Contextual data
The Contextual data panel in the Workspace brings every mental model related to an entity into one place. It does not distinguish between where a model lives in architxt — plain, derived, or contextual — it simply shows what is available for the selected entity and lets you open it.
What the panel shows
The panel lists entities that have at least one attached mental model. Each row is an entity; expanding the row reveals the models attached to it. The same entity can appear because it has:
- A plain mental model stored directly on the entity in architxt.
- A derived mental model produced by a template role applied to the entity's catalog record.
- A contextual mental model generated from the contextual graph node.
- An edge-context model that explains relationships between this entity and others.
All of these are grouped under the entity's identity. The panel uses the entity name from the catalog or contextual graph, and shows the underlying entity ID in a monospace label so you can always see which object the models belong to.
Model scope badges
Each attached model shows a scope badge that tells you how it is attached:
| Badge | Meaning |
|---|---|
| PLAIN | A plain mental model stored directly on the architxt entity. |
| DERIVED | A template-derived mental model from the entity's catalog record. |
| NODE | A contextual mental model generated for this graph node. |
| EDGE | An edge-context model describing a relationship to another entity. |
| SEED | A discovery/seed contextual model around this entity. |
The badge tells you how the model was produced; the role label next to it tells you what kind of model it is — for example, summary, capabilities, edge context, or a custom template role.
Filtering and search
The panel header contains controls that help you find the right model quickly when the entity list is large.
| Control | Purpose |
|---|---|
| ALL / NODES / EDGES | Scope filter. ALL shows every attached model. NODES limits the list to entity-level models (plain, derived, and node/seed scopes). EDGES shows only edge-context models that describe a relationship. |
| Search box | Freetext filter across entity names, entity IDs, scope badges, role labels, and model titles. Typing multiple words performs an AND search, so acme products only keeps rows that contain both words somewhere in their visible fields. |
| Counter | Shows the total number of items in scope and how many match the current filter, for example 12 (4). |
The search applies to every item in the panel, including models inside collapsed entities. If an entity has no matching models under the current filter, it is hidden entirely.
Deduplicated edge-context items
Edge-context models can be discovered from both endpoints of a relationship. The panel deduplicates these so a single edge appears only once, sorted with the alphabetically earlier endpoint first. This keeps the list compact and avoids mirror duplicates when both entities are in scope.
Interaction model
| Action | What happens |
|---|---|
| Click a model | Opens the model content in the active Preview tab (or replaces the current view tab). |
| Open in new tab | Opens the model in a new read-only View tab, so you can keep the current tab open. |
| Expand / collapse entity | Shows or hides the attached models for that entity. |
| Filter by scope or search | Narrows the visible items without changing what is attached to each entity. |
Why this matters
In other parts of architxt, mental models are managed in different places:
- Plain and derived models are created and edited on the main Mental Models page and in the Hindsight sync flows.
- Contextual models are produced automatically by the Context Patches sync job.
- Edge-context models are built from the relationships in the contextual graph.
The Contextual data panel is the only place where all of these are brought together under the entity they describe. This makes it the right surface for:
- Checking whether an entity has a summary before running a Reflect query.
- Comparing how the same entity is described by different model types.
- Opening a model in a new tab for side-by-side reading.
- Finding the edge context for a relationship you want to include in a curated page.
Example
Consider an entity labeled Acme Corp.
| Model | Scope | Role | How you see it |
|---|---|---|---|
Acme Corp summary | DERIVED | sys_entity_summary | From the catalog-derived models. |
Acme Corp capabilities | NODE | sys_entity_capabilities | From the contextual graph node. |
Acme Corp → Products division | EDGE | sys_edge_context | From a graph relationship. |
Acme Corp strategy | PLAIN | Custom template role | Created directly in architxt. |
All four rows appear when you expand Acme Corp in the Contextual data panel, even though they came from different sources.
Relationship to other surfaces
- Context Patches → Mental Models tab shows every contextual/role-based model for the bank at once. The Contextual data panel filters that same information to a single entity and mixes in plain and derived models. That tab now also supports scope filters (All / Empty / Failed) and search, which mirrors the filtering behavior in the Contextual data panel.
- Mental Models (main page) is where you create and edit plain and derived models; their content flows here.
- Workspace → Preview / View tabs is where you land when you click or open a model from this panel.
- Workspace → Chat reuses the same contextual-item list, scoped to the entities the agent resolved in a reply.
- Curated pages let you copy content from previewed models into a composed page.
Summary
The Contextual data panel is the collation point for everything an entity knows. It gathers plain, derived, contextual, and edge-context models under one identity, lets you preview them in place or open them in a new tab, and connects directly to the rest of the Workspace for reading and composition.