Commit 0cddec03 authored by Falko Sieverding's avatar Falko Sieverding 🔰
Browse files

Add DAP Rapid field issue reporting guide for SAs

parent 5bc97cc5
Loading
Loading
Loading
Loading
+7 −2
Original line number Diff line number Diff line
@@ -23,6 +23,10 @@ For all other situations with existing customers, Solutions Architects should ad

If a ticket later becomes a priority, the assigned Support Engineer can escalate it to the responsible backend engineering team through a [GitLab Request for Help](https://gitlab.com/gitlab-com/request-for-help/) (RFH) issue. In exceptional circumstances such as a blocker that could prevent closing business, Solutions Architects may raise an RFH issue directly. If doing so, please ensure you carefully follow the documented process in the [RFH README](https://gitlab.com/gitlab-com/request-for-help/-/blob/main/README.md) and, when closing the issue, apply the correct [closure label](https://gitlab.com/gitlab-com/request-for-help/-/blob/main/README.md#applying-the-correct-closure-labels) to align with reporting requirements.

### DAP Rapid — Reporting Customer-Blocking Issues

When SAs discover bugs or customer-blocking issues during DAP trials, POVs, or customer deployments, use the [DAP Rapid field reporting process](/handbook/solutions-architects/tools-and-resources/dap-issue-reporting/) to route them to engineering with appropriate severity and tracking labels. SAs can file bugs directly — no Zendesk ticket required.

### SaaS Trials for Existing Customers

Paid customers must set up a new namespace on SaaS when trialing Ultimate.  There are two primary reasons for this:
@@ -72,13 +76,14 @@ Solutions Architects frequently interact with customers for demos, presentations
- [Rectangle](https://rectangleapp.com/) to quickly move and resize windows in macOS using keyboard shortcuts or snap areas
- [Dark Reader](https://darkreader.org/) enables browser dark mode to better fit room lighting
- [Postman](https://www.postman.com/) for API interaction
- [Visual Studio Code](https://code.visualstudio.com/) lightweight IDE text editor
- [VSCodium](https://vscodium.com/) open-source IDE preferred by some SAs for demos and development
- [Visual Studio Code](https://code.visualstudio.com/) lightweight IDE text editor

#### Related macOS tips

- [Switch between full screen applications](https://www.intego.com/mac-security-blog/how-to-enter-and-exit-full-screen-mode-in-macos/) using the trackpad, Command keys or other options
- Use the [Zoom Accessibility Features](https://www.imore.com/how-use-zoom-mac) to zoom in on targeted screen locations
- [Work in multiple spaces on a single monitor](https://support.apple.com/en-gb/guide/mac-help/mh14112/mac) to keep multiple app windows or browser tabs open in fullscreen mode
- [Work in multiple spaces on a single monitor](https://support.apple.com/en-gb/guide/mac-help/mh14112/mac) to keep multiple app windows or browser tabs open in fullscreen mode
  - Enables switching between windows or tabs with trackpad gestures, keeping display screen clean and uncluttered

### Useful Customer Facing Presentations
+115 −0
Original line number Diff line number Diff line
---
title: DAP Rapid — Field Issue Reporting
description: How Solutions Architects report customer-blocking DAP issues to engineering
---

## Overview

**DAP Rapid** is the Duo Agent Platform program for handling Sev1 and Sev2 customer-blocking issues with clear SLOs, tracking, and escalation paths across 14 contributing engineering groups.

As a Solutions Architect, you can:

- **File bugs directly** in GitLab issues — no Zendesk ticket required
- **Propose severity labels** (engineering makes the final call)
- **Escalate critical issues** through RFH or incident.io

Use this process whenever you discover bugs or customer-blocking issues during DAP trials, POVs, or customer deployments.

## When to report

Report issues that block or significantly impair a customer's use of the Duo Agent Platform.

| Severity | Description | Examples |
|----------|-------------|----------|
| **Sev1 — Critical** | System down, major functionality broken, customer blocked | Agent cannot start, tool execution fails for all users, data loss |
| **Sev2 — Major** | Significant functionality impaired, workaround available | Specific tool fails intermittently, degraded performance, incorrect output with known workaround |

For Sev3/Sev4 issues (minor bugs, cosmetic problems), file a standard bug issue with the labels below — the DAP Rapid SLOs do not apply.

## How to report

Choose the channel based on severity and urgency:

```text
Customer-blocking issue discovered

    ├─ Sev1 (system down, no workaround)
    │   └─ Raise incident through incident.io OR RFH
    │       → Escalation required if not resolved in 12h

    ├─ Sev2 (impaired, workaround exists)
    │   └─ File bug in gitlab-org/gitlab OR raise RFH
    │       → Escalation if not resolved in 24h

    └─ Sev3/4 (minor)
        └─ File bug in gitlab-org/gitlab with labels
```

### Channel 1: Direct bug filing (most common)

File an issue in [`gitlab-org/gitlab`](https://gitlab.com/gitlab-org/gitlab/-/issues/new) with the required labels listed below. This is the fastest path for field-discovered bugs.

### Channel 2: Request for Help (RFH)

Use [GitLab Request for Help](https://gitlab.com/gitlab-com/request-for-help/) when you need engineering to investigate a support-adjacent issue, or when the root cause is unclear. Follow the RFH README and apply the correct closure labels.

### Channel 3: incident.io (critical production incidents)

For Sev1 production incidents on GitLab.com, [declare an incident using incident.io](/handbook/engineering/infrastructure-platforms/incident-management/#reporting-an-incident) and ensure the "Affects Duo Agentic Platform (DAP)" field is set to YES.

## Required labels

When filing a bug issue in `gitlab-org/gitlab`, apply:

| Label | Purpose |
|-------|---------|
| `~"Duo Agent Platform GA Fast Follow"` | Routes the issue into DAP tracking |
| `~"type::bug"` | Bug identifier |
| `~"severity::1"` through `~"severity::4"` | Proposed severity (engineering makes final call) |
| `~"group::[team name]"` | Owning AI engineering group |

## Resolution SLOs by deployment type

These are internal targets (SLOs), not customer-facing SLAs. The primary goal is to **unblock customers within 24 hours** for all Sev1 and Sev2 issues.

| Deployment | Severity | Target unblock time | Fix/patch deployment |
|------------|----------|---------------------|----------------------|
| **GitLab.com (SaaS)** | Sev1 | 30 min – 12 hours | Immediate (deploy within ~6h of MR merge on business days) |
| | Sev2 | 4 – 24 hours | Target: 3 business days |
| **Self-Managed** | Sev1 | 30 min – 12 hours | Next [patch release](/handbook/engineering/releases/patch-releases/#patch-release-overview) (~2 weeks, 2nd/4th Wednesday) |
| | Sev2 | 4 – 24 hours | Next patch release |
| **Dedicated** | Sev1 | 30 min – 12 hours | Next patch release with [backporting](https://docs.gitlab.com/policy/maintenance/#backporting-to-older-releases) (2–4 weeks) |
| | Sev2 | 4 – 24 hours | Next patch release with backporting |

**Notes:**

- Configuration or setup issues should be unblocked within the target time without a code change.
- Code fixes requiring deployment follow the timelines above.
- For Self-Managed and Dedicated, communicate fix timeline to the customer within 12h (Sev1) or 24h (Sev2).
- Dedicated customers run one version behind; backporting is limited to the previous two minor versions.

## Tracking dashboards

| Tracker | Link | Owned by |
|---------|------|----------|
| Zendesk DAP issues | [Zendesk dashboard](https://gitlab.zendesk.com/explore/studio#/dashboards/65C06156A78CBD32904C3E92205D82AC4A6B49D51D5D149FF5D324C97ADAD8F6) | Support |
| incident.io DAP issues | [incident.io dashboard](https://app.incident.io/gitlab/incidents) | Engineering & Support |
| Bug tracker (GitLab issues) | [DAP Bug Tracker wiki](https://gitlab.com/groups/gitlab-org/ai-powered/-/wikis/home/Program-Duo-Agent-Platform/Program-Bug-Tracker) | AI EM & PM |
| RFH tracker | [DAP RFH Tracker wiki](https://gitlab.com/groups/gitlab-org/ai-powered/-/wikis/home/Program-Duo-Agent-Platform/Program-Bug-Tracker/DAP-RFH-Tracker) | Support Engineering |
| Reliability (Grafana) | [DWS log-based dashboard](https://dashboards.gitlab.net/d/3d9c7954-2669-4782-9206-b714c8a589fa/dws-log-based-dashboard) | Engineering |
| Latency (Snowflake) | [Language server metrics](https://app.snowflake.com/ys68254/gitlab/#/language-server-metrics-dd7LWVgnL) | Engineering |
| Tool call error rate | [Duo Workflow Service dashboard](https://dashboards.gitlab.net/d/duo-workflow-svc-main/duo-workflow-svc3a-overview) | Engineering |

## Slack channel

Use [`#f_duo-agent-platform`](https://gitlab.slack.com/archives/f_duo-agent-platform) as the consolidated channel for all DAP inquiries — questions, issue triage, and status updates.

Do **not** use `#building-on-duo-agent-platform`, `#usage-billing-help`, `#dap-customer-feedback`, or individual AI group channels for issue reporting. Redirect to `#f_duo-agent-platform`.

## Related resources

- [Support Super Form](https://support-super-form-gitlab-com-support-support-op-651f22e90ce6d7.gitlab.io/) — request Support assistance during pre-sales
- [RFH process](https://gitlab.com/gitlab-com/request-for-help/) — Request for Help from engineering
- [AI POV scope and acceptance](/handbook/solutions-architects/tools-and-resources/pov/ai/) — running DAP customer trials
- [GitLab Support SLAs](https://support.gitlab.com/hc/en-us/articles/11626483177756-GitLab-Support) — official customer-facing SLAs
- [Incident escalation process](/handbook/engineering/infrastructure-platforms/incident-management/tier2-escalations/) — PagerDuty / incident.io escalation
+8 −4
Original line number Diff line number Diff line
@@ -39,13 +39,13 @@ We need to proactively identify languages, testing questions and teams with them
    - Invitation to Office Hours
- Bias for guided, hands-on activities. See POV activities for inspiration
- When POV goals are to surface quantitative indicators of improvements brought by AI-powered features, it is tempting to adopt a comparative study approach for the workshops or hackathons described above.
- It consists of assigning the same task to two groups of developers. One group works as usual, while the other is allowed to use AI-powered features. Optionally, a second assignment is given, and the two groups are switched around.
- A drawback of this approach is that working without AI-powered features isnt very fun. Considering how precious it is in enterprise settings to have time dedicated to this kind of initiative, spending it on working as usual may not be ideal.
- It consists of assigning the same task to two groups of developers. One group works "as usual," while the other is allowed to use AI-powered features. Optionally, a second assignment is given, and the two groups are switched around.
- A drawback of this approach is that working without AI-powered features isn't very fun. Considering how precious it is in enterprise settings to have time dedicated to this kind of initiative, spending it on "working as usual" may not be ideal.
- As an alternative, ask participants to log, for each task
  - Before: a time estimate if they were working as usual
  - After: the time it actually took, with the help of AI-powered features.
  - After: include qualitative feedback about the comfort of working, quality of software, etc.
- Familiarize yourself with [Duo Chat best practices](https://about.gitlab.com/blog/2024/04/02/10-best-practices-for-using-ai-powered-gitlab-duo-chat/) and [Duo Code Suggestions top tips](https://about.gitlab.com/blog/2024/06/11/top-tips-for-efficient-ai-powered-code-suggestions-with-gitlab-duo/). There are some great tips and tricks in this blog to assist with integrating Duo Chat and Code Suggestions into a customers workflow. Check the documentation for more practical [GitLab Duo uses case](https://docs.gitlab.com/ee/user/gitlab_duo/use_cases.html).
- Familiarize yourself with [Duo Chat best practices](https://about.gitlab.com/blog/2024/04/02/10-best-practices-for-using-ai-powered-gitlab-duo-chat/) and [Duo Code Suggestions top tips](https://about.gitlab.com/blog/2024/06/11/top-tips-for-efficient-ai-powered-code-suggestions-with-gitlab-duo/). There are some great tips and tricks in this blog to assist with integrating Duo Chat and Code Suggestions into a customer's workflow. Check the documentation for more practical [GitLab Duo uses case](https://docs.gitlab.com/ee/user/gitlab_duo/use_cases.html).

## Pre-requisites

@@ -190,6 +190,10 @@ After the workshop
- Ask the customer POV lead about the feedback they had internally
- Give feedback about the workshop, in the appropriate GitLab project or Slack channel

### Reporting issues during trials

When you discover bugs or customer-blocking issues during a DAP trial, follow the [DAP Rapid field reporting process](/handbook/solutions-architects/tools-and-resources/dap-issue-reporting/). SAs can file bugs directly in GitLab issues with severity labels — no Zendesk ticket required. For issues that meet the [definition](/handbook/engineering/infrastructure-platforms/incident-management/#incident-management) of an incident, [declare an incident using incident.io](/handbook/engineering/infrastructure-platforms/incident-management/#reporting-an-incident) and ensure the "Affects Duo Agentic Platform (DAP)" field is set to YES.

### Results

### Tracking
@@ -272,7 +276,7 @@ While this could typically be interpreted as "failing" in a regular context, it
  - Mention that this is expected, and inherent to Gen-AI. It is neither a bug that will be fixed nor something specific to GitLab's Gen-AI features.
- As a presenter, be prepared for these "happy incidents"
  - Be prepared to brush it off of laugh it off. One wrong suggestion is not a big deal. After all, it's not a true demo until something goes wrong.
  - Use them as opportunities to adopt [pioneer/learning/playful mindset](https://gitlab.com/gitlab-com/content-sites/handbook/-/edit/pdumaitre-main-patch-06406/content/handbook/solutions-architects/tools-and-resources/pov/ai.md?from_merge_request_iid=3693#elements-of-perspective-and-communication-on-gen-ai-features), or convey the messaging that AI is meant as an assistant to development teams.
  - Use them as opportunities to adopt pioneer/learning/playful mindset, or convey the messaging that AI is meant as an assistant to development teams.
    > "If you want to dig into this later, I'm sure we'll be able to find the right prompt to make this work"
    > "Good thing I'm following best practices and reviewing AI suggestions"
    > "Glad to see I'm not obsolete just yet!"