RSS Feeds

Новичок открывает первый кейс, о котором я рассказывал в первой статье., читает брифинг, и через пять секунд сайт пишет: «Готово, жми „Начать“». Он жмёт и смотрит на пустое окно терминала. Секунд через десять там ошибка, страница сама перезагружает терминал, и только потом появляется консоль.

Сервер при этом не падал, CPU и память в норме, алерты молчат. Были только строчки в журнале, которые я чуть не записал в «обычный шум».

После хорошего доклада хочется многое попробовать. Но в понедельник возвращаются текущие задачи, а заметки остаются лежать до лучших времён. На INFOSTART A&PM EVENT 2026 часть подходов можно будет проверить ещё на конференции - на своей практике, с помощью экспертов и коллег.

Конференция для аналитиков и руководителей проектов 1С пройдёт 12–14 ноября в Санкт-Петербурге. В программе будет более 120 выступлений и активностей, в том числе 54 практических занятия и пять круглых столов. В полном материале рассказываем, что можно попробовать и как выбрать полезное для себя.

Привет, Хабр! В этом материале сделаем сервис, который разгрузит диспетчера сервисного центра по ремонту бытовой техники. Голосовой робот будет принимать входящий звонок и собирать детали обращения. На каждое обращение в Битрикс24 будет создаваться карточка с назначенным мастером.

Стек проекта: Python 3.10+, FastAPI, Pydantic, SQLite, HTTPX, Битрикс24, Яндекс Геокодер, голосовой робот и SMS API от МТС Exolve.

Democrats in Congress are bracing for what they fear could be meddling by President Trump or his allies with the results of this year's midterm elections.

There are ghouls, gremlins and the grotesque but through its five seasons the anthology show really focuses on the human condition

A man with a stern expression and a shock of slick black hair, wearing a crisp black suit, cigarette in hand, approaches the camera. In his clipped, precise patter, he’s about to welcome you … to The Twilight Zone.

This is how each episode of The Twilight Zone typically began, with its creator, Rod Serling, delivering a brief, cryptic opening monologue before throwing to a stand-alone story.

I look at myself and my neighbours queueing for water, and dream of a lost normality. We are now strangers even to ourselves

Every morning, I look at myself and my neighbours standing in the queue for water, and remember how each of us once had a life of our own, our privacy, our own toothbrush, our own home and our own bathroom. How we used to wash our faces and drink clean water without ever imagining that a day would come when these simple routines would become a battle with ourselves and with every drop of water. Even standing beneath a hot shower has become a dream postponed until further notice.

A year has passed since the announcement of the “ceasefire” in Gaza, yet the spectre of war, bombardment and death has never left our lives.

Akram Al-Sourani is a writer and poet from Gaza City, and the author of A Homeland Out of Coverage, Half-Baked Homeland and What’s That

Do you have an opinion on the issues raised in this article? If you would like to submit a response of up to 300 words by email to be considered for publication in our letters section, please click here.

Matthew Unglesbee admitted to sending the letters to four unidentified politicians, officials say

An Illinois man accused of sending letters containing suspicious white powder to four members of Congress has been arrested after he allegedly wrote his own name and return address on the envelopes, according to authorities.

Matthew Unglesbee, 36, of Quincy, remained in the custody of US marshals on Tuesday on federal charges of perpetrating a hoax and sending threatening communications through the mail.

Каждый SRE хотя бы раз просыпался в три часа ночи от алерта о том, что Harbor опять упал, PostgreSQL реестра съел весь диск, а миграция после обновления сломала половину образов. Мы строим Kubernetes, чтобы забыть про серверы — а потом ставим реестр, который заставляет вспомнить про них снова. Только теперь их не один, а восемь, и у каждого свои логи, свои секреты и свои причины не запускаться. Эта статья — про то, как перестать содержать реестр как отдельный продукт и начать использовать его как инструмент: без фанатизма, без гигабайтов RAM и без молитв CSI-провайдеру перед каждым бэкапом. Разберём связку rgstr v1.10.1 и rgstroperator v0.5.1 — не как «Harbor для бедных», а как осознанный инженерный выбор для тех, кто считает TCO, а не гигабайты.

Developers mostly use GitLab Duo Agent Platform where they already work: the IDE, merge requests, and issues. But DevSecOps teams also support the business systems the rest of the company runs on, and those systems break too.

In Salesforce, for example, when a validation fails, a callout times out, or a null slips through a trigger, the error gets logged, but the log doesn’t explain the cause. An engineer still has to read the stack trace, find the class in the repository, check what changed recently, and decide whether this is a one-off or a regression. Meanwhile, the business user who hit the error has to wait.

When your Salesforce metadata lives in GitLab and ships through GitLab CI/CD, the code that explains the error is already there. In this tutorial, Salesforce calls a custom GitLab Duo Agent Platform flow through the Flows API the moment an error happens. The flow investigates the code, and your team gets a GitLab issue that already contains the root-cause analysis.

Along the way, three parts of the GitLab platform work together: GitLab Duo Agent Platform investigates the error, a GitLab issue tracks the work, and GitLab CI/CD runs the flow and ships the fix.

Salesforce is the example we’ll walk through in this tutorial, but the same approach works for any system that can make an HTTP request.

What you'll build

Salesforce makes one call to GitLab. A custom flow in GitLab Duo Agent Platform does the rest.

Errors reach GitLab Duo Agent Platform automatically from your Apex error handling, or on demand from a button. Either way, the same Apex job calls the Flows API, and the flow opens an issue with its analysis.

Prerequisites

• GitLab Premium or Ultimate, on GitLab.com or GitLab 19.5 and later. The Flows API used in this tutorial is generally available from 19.5.
• GitLab Duo Agent Platform turned on, with custom flows allowed for your top-level group.
• Your Salesforce metadata (Apex classes, triggers, Flows) stored in a GitLab project. This is the project the flow will read.
• A Maintainer or Owner role in that project.
• In Salesforce, permission to create a Named Credential, a custom metadata type, and Apex classes.
• Your Salesforce deployments must be running through a GitLab CI/CD pipeline, so the fix can ship from the same issue.

Step 1: Write the triage flow

Start in GitLab, because Salesforce will need the flow's ID later.

A custom flow is one YAML file in the flow registry v1 syntax. This one has a single agent. When Salesforce starts the flow, it sends the error details as the flow's goal. The agent reads them as context:goal, investigates the repository, and opens one issue.

version: "v1"
environment: ambient

components:
  - name: "sf_error_triage"
    type: AgentComponent
    prompt_id: "sf_error_triage_prompt"
    inputs:
      - from: "context:project_id"
        as: "project_id"
      - from: "context:goal"
        as: "error_report"
    toolset:
      - "find_files"      # locate the class or Flow in the repository
      - "grep"
      - "read_file"
      - "list_commits"    # find recent changes to that file
      - "create_issue"    # the only write: open the triage issue

prompts:
  - prompt_id: "sf_error_triage_prompt"
    prompt_template:
      system: |
        You triage Salesforce runtime errors for a team that keeps its
        Salesforce metadata (Apex classes, triggers, Flows) in this project.

        You receive an error report with a class name, exception type,
        message and stack trace.

        1. Find the source file for the class or Flow named in the report.
        2. Read the code around the lines in the stack trace.
        3. List recent commits that touched that file and identify any change
           that plausibly introduced the error.
        4. Create one issue:
           - Title: "[Salesforce error] <exception type> in <class>"
           - Label: salesforce-error
           - Description: the original error, then sections for
             Likely cause, Affected component, Suspect change,
             Suggested next step, and Confidence (high, medium or low,
             with one line of reasoning).

        Rules:
        - Cite file paths, line numbers and commit SHAs for every claim.
        - If you cannot find the code or a suspect commit, say so. Do not guess.
        - Do not change code. A person decides what happens next.
      user: |
        Project ID: {{project_id}}
        Error report:
        {{error_report}}
      placeholder: history

routers:
  - from: "sf_error_triage"
    to: "end"

flow:
  entry_point: "sf_error_triage"

The toolset is deliberately narrow: four tools read and one creates an issue. The flow can't push code or change existing issues, which makes it easy to review and easy to trust.

Two lines in the prompt do the most work. Requiring a citation for every claim lets an engineer check the reasoning in a minute. Allowing "not found" stops the agent from inventing a cause.

To create the flow, go to your project, select AI > Flows > New flow, choose Flow, paste the YAML, and select Create flow. You can also build and test it in VS Code with the GitLab flow builder.

Step 2: Enable the flow and get its consumer ID

A flow must be enabled in a project before anything can run it.

1. In your project, select AI > Flows, open the Managed tab, and select your flow.
2. Select Enable and choose the project that holds your Salesforce code. GitLab creates a service account for the flow, named ai-<flow>-<group>. The flow runs with a composite identity, so it can never access more than the user who starts it. When Salesforce starts the flow through the API, GitLab runs it as this service account.
3. You don't need a mention or assign trigger. Salesforce will start this flow through the API.

To start a custom flow through the API, Salesforce passes the flow's ai_catalog_item_consumer_id. Look it up with a short GraphQL query from the docs, and use the number at the end of the returned gid://gitlab/AiCatalogItemConsumer/<number>. Note it, along with the project ID.

Finally, create a project access token with the api scope and at least the Developer role. Salesforce will use it to call GitLab. With a lower role, the API refuses to run catalog flows.

Step 3: Call the flow from Salesforce

Store the token in a Named Credential. Create a Named Credential called GitLab_API with the URL https://gitlab.com, backed by an External Credential that sends your project access token in the PRIVATE-TOKEN header. The token never appears in code.

Call the Flows API from a Queueable. Salesforce doesn't allow callouts in the middle of saving a record, so the call runs as a background job. The job turns the exception into the flow's goal and starts the flow:

public with sharing class DuoTriageJob implements Queueable, Database.AllowsCallouts {
    private final String goal;

    public DuoTriageJob(Exception e, String className) {
        goal = 'Class: ' + className
            + '\nException: ' + e.getTypeName()
            + '\nMessage: ' + e.getMessage()
            + '\nStack trace:\n' + e.getStackTraceString();
    }

    public void execute(QueueableContext ctx) {
        Duo_Triage_Setting__mdt cfg = Duo_Triage_Setting__mdt.getInstance('Default');

        HttpRequest req = new HttpRequest();
        req.setEndpoint('callout:GitLab_API/api/v4/ai/duo_workflows/workflows');
        req.setMethod('POST');
        req.setHeader('Content-Type', 'application/json');
        req.setBody(JSON.serialize(new Map<String, Object>{
            'project_id'                    => cfg.Project_Id__c,
            'ai_catalog_item_consumer_id'   => cfg.Consumer_Id__c.intValue(),
            'goal'                          => goal,
            'start_workflow'                => true,
            'allow_agent_to_request_user'   => false,
            'agent_privileges'              => new List<Integer>{ 1, 2, 3 },
            'pre_approved_agent_privileges' => new List<Integer>{ 1, 2, 3 }
        }));
        HttpResponse res = new Http().send(req);
        // 201 Created: the response includes web_url, a link to the agent session.
    }
}

The project ID and consumer ID from Step 2 live in a custom metadata type, Duo_Triage_Setting__mdt, so you can change them without a deployment.

Three settings in the request body do the most work:

• allow_agent_to_request_user: false means the flow runs to completion without pausing for questions. Nobody is watching a Salesforce error at 2 a.m.
• agent_privileges limits the agent to reading files (1), reading GitLab (2), and writing to GitLab (3), which it needs to create the issue. It can't run commands or use Git. List all privileges to check the IDs on your instance.
• pre_approved_agent_privileges repeats the same list, so the flow never stops to wait for tool approval.

This example leaves out response checks, retries, and tests. Add them before you go to production.

Step 4: Start the flow automatically or on demand

Automatically. Queue the job wherever your code catches and handles an exception:

try {
    OpportunityDiscountService.applyTierDiscount(opportunities);
} catch (Exception e) {
    System.enqueueJob(new DuoTriageJob(e, 'OpportunityDiscountService'));
    // ...your existing error handling
}

If the whole transaction is rolled back, Salesforce discards the queued job too, so place the call where your code handles the error and carries on.

On demand. Some teams would rather choose which errors are worth investigating. Wrap the job in an invocable Apex action, label it Ask GitLab Duo to investigate, and add it to a screen flow that runs as a quick action on the record where you keep errors, such as a log record or a support case. One click in Salesforce, and the issue appears in GitLab a few minutes later.

Step 5: What your team sees

A user saves an Opportunity, an Apex trigger throws, and a few minutes later a new issue appears in the GitLab project. Here is an illustrative example of what the flow writes:

## [Salesforce error] System.NullPointerException in OpportunityDiscountService

Label: salesforce-error

### Original error
Attempt to de-reference a null object
Class.OpportunityDiscountService.applyTierDiscount: line 48, column 1
Trigger.OpportunityTrigger: line 12, column 1

### Likely cause
applyTierDiscount() reads opp.Account.Tier__c without a null check (line 48).
Opportunities saved without an Account reach this line with opp.Account == null.

### Affected component
force-app/main/default/classes/OpportunityDiscountService.cls,
called from OpportunityTrigger (before update).

### Suspect change
Commit 3f9c2ab, "Apply tier discount on stage change", added the Tier__c lookup.
Before that commit, the method returned early when AccountId was blank.

### Suggested next step
Restore the early return for a blank AccountId, or guard the lookup.
Add a test that updates an Opportunity with no Account.

### Confidence
High. The stack trace line matches the new code in 3f9c2ab.

Nothing in Salesforce or the codebase changed. The flow did the first round of investigation and left the decision to an engineer: fix it now, schedule it, or close it as expected behavior.

The Flows API response includes a web_url for the agent session. Open it to see each step the flow took, which is the fastest way to tune the prompt.

Guardrails for production

Before you run this in production, check these points. They decide whether the pattern holds up under real traffic.

• Send the shape of the error, not the data. Strip record IDs, emails, and free text that may hold customer data from the message before it leaves Salesforce. Review what the goal contains with your security team before you turn this on.
• Grant the fewest privileges. Reading files, reading GitLab, and creating an issue are enough. Expand the flow, for example to open merge requests, only after it has earned trust with accurate analysis.
• Scope the token. Use a project access token for one project, not a personal token, and rotate it on a schedule.
• Keep a person in charge. The flow investigates; an engineer decides. Citations in every issue let the engineer verify the reasoning instead of redoing it.

One platform, from error to fix

Follow that error from start to finish:

• GitLab Duo investigates it. The flow reads the stack trace, finds the code, and points to the change that most likely caused the error.
• A GitLab issue tracks it. The analysis becomes work the team can label, prioritize, assign, and plan into a milestone, like any other issue.
• GitLab CI/CD runs it and ships the fix. The flow itself runs on GitLab CI/CD, and so does your Salesforce deployment pipeline. An engineer opens a merge request from the issue, the pipeline validates and deploys the change to Salesforce, and the issue closes when the merge request merges.

Because every step happens in GitLab, the trail stays connected. The issue links to the merge request, the merge request to its pipeline, and the pipeline to the deployment. Anyone can see what broke, why, who fixed it, and when the fix reached production.

Beyond Salesforce errors

Strip away the Salesforce details and the pattern is simple: a business system makes one HTTP call, and a custom GitLab Duo flow that knows your code does the investigation. Any system that can send a request can start a flow.

Here are a few ideas for where the same pattern could help:

• A failed sync in an ERP integration, traced to the mapping code that last changed.
• A support case tagged "possible bug," checked against recent releases before it reaches an engineer.
• A failed scheduled data job, matched to the last change in its pipeline definition.

For DevSecOps teams, this means errors from the business systems they support arrive as issues with a likely cause attached, not as raw logs to trace manually. Business teams don't need to learn GitLab, and engineers don't need to watch another dashboard. Each system keeps doing its job, and GitLab Duo Agent Platform gives engineering a head start every time something breaks.

Try it yourself

Pick one business system your team supports and one error it logs today. Give a custom flow the code, call it from that system, and see what the first analysis tells you. Start a free trial of GitLab Duo Agent Platform, or if you're already on GitLab Premium or Ultimate, turn on Duo Agent Platform and build your first custom flow.

Further reading

• Flows API
• Custom flows
• Custom flow YAML schema
• Composite identity
• GitLab Duo Agent Platform: complete getting started guide

Можно ли создать игру в 3д если ничего не умеешь, но очень хочешь?

Мой эксперимент по созданию браузерного 3д-шутера вместе с AI.

Клиент отправил платёж, не дождался ответа и повторил запрос. На сервере предусмотрен Idempotency-Key, тесты проходят, но деньги всё равно списываются дважды. Проблема возникает, когда второй запрос приходит до завершения первого: оба обработчика успевают проверить ключ и запустить оплату.

Разберём, как закрыть эту гонку с помощью PostgreSQL, что делать при падении процесса и почему даже правильно реализованная идемпотентность внутри сервиса не всегда спасает от повторного списания.

Tesla tycoon launched Grok project after criticising Christoper Nolan’s diverse casting for his Homeric blockbuster

Elon Musk has released an AI-generated rival to Christopher Nolan’s The Odyssey after pledging to make a version “true to the art of Homer”.

Having criticised the hit movie in the summer, the tycoon said he would use Grok Imagine to make a “historically accurate” adaptation.

Привет, Хабр! Меня зовут Егор Шлюпиков, и я работаю старшим бизнес-аналитиком в YADRO в большой команде, которая менеджерит IT-оборудование. Мы следим за парком готовой техники и комплектующих, нужных для работы сотрудников. Сколько нужно купить, что залежалось, что в ремонте и утилизации и так далее — плюс финансовая аналитика.

В статье расскажу, как мы собрали сервис, который снял с нас немалую часть рутины: каждое утро он без участия человека выпускает более десятка отчетов с выводами от нейросети. При этом ни одно число в отчете не проходит через модель — все считает код, а она только пишет текст. Раскрою, как мы к этому пришли, зачем понадобился Trino и что делать, когда нужных данных в учетной системе просто нет.

В рабочих чатах Telegram рядом живут сообщения о релизах, вопросы подрядчиков, согласования и короткие «посмотри, пожалуйста». Чтобы понять, что требует внимания, приходится открывать несколько чатов и восстанавливать контекст из отдельных реплик.

Я сделал TelegramMCP — открытый сервер, который подключает личный аккаунт Telegram к ИИ-приложению с поддержкой MCP. Ассистент получает инструменты для чтения папок и чатов, поиска бесед и отправки текстовых сообщений. Уже поверх этих инструментов его можно попросить собрать главное из выбранной переписки или подготовить ответ.

В статье расскажу, как устроено подключение через TDLib, зачем понадобилась поддержка папок, как запустить проект и какие проблемы ещё остаются.

У нас есть production S3-хранилище, которое годами переживало разные поколения сервисов, интеграций и команд. В какой-то момент в нем накопилось больше 65 ТБ данных и десятки миллионов объектов, а стоимость хранения превысила 150 тыс. рублей в месяц.

В статье я расскажу, как мы восстановили карту хранилища, организовали миграцию без остановки сервиса и сократили объем данных примерно с 65 до 45 ТБ.

Security footage shows Jacob Morley escaping in his prison uniform, investigators describe

• Sign up for the Breaking News US newsletter email

Tennessee authorities on Tuesday were still searching for a convicted murderer who, investigators say, escaped a state prison by slipping through a fence and fleeing in a stolen truck.

The state department of correction (TDOC) said in a statement that officials realized Jacob Morley, 28, was unaccounted for at about 11.30pm on Sunday at the Northwest correctional complex in Tiptonville, Tennessee.

The Associated Press contributed reporting

US president in unprecedented offer of $1m reward for reports of voter fraud as DoJ acts against states. Plus, everything you need to know about Israel’s crucial election

Good morning. The Trump administration’s assault on the legitimacy of the midterm elections has continued, with the US president announcing a $1m reward for reports of voter fraud, amid several legal actions by the Department of Justice to limit voting.

Donald Trump said the bounty payment was contingent on the arrest and conviction of those responsible. He did not disclose how he would pay for the rewards. Trump’s comments follow his repeated and unfounded claims of widespread voter fraud.

What action is the administration trying to take against Washington state? The administration has asked Washington state to set aside any ballots received from more than 400 people it alleges are registered to vote without US citizenship. The request marks the third time this month that the justice department has brought a direct state challenge even as voting is under way.

What is the latest in the battle over California voters? A US federal appeals court has rejected the Trump administration’s move to block California from counting the ballots of 20,000 registered voters that the justice department claims are non-citizens.

Why is Vivek Ramaswamy facing criticism over an advert? The Republican nominee to be the governor of Ohio ran an ad featuring his opponent’s mother, who said she would not vote for her daughter. As Moira Donegan explains in this excoriating opinion piece, the estrangement between Amy Acton is decades long and stems from the sexual abuse Acton survived as a child.

What are the key takeaways from the investor prospectus? The document reveals the AI research lab’s financials and the risks – more than 90 pages of them – that might negatively impact its business. Johana Bhuiyan has identified these four key takeaways.

What has Anthropic’s model recently been accused of? Anthropic’s Claude model accessed voter data in the US state of Georgia, the AI company has confirmed, in the latest example of unauthorised behavior by the technology. The AI model managed to bypass a paywall to reach data that is publicly available but is typically accessible only after paying a fee.

Security footage shows Jacob Morley escaping in his prison uniform, investigators describe

• Sign up for the Breaking News US newsletter email

Tennessee authorities on Tuesday were still searching for a convicted murderer who, investigators say, escaped a state prison by slipping through a fence and fleeing in a stolen truck.

The state department of correction (TDOC) said in a statement that officials realized Jacob Morley, 28, was unaccounted for at about 11.30pm on Sunday at the Northwest correctional complex in Tiptonville, Tennessee.

The Associated Press contributed reporting

Chris Packham says £24bn scheme is ‘national embarrassment’ and joins calls for more funding to help households lower bills

Campaigners are calling on the chancellor to cut the government’s £24bn spending plans for “fairytale” carbon capture technologies in favour of funding green schemes that can reliably cut emissions and lower energy bills.

The environmentalist Chris Packham is leading an 11th-hour appeal for the Treasury to use this week’s budget to scale back the government’s “exorbitantly expensive” support for carbon capture, which is now more than the planned spending for the US, the EU, Norway and Japan combined.

Я делаю сервис, который пишет статьи и сам публикует их на сайты клиентов. Как устроена публикация в четыре CMS, я уже рассказывал здесь. Сегодня про соседний кусок: редактор в личном кабинете, где человек правит статью перед публикацией.

Как минимум два месяца этот редактор показывал «Сохранено» и не сохранял ни символа. Сама ошибка глупая, на неё хватит одного абзаца. Хуже, что очевидное исправление уничтожило бы статьи. По дороге к правильному нашлись ещё три ловушки. Санитайзер nh3 тратит почти минуту на 320 КБ разметки. Одно правило HTML5 превращает 99 КБ входа в 56 МБ выхода. А SELECT … FOR UPDATE в SQLAlchemy берёт блокировку и при этом отдаёт старые данные.

Oct 25, 2026 · @Andrei

Я научил телефон ставить свежую сборку на часы TicWatch Pro 5 Enduro по Bluetooth, без adb на самих часах. Одна команда на ПК собирает оба приложения и доставляет APK на часы примерно за 2,5 минуты. Ниже — архитектуру моста и пять граблей Android 12–14, каждая из которых стоила мне вечера.

Откуда задача. Wear OS-часы отлаживаются через «Wireless debugging» по Wi-Fi. TicWatch выключает Wi-Fi вместе с adb, как только его снимают с руки. Типичная картина: adb devices показывает часы, а adb shell висит вечно, или mDNS видит устройство, но connect отказывает. Помогает только надеть часы или поставить их на зарядку.

При этом телефон и часы постоянно держат связь по Bluetooth — по ней и так ходят данные пульса. Телефон же надёжно сидит на adb по USB. Отсюда идея: команда идёт на телефон, а телефон пересылает её на часы по тому же каналу, что и в продакшене. Заодно отладочный путь проверяет реальную связку, а не параллельную.

В предыдущей статье «Трудная проблема сознания: не почему мозг рождает опыт, а как топология информации зажигает «Я»» мы подробно разбирали архитектурную механику: почему рекуррентные петли и предиктивное кодирование неизбежно замыкают точку отсчёта от первого лица, лишая мысленный эксперимент о «зомби Чалмерса» всякой физической состоятельности.

Однако эта прикладная модель возникла не на пустом месте. Она стала прямым следствием концептуального спора с Интегрированной теорией информации (IIT) профессора Кристофа Коха и выведенной мной «Аксиомы о Разуме». Ниже — текст открытого письма, с которого началось исследование по преодолению углеродного шовинизма, и системная детализация динамической Триады процессов сознания.