounding runs from the same region as the Vertex AI location I set (e.g. europe-west1), or is search handled globally regardless of inference region?

[manic geyser]

Is it possible to only get retrieved contexts (without the Generation step) in Vertex AI search without creating a RAG corpora? Maybe I misunderstood but when I use a datastore I can use Vertex AI Search with grounding without creating a RAG corpora

[orchid mountain]

hello

[livid verge]

Hi

[orchid mountain]

How can I help you?

[livid verge]

Uhhhh.....Some exotic Google Tensor board maybe? I don't know who you are. Who am I speaking to?
The FBI to a currently unknown address but a person in Rochester, New York. I don't know. Really?

[cosmic tiger]

<@&1009526435276394496> 👆 spam

[sacred forum]

Thank you!

[pale cloak]

Hey, does anyone know a way to ask for help about an billing issue that I have? I can't find it D:

[verbal berry]

What help do you need?

[pale cloak]

Google is charging me for a service in an account that i never use xD

I dont know what is cloud spanner haha

[pine charm]

Cloud Spanner is a database.
You should check the project it is billed to, make sure it is not enabled, and contact billing on that account. Your bill should have contact info.

[pale cloak]

Im on it, chatting with support team

[manic geyser]

Any idea?

[manic geyser]

Or is there any option to create a RAG Corpora without Cloud Spanner?

[pine charm]

You can look at something like Vertex AI Vector Search

[quick pumice]

<@&1009526435276394496> spam

[hot folio]

Hi Everyone,
I just joined this amazing community. hope to get in touch with everyone 🙂

[ocean lily]

Hi guys, I applied for Gnerative AI Certification in google cloud
Did anyone get any update regarding this?

[manic geyser]

But regarding Vertex AI Search: which vector database uses Vertex AI Search when there is no Cloud Spanner

[pine charm]

Note that I said Vertex AI Vector Search. Which is a different product than Vertex AI Search.

I was also unaware that VAI Search used Spanner

[compact lance]

Hii !
I am facing an error trying to deploy a FastAPI app on GCP Cloud Run, via a github action.

This is the step throwing an error

- name: Deploy to Cloud Run
              env:
                  APP_NAME: ${{ secrets.APP_NAME }}
                  DB_USERNAME: ${{ secrets.DB_USERNAME }}
                  DB_PASSWORD: ${{ secrets.DB_PASSWORD }}
                  DAGSHUB_TOKEN: ${{ secrets.DAGSHUB_TOKEN }}
                  MONGO_DB_URI: ${{ secrets.MONGO_DB_URI }}
              run: |
                  gcloud run deploy ${{ secrets.SERVICE_NAME }} \
                    --image docker.io/${{ secrets.DOCKERHUB_USERNAME }}/networksecurity_ml-project:latest \
                    --platform managed \
                    --region ${{ secrets.GCP_REGION }} \
                    --allow-unauthenticated \
                    --port 8080 \
                    --memory 2Gi \
                    --cpu 1 \
                    --set-env-vars "APP_NAME=${APP_NAME},DB_USERNAME=${DB_USERNAME},DB_PASSWORD=${DB_PASSWORD},DAGSHUB_TOKEN=${DAGSHUB_TOKEN},MONGO_DB_URI=${MONGO_DB_URI}"

And this is the error statement:

ERROR: (gcloud.run.deploy) [SERVICE_ACCOUNT_EMAIL] does not have permission to access namespaces instance [PROJECT_ID] (or it may not exist): Permission 'iam.serviceaccounts.actAs' denied on service account <[e-mail]> (or it may not exist).

Did anyone else encounter this error?
Any help would be very much appreciated🥲

[lavish sun]

the service account you're deploying your Cloud Run service with (the one in GH actions) does not have actAs permission which is required to assign a service account to CLoud Run service (while you're not explictly specifying any, your default compute SA is used - the one with project-number-compute).
Add missing permission to the SA you're deploying with and it should be all good

[compact lance]

I did add Cloud Admin role and Service account user role, which I understand has the actAs permission, to the service account.

Just in case, I also tried adding the service account user role to the default compute service account, but the error still persists

Also, the default compute SA, when I looked it up on Service accounts under IAM had a query parameter or something attached to it like, [e-mail]?u=1234....

The SA mail shown in the error message doesn't contain anything like this query parameter, I was wondering if that is why nothing's happening even after adding permissions

[manic geyser]

Okay but I don't understand why Vertex AI Search doesn't use Cloud Spanner for RAG while the other options do. I just want to use Grounding with Vertex AI Search but without the generation but only the retrieval since I am only interested in the retrieved chunks

[uncut glen]

Hello everyone, I’m using the Vertex AI Gemini models in my iOS application. Today, I’ve noticed that the output of Gemini Pro and Flash is too small or even malfunctioning. It doesn’t react to the JSON structure it should like before, and the output is only 200-300 tokens. Has anyone else noticed this issue?

[frozen light]

"Hello, I'm from Paris and I'm a Data Analyst / Data Scientist. I want to learn cloud computing for the first time — what would you advise me? Which YouTube channels should I watch, and which courses should I take to pass a certification?

[acoustic fern]

Hi everyone. I have seen that recently Google introduced an integration between Vertex AI deployments and Model Armor (https://cloud.google.com/security-command-center/docs/model-armor-vertex-integration). Has someone tested it?

[queen spoke]

First, choose one of cloud platform - GCP, AWS, Azure.

And create free account and try to do this.

• Launch a virtual machine
• Upload data to a storage bucket
• Deploy a small web app

And study about DevOps

Docket, Kubernetes etc..

[frozen oak]

Hi everyone, I’ve registered for the Google Cloud Skill Boost Edition 3 and have received confirmation of my participation. The email mentions that the program begins on October 22 and requires registration on Google Cloud Skill Labs using a corporate email ID. However, I’ll be leaving my organization on October 30. Is there any way to update or change my registered email address? Your assistance would be greatly appreciated.

[cinder ginkgo]

Hey! I can help with that. It sounds like the issue is that your Google Cloud Skill Boost registration is tied to your current corporate email, which you’ll lose access to after leaving your organization. Could you confirm if you’ve already created your Skill Labs account with that same email, or just used it for registration so far?

[frozen oak]

Heyy thanks I didn't yet created the skill labs account with the corporate email. I've just used it for registration.

[frozen light]

Thank you sso much for your response

[dull wren]

Does Google Cloud has or planning to have AI model deployment/APIs as a service like Amazon Bedrock and Azure AI Foundry?

oh it exists, it's vertext ai model garden

https://console.cloud.google.com/vertex-ai/model-garden

[proud spire]

I'm gonna lose my mind... is the gdrive 2TB useful for anything at all? I can't even stream a dataset cuz It's drive... and google colab has some horrible compute power costs to begin with

has anyone found a workaround to use the 2TBs in a productive manner?

not going back to gcolab, no ssh access makes it a hassle to work with

[pine charm]

It is generatlly meant to be used for workspace operations, not data processing operations. Tho I think there are integrations with BigQuery which may be useful for colab integration.

[cosmic tiger]

Are you trying to find a problem for gdrive to solve or do you have a problem that gdrive is not addressing?

[proud spire]

honestly, I just wanna use gdrive for dataset streaming but it seems like that's not possible w/o gcolab. On the other hand I can't think about a single use case for gdrive besides ||using it as computer storage and having google screwing you over later on||

[proud spire]

just figured out it has some handy integration with other services, glad to know that. Thanks

[cosmic tiger]

gdrive wasn't necessarily built with that intention. Do you need to use gdrive or you just want to use gdrive. If you have to use gdrive, what are not able to do that you need it to do?

[proud spire]

some api to gather my data that is

[cosmic tiger]

I think you misunderstood my questions. Are you looking to use gdrive in a specific way for a specific reason? Like, can you just load all the data off gdrive or is there too much data and you want to stream it. Or, you figure you have 2TB included/free with services you already consume and you don't want to spend money on a more appropriate technology? Or, you're just curious and you want to hack gdrive to do something that would be interesting?

For example, if you have a reasonablly small dataset and you just want to use gdrive for storage it'll work if that's all you need.

[proud spire]

I thought I could use those 2 TB to stream my dataset, I don't wanna spend much money on an external service besides cloud compute power, It's just that drive doesn't seem to offer such capabilities right off the bat, was wondering if anyone has figured out an use case besides just using it to store personal data

storing the data alone is great, but yet again it is not feasible to use for cloud compute ai training, which I guess is fairly out of gdrive's scope

maybe some workaround with google colab, the fact that it still doesn't offer an official linux client or cli utility, makes gdrive not fit for developer use case, at least in my opinion.

anyways, would love to hear if anyone has found some cool workaround

[cosmic tiger]

Do you need to "stream" the data? Also, what is you definition of stream in this context? Not sure what you mean by "store personal data" as Google drive can store data for many different uses.

As a basic datastore, it's not any different than any other basic datastore so a variety of applications can be built on top of it.

[proud spire]

by streaming I mean retrieving the data with an api temporally to train an ai model

I can't really fit the entire database in a cloud compute solution

maybe I could shard it, and use the existing hacky solutions to gdrive to download the shards

but that's beyond what I know to do

might be worth a try

[cosmic tiger]

Don't know how large your dataset is or what kind of data it is or what format. But, you can load documents individually from gdrive or open sheets and read line by line, etc.

gdrive can be treated like a local drive and you can work with files the same way.

Though it seems you want to use 2TB, which in that case you really need to be using a solution that was built for that much data

[proud spire]

good point

[buoyant obsidian]

Depends how new to Cloud & programming as a whole you are I would say
If you're brand new to it, I would recommend doing first the Cloud Digital Leader certification, then the Professional Cloud Architect one :)

Those should give you the fundamentals about GCP and more generally about the Cloud.
▶️ To prepare for the certifications, there are dedicated lessons, labs & learning paths available on skills.google (i strongly recommend starting here)

[frozen light]

thank u

[buoyant obsidian]

No problem!
If you have other questions or need help feel free to ask!
(by the way, I'm also from Paris :D)

[thorn cliff]

Hey guys! I really wish to attend The Agentverse Austin in Austin, TX on Oct 21st but I am still waitlisted. Is there any way for me to contact the organizers (yay if they're here) and get accepted if there's one space (pls pls)?

@hidden monolith

[slender reef]

hi everyone

i have a question how to do exception for user level not project

i want to restricte creation public ip just to some users

[shadow otter]

What kind of roles can one target after completing GCP Cloud Architect certification? Of course, after having done some hands on projects!

[novel silo]

The new Skills landing page is very nice! 🙏

[safe solstice]

Its' definiitely a huge improvement and looks great! I love the gamification. 🥳

[novel silo]

Definitely! I really hope they expand on the AI Boost Bite section. I don't always have time for a full module, but reading or watching a 10 minute video in carline while picking up my kid is great.

[fast relic]

how can we join nvedia

[silent vine]

@tawny valve Please do not spam across channel. Use only #chrome-built-in-ai . Thank you

[quaint gale]

Very good idea please don't share or do spam please

[marble hound]

is there any course of based on cloud will launched by the Google from 10 nov to 15

[vernal phoenix]

🎉 Memory Revisions now available in Preview for Vertex AI Agent Engine

Hi everyone,

Wanted to share a new capability for those of you building agents with Vertex AI Agent Engine. Memory Revisions is now in Preview.

This feature addresses the common challenge of managing and debugging agent state by providing an automatic, immutable version history for Memory resources.

How it works:

• Each modification (creation, update, or deletion) of a memory generates a new MemoryRevision resource.
• Memories can be restored to a prior state using the RollbackMemory method, specifying a target_revision_id.
• For tracking purposes, GenerateMemories requests support revision_labels to categorize state changes, which can then be filtered using ListMemoryRevisions.
• When a memory is deleted, a final revision with an empty fact is created. Child revisions remain accessible for recovery for a period of 48 hours.

It's enabled by default with a 365-day TTL and you can customize it at the instance or request level.

You can find docs and code here.

On Vertex AI Agent Engine, we released so many other things and I will try to share content with you this week. Happy building!

[vernal phoenix]

🚀 Vertex AI Agent Engine now supports Inline Source Deployment!

Hey all, quick heads-up if you're deploying agents on Vertex AI Agent Engine.

We finally added Inline Source Deployment. You can stop serializing your agents into pickle files and uploading them to GCS buckets.

How it works
Pass your local directories (core/, deployment/) directly in the API call.
The SDK handles the tar/base64 encoding and kicks start the deployment process

Why it matters
You can actually git diff your deployments now.
Standard SAST tools can scan the code before it pushes (impossible with binary blobs).
You don't need to manage a staging bucket anymore.

You can find code and blog below to get started!

• Notebook: https://github.com/GoogleCloudPlatform/generative-ai/blob/main/agents/agent_engine/tutorial_deploy_your_first_adk_agent_on_agent_engine.ipynb
• Blog: https://discuss.google.dev/t/deploying-agents-with-inline-source-on-vertex-a

Happy building!

[fast rover]

Hi team i have to change my legal name in certmetrics of my profile ,i have raised ticket multiple times,but i didn’t get response,I have examination on coming sunday,please provide the solution

[copper nexus]

Hi, I checked and it appears that the only way to change the name is to send a request through following link. Is that the link you used?
Make sure you send the request from the same email as your exam related account: https://support.google.com/cloud-certification/contact/Cloud_Certifications_Contact

[paper summit]

I don't anything about cloud of google and how can I learn that any suggestions or guide lines from the people who have kindness

[novel silo]

Google Skills and the Google for Developers portal has many courses, codelabs, and certification paths. Might have a look through there. You can start at basics with gcloud, learning the UI, etc.

[paper summit]

Ok

[floral crystal]

guys i got a query , how do you guys maintain long term memory in vertex ai , memory bank or something else??

[stiff flicker]

Hey those who did Cloud study Jams for EMEA did you receive any emails after the 20th

[queen spoke]

In Google Colab, I have to train my model more than 12 hours, but colab's max duration time is 12 hours, how can I increase this?

[pine charm]

It depends on your needs and what, exactly, you're trying to store. I've used Agent Engine Sessions and Memory Bank in the past. They're both pretty straightforward.

[rugged iris]

Looks like Vertex forgot to update the SynthID detection model that's on the Imagen playground, because it's very hit or miss (mostly miss) on Nano Banana Pro outputs.

[agile talon]

Hi everyone,

Over the past five months, I’ve been working on a book focused on building AI applications with Genkit and Go. I’m happy to share that it’s now available, and fully open source.

1. Online version: https://mastering-genkit.github.io/mastering-genkit-go
2. Downloadable copies: https://github.com/mastering-genkit/mastering-genkit-go/releases
3. Physical copy on Amazon: https://www.amazon.com/dp/B0G4HKJRPB

[hallow current]

Hey all, Im having some issues here is a detailed question I made with Gemini:

I’m deploying a custom ADK agent (LlmAgent subclass) to Vertex AI Agent Engine (Reasoning Engine) and I'm running into an issue where custom telemetry events seem to be stripped or filtered out before reaching the client.

The Goal: I need to send real-time telemetry (cost per turn, energy usage, RAG source URIs) to my Nuxt 3 frontend. I implemented a "Sidecar Pattern" where I yield a final, synthetic Event containing a JSON payload after the LLM generation finishes.

The Problem:

1. Local Dev Works: When running locally (adk web), the event arrives perfectly.

2. Cloud Deployment Fails: When deployed to Vertex AI, the stream closes immediately after the final LLM token. The custom sidecar event never arrives at the client.

3. Logs Confirm Injection: My Cloud Logging stderr confirms the code is executing and injecting the event: INFO: 💉 Injecting Sidecar Telemetry: 2072 bytes

4. Client Stream: The client receives all standard text chunks, but the stream ends without receiving the final JSON chunk.

Questions:

1. Does the Reasoning Engine enforce a strict schema that drops events with custom author tags (like system_telemetry)?

2. If I set author="model" (which I tried), does the engine filter out events that look like "data" or don't match the internal generation state?

3. Is there a recommended way to pass custom metadata (like citations or cost) through the Reasoning Engine stream without it being stripped?

[vernal phoenix]

📣 NEW VERTEX AI ENGINEERING BLOG: Implementing EAGLE-3 at scale on Vertex AI with SGLang

Hi all,

Vertex AI just launched a NEW engineering blog to document its applied research.
The first post in collaboration with SGLang details how they implemented EAGLE-3 at scale and what they learned .
Check out the blog. They open-sourced the notebook if you want to reproduce some benchmark results yourself.

• Blog: https://docs.cloud.google.com/vertex-ai/docs/blog
• Notebook: https://github.com/GoogleCloudPlatform/generative-ai/blob/main/open-models/benchmarking_eagle_on_vertex_ai.ipynb

Happy learning!

[novel silo]

Really pushing those ads for Google Cloud Next 🤣☁️ I already wanted to go but have no reason 😭

[vernal phoenix]

📣 NEW Tutorials about bow to prepare preference data and use custom metrics for Gemini Tuning on Vertex AI

Hey everyone!

Many of you have asked for guidance on Gemini Tuning with Vertex AI, specifically: "How do I prepare tuning/preference data?" and "How can I measure improvements?"

Together with Haichao and Jessica from the Vertex AI Engineering team, we have published two new tutorials to answer exactly that.

What you'll learn:

• Inject custom metrics (like F1 score or JSON validation) directly into the Supervised Fine-Tuning loop.
• Execute custom code during the tuning job for tailored performance evaluation.
• Use the Vertex AI Gen AI Evaluation SDK to automatically score preference datasets and visualize quality distributions.
• Filter out noisy data by creating a clear quality gap between "chosen" and "rejected" responses.

You can find notebooks below:

• Data Prep: https://github.com/GoogleCloudPlatform/generative-ai/blob/main/gemini/tuning/dpo_gemini_data_prep_tuning.ipynb
• Custom Metrics: https://github.com/GoogleCloudPlatform/generative-ai/blob/main/gemini/tuning/sft_gemini_custom_metric_evaluation.ipynb

[daring fable]

Hey Ivan Nardini! Awesome update on custom metrics for Gemini Tuning on Vertex AI. This is a massive improvement for quality control.

From a deployment and DevOps perspective, how do these custom metrics and the fine-tuning process affect the model's overall latency and 'cold start' times when deploying the tuned model in a production environment?

[daring fable]

@sinful nexus (Moderator), this marketplace/spam from @Matt has now hit #cloud, disrupting important discussions on Gemini Tuning (like the one I started with @Ivan Nardini). Please remove the user, as this violates server rules across multiple technical channels. Thank you.

[sinful nexus]

Thank you for reporting it but I wasn't online at that time. It was handled by other mods.

[scenic shadow]

It is wondering that AI supports the Cloud. I m new here and wanna learn about it.

[slow trout]

Hi

I need help in vertex ai
How i can add eventarch when my batch job(gemini model) completed

[vernal phoenix]

Hi all,

Vertex AI just launched the Cloud API Registry integration in Vertex AI Agent builder.

Instead of manually writing wrappers and managing fragmented secrets, you can now use a centralized catalog of MCP servers.

Key Benefits:
🌐 Centralized Governance: Provides a unified view of tools across the organization.
🛠️ Zero Boilerplate: We can consume standardized tools (e.g., BigQuery execute_sql) without writing implementation code.
🔐 Unified Security: Handles auth via standard IAM policies rather than granular API keys.

Here you can find new guide with tutorial notebook on how to deploy a Data Analyst Agent on Vertex AI Agent Engine with Cloud Registry API:

• Guide: https://discuss.google.dev/t/where-is-the-mcp-server-deploy-your-agent-with-cloud-api-registry-on-vertex-ai-agent-engine/298130
• Tutorial: https://github.com/GoogleCloudPlatform/generative-ai/blob/main/agents/agent_engine/tutorial_get_started_with_cloud_api_registry.ipynb

Happy building!

[vernal phoenix]

Hi Gabriel, the tuning service is based on low-latency and scalable Gen AI Eval service on Vertex AI. I would not expect significant impact during tuning. And becuase the evaluation happens during tuning process, there would be no impact on the deployment of the resulting tuned model. Hope I provide clarification you were looking at. In case not, feel free to reach out!

[vernal phoenix]

Hi all,

Vertex AI just released Agent Designer, a low-code visual interface that allows you to orchestrate agents and subagents on a canvas, then export the logic directly to the Agent Development Kit (ADK) for code-level refinement.

With Agent Designer, you can:

• Design agents, subagents, and control flow on a canvas.
• Test agentic flow with a chat interface
• Hit "Get Code" to generate the Python scaffold for the Agent Development Kit (ADK) and move straight to your IDE.
• Toggle on Google Search, URL context, and connect Vertex AI Search Data Stores (RAG) via the UI.
• Experiment with connecting Model Context Protocol servers (currently unauthenticated only).

Note: It’s in Preview. It’s great for scaffolding, but it doesn't support advanced ADK patterns yet, so complex logic still lives in the code.

Here the documentation to get started with the new Designer: https://docs.cloud.google.com/agent-builder/agent-designer

Questions or feedback? Connect with me on LinkedIn (in/ivan-nardini) or X/Twitter (@vernal phoenix)

Happy building!

[rough vector]

Hi @vernal phoenix where can I find alphaenvolve for test? I don't see any form to sign up for testing. It's not available at agents either.

[night magnet]

Hello everyone, I would like to contact a Cloud Google Developer Expert.

[quick pumice]

see #gde-public

[novel silo]

This is pretty cool. Unified cloud docs page!
https://docs.cloud.google.com/

[novel summit]

Thanks

[buoyant obsidian]

Np! Good luck!

[novel summit]

Thanks, Neïl! I'm actually starting a project with Shield Gemma for AI Red Teaming right now. Really excited to dive deeper into the Google Cloud ecosystem.

[finite karma]

Anyone else not able to load https://console.cloud.google.com/welcome

It's taking forever

[remote roost]

Good evening guys

Please I need your help, I need to know what Cloud resources are connected to the Trial GenAI credit I got on google.

Essentially, I just need to know how exactly to use these credits.

[cinder ginkgo]

The credits aren’t attached to any Cloud resource yet. The problem is there’s no automation or connection between two systems, so nothing is triggering the GenAI service to use the credits.
Once you connect a data source to GenAI App Builder and set an automatic trigger, the credits start applying immediately.
I can help you set that up so everything syncs instantly.
Quick question: what are you trying to connect it to (Vertex AI, a web app, or an external tool/API)?

[remote roost]

Thanks for responding Joshua

I am trying to connect it to Vertex AI, and then if possible a firebase project.

[cinder ginkgo]

That setup is possible. Once Vertex AI and Firebase are under the same Google Cloud billing account, the GenAI trial credits apply automatically. Nothing needs to be manually attached.
I can set this up for you end-to-end and make sure everything is connected and the credits are actually being used.
If you’re open to it, we can handle this privately and get it done quickly.

[novel silo]

👋 I used a gcloud trial credit as well for a few projects. My general workflow for using it was create a new project on gcloud, link billing account. There is likely two or more options if you previously linked billing, your trial account and your normal billing account, link your trial account to the project and set cloud limits aka budgets even though it is a trial, it's good practice, then link the project in gcp to whatever resources you are using.

[remote roost]

Okay, thanks a lot, I'll try linking the a new project directly to it!

[remote roost]

Thanks, this was a lot of help, I can manage from here.

I really appreciate you taking the time to respond!

[stuck temple]

hi guys

i have a ask

venezuela is a avaible country for pay/use google cloud officially?

i don't see venezuela on country list

[solid oriole]

to use, yeah, but to pay, mostly not

[last flame]

Hello, has anyone tried using the Deployment Resource Pools? I was trying to deploy an LLM model using that but having trouble with it due to the TF/pytorch version in the prebuilt containers, anyone used it for something similar?

[novel silo]

A cool resource I found
https://github.com/amitkmaraj/course-creation-ai-agent-architecture

[woeful python]

Okay.

[raven frigate]

Claude Sonnet 4.5 is enabled in the model garden, but quota still 0, submitting quota request will be denied, i wonder why...🤔

Anyone has gone through this complex process of quota systems in gcp?

[raven frigate]

<@&1009526435276394496> spammer

[buoyant tartan]

🔄 This channel is back from the archive, you can still access previous topic history in: https://discord.com/channels/1009525727504384150/1464300723570213048

[finite rapids]

Is this things on

[pine estuary]

hello why genai file api upload file is so slow?

300mb takes about forever

[2026-04-26T04:42:56.993Z] [info] Still uploading video to Google Files API
{
"videoFileName": "f.mp4",
"sizeHuman": "315.6 MB",
"elapsedMs": 1240196,
"note": "The Google SDK does not expose byte-level upload progress."
}

[vital basin]

Hmm

[hallow abyss]

Hii can someone tell me what is cloud and how it's work

[meager lichen]

hey wehen i try to use my credits i get "error Your prepay credits are depleted. "

[winter frigate]

you get servers do what u want, pretty fun, gets advanced very quick

any ideas on when the rollout will complete?

pls not say 2-3 weeks

can we bypass it with gcloud cli or something

[craggy junco]

💀

[last stream]

Any chance OAuth MCP support landing in Agent Designer is part of I/O?

could really push my build along

[runic swan]

Just message to support and it will be resolved in next 24 hours

[winter frigate]

cant

needs support plan

which i dont want for a quota only lol

[runic swan]

What service you are looking for just compute capacity?

[winter frigate]

i need c4 family 4 vcpus committed quota

i have the vm just want to make 3 year commitment

so its cheap

[runic swan]

Nice topic

[neon canopy]

Hello mate , new here looking for some experience and hand on cloud

[lucid coyote]

If upgrade Google Cloud free account to paid account, how to credits that still remaining? Can used for try Gemma using GPU or how to?

[median lark]

I need to use a coupon on an existing billing account to pay for a bill

[void fjord]

Yes your remaining credits usually still work until they expire after upgrading.

And yes, you can use them for GPU stuff like running Gemma, but you’ll need billing enabled and GPU quota access (Vertex AI / VM).