Features are the product building blocks that Aira generates from your insights. The roadmap is the visual representation of those features over time.
What features are
A feature in Aira is a scoped unit of work derived from one or more knowledge atoms (insights). Each feature references the specific atoms it was synthesized from, maintaining a traceable chain from raw source material to planned work.
Features include:
- Title and description — What the feature does
- Priority — Based on RICE scoring (Reach, Impact, Confidence, Effort)
- Status — backlog, planned, in_progress, done, shipped
- Effort estimate — How much work it represents
- Rationale — Why this feature matters, grounded in specific insights
- Linked atom IDs — The atoms that justify this feature's existence
Generating features with AI
From the insights page or the onboarding flow, click "Generate Features" to trigger AI feature generation.
How generation works
Aira uses iterative generation — features are created one at a time, not in a single batch:
- The AI generates one feature proposal based on project context and available insights
- The feature is immediately persisted and streamed to your browser
- The AI assesses coverage: are all significant insights addressed? What gaps remain?
- If gaps exist, it generates the next feature with full context of what came before
- This continues until the AI determines the roadmap is complete or hits a safety limit
Each feature appears in the UI as it's generated with a fade-in animation. You see progress as it happens, not a spinner for 60 seconds.
The quality gate
Generated features pass through a quality gate (Evaluator → Reviser loop):
- Evaluator — Assesses the feature against scoping criteria: is it well-defined? Does it address real insights? Is the effort estimate realistic?
- Reviser — If the evaluation fails, the content is revised based on feedback
- This loops up to a configured maximum (typically 1 iteration for features)
Artifact stability states
Features carry an explicit stability state:
| State | Meaning |
|---|---|
| draft | Generated with incomplete evidence coverage. Not used for downstream planning. |
| preview_stable | Preview-lane quality bar met. Suitable for onboarding progression but marked as refining. |
| stable | Deep-lane validation complete. Used for sprint planning and automation. |
The roadmap view
The roadmap page offers two views:
Gantt chart
A timeline visualization with 6 zoom levels (day, week, 2-week, month, quarter, year). Features appear as horizontal bars that you can:
- Drag to change start dates
- Resize to adjust duration
- Reorder by dragging vertically
List view
A sortable table of all features with columns for status, priority, effort, and dates.
Feature detail dialog
Click any feature to open its detail dialog. Here you can:
- Edit the title, description, and status
- View and modify priority and effort
- See the linked atoms that justify the feature
- View tasks broken down from this feature
- Track the artifact stability state
Manual feature creation
You can create features manually from the roadmap page. Manual features don't have linked atoms — they exist as standalone planned work. You can later associate them with insights if relevant.
Feature lifecycle
backlog → planned → in_progress → done → shipped
When all tasks under a feature are complete, the feature can move to done. When deployed, it moves to shipped, and any linked insights transition to implemented status.