Skip to main content

​
AI-DLC user flow

AI-DLC is designed so that planning, execution, and handoff happen through visible artifacts instead of hidden chat context.

​
Typical journey

1

Describe the outcome you want

You start with a workspace-level request such as a new capability, a major enhancement, or a cross-project change.
2

Let Inception shape the plan

The Inception agent creates an intent, decomposes it into unit artifacts, writes story artifacts, and groups related stories into bolt artifacts.
3

Review the planned execution shape

You review the plan as documents: intent summary, unit boundaries, user stories, and bolt groupings.
4

Select a bolt for construction

Construction works against one planned bolt. It can refine the execution plan, but it stays grounded in the bolt and its associated stories.
5

Watch Construction create working docs while coding

As Construction works, it creates or refines design and execution artifacts such as domain_design, logical_design, system_context, implementation_plan, test_report, and walkthrough.
6

Review the evidence

You read the design docs, the execution walkthrough, and the verification report as document-style artifacts, not raw arrays.
7

Hand off to operations when needed

If the bolt changes deployment-relevant behavior, AI-DLC emits or updates a deployment_unit so Operations has a structured handoff artifact.

​
What you should expect in each phase

​
Inception

Inception is where AI-DLC turns a broad request into an executable plan. You should expect:
  • one root intent
  • one or more unit artifacts
  • one or more story artifacts under each unit
  • one or more bolt artifacts that group related stories into planned execution batches

​
Construction

Construction works against a planned bolt. You should expect:
  • code changes tied to the selected bolt
  • design outputs such as domain_design and logical_design
  • execution outputs such as walkthrough and test_report
  • an explicit record of touched workspace projects

​
Operations

Operations works from a structured handoff rather than reconstructing intent from chat history. You should expect:
  • a deployment_unit
  • deployment notes
  • rollout stage tracking
  • verification state
  • incidents or mitigation notes when relevant

​
What the UI should feel like

AI-DLC should feel like three coordinated surfaces, not a pile of disconnected artifact tabs. 
+----------------------+--------------------------------------------------------------+
| Sidebar              | Main workspace area                                          |
|                      |                                                              |
| [AI-DLC] [Agents]    | [Dashboard] [Artifact Viewer tabs...]                        |
| Group by: Status     |                                                              |
|                      | Dashboard                                                    |
| Draft                | +-------------+-------------+-------------+                  |
|   INT-001            | | Draft       | In review   | Approved    |                  |
|   STORY-014          | | cards...    | cards...    | cards...    |                  |
| In review            | +-------------+-------------+-------------+                  |
|   BOLT-004           |                                                              |
| Approved             | Artifact viewer                                             |
|   TEST-002           | +--------------------------------+ +----------------------+ |
|                      | | Selected artifact document     | | Context rail         | |
|                      | | markdown page + meta row       | | tree, relations,     | |
|                      | | read-only document body        | | chats, activity      | |
|                      | +--------------------------------+ +----------------------+ |
+----------------------+--------------------------------------------------------------+
Use the surfaces like this:
  • sidebar for global find, grouping, and quick opening
  • dashboard for status-oriented overview across the workflow instance
  • artifact viewer for deep work in one root context

​
What the artifact viewer should feel like

When AI-DLC opens an artifact in context, the reading experience should stay very plain:
  • title first
  • a minimal metadata row under the title
  • markdown sections rendered like a document preview
  • a slim right rail for context, not a second document
The right rail is where AI-DLC keeps:
  • the root tree
  • relations for the selected artifact
  • linked chats
  • activity, collapsed by default
This keeps the document body focused on the artifact itself instead of mixing narrative content with workflow chrome.

​
How navigation should resolve context

The important behavior is not “open the clicked artifact in isolation.” It is “open the clicked artifact in context.” Fabriqa should resolve a root viewer context like this:
  1. Start from the clicked artifact.
  2. Walk up its parent chain.
  3. Stop at the first ancestor whose parent is in a different phase, or at the top root if there is no phase boundary.
  4. Open or focus the viewer tab for that root artifact instance.
  5. Highlight the clicked artifact inside the tree and show its document on the left.
In AI-DLC that gives intuitive results:
  • click a story from the sidebar: open the surrounding intent context and select the story
  • click a walkthrough: open the surrounding bolt context and select the walkthrough
  • click a test_report: stay inside the same bolt execution context rather than opening a disconnected tab
Sidebar click: STORY-014
  -> viewer root: INT-001
  -> open/focus INT-001 tab
  -> tree selects STORY-014

Dashboard click: WALKTHROUGH-003
  -> viewer root: BOLT-004
  -> open/focus BOLT-004 tab
  -> tree selects WALKTHROUGH-003

​
How AI-DLC handles change requests

AI-DLC is not a one-way conveyor belt. If you change your mind during construction:
  • Construction can refine the current bolt if the change still belongs in that scope.
  • Construction can create follow-up planning artifacts if the change expands scope.
  • Construction can effectively re-enter planning behavior, but it should do so through workflow artifacts instead of hidden reasoning.
The important rule is that the new scope still becomes explicit artifacts. It does not live only inside the chat transcript.

​
Prompt-driven, not automatic

AI-DLC still relies on agent prompts to decide:
  • when a bolt is ready to execute
  • when a design document is necessary
  • when an implementation plan should be written
  • when a walkthrough or test_report must be created
The workflow definition makes those artifact contracts explicit, but it does not auto-run the flow.

​
Continue reading

AI-DLC overview

Go back to the AI-DLC overview page.

AI-DLC artifacts

Continue into the artifact model and execution artifact structure.

AI-DLC definition

See the same flow expressed as a Workflow DSL definition.

Workflow glossary

Review the shared terms used across AI-DLC and the Workflow DSL.