A FrameGraph applies a chosen analytical framework over a topic graph to produce a navigable, sourced, reusable view of what is known about a subject. Where search returns a list and a copilot returns one answer, a FrameGraph keeps the structure of understanding: each angle on the subject, drillable, attributed, and accumulating into a knowledge cluster you can extend, export, and turn into a governed knowledge base.
Ask a search engine and you get a list to reconcile yourself. Ask a copilot and you get one fluent paragraph that collapses a rich subject into a single narrative. Both are useful, and both discard the thing a person actually reasons with: the different angles on a subject, how they connect, and where each claim comes from.
When you genuinely want to understand something, whether a sport, a market, a regulation, or a competitor, you do not want one answer. You want to know who the actors are, what is at stake, when the key moments fell, where it happens, why it works the way it does, and how it operates, and you want to move between those angles, follow a thread, and keep what you find. A FrameGraph is built around that way of thinking rather than against it.
FrameGraph is both a method and the artifact it produces. The method, framegraphing, takes any starting point and lays a selectable framework across it. The artifact, the FrameGraph, is the navigable result.
The topic graph is the layer of entities and relationships beneath the subject: its people, organisations, places, events, and concepts, and the links between them. A framework is a set of aspects, the angles you choose to read the subject through. 5W1H (who, what, when, where, why, how) is the default, but PESTEL, SWOT, the ecosystem and five forces, jobs to be done, a decision frame, and an ontology view are all selectable, and the same subject can be re-read through any of them. Every aspect is generated, sourced, and given a default visualisation, and as you drill and retrieve, the entities you touch accumulate into a knowledge cluster you can save, filter, and grow.
The pipeline is the same whatever the subject or the framework.
A quick pass over the topic, url, document, or prompt to size the subject and recommend a framework.
Pick a framework, or accept the recommendation. Its aspects become the angles the subject is read through.
Each aspect is generated from the chosen source and tiers, with its own entities and a default visualisation.
A cross-cutting pass writes the whole-subject summary and infers the relationships between aspects.
Drill across ten views; touched entities accumulate into a cluster you can save as a Snap and export.
An aspect set is not tied to one picture. Every FrameGraph can be read through ten visualisations, and each aspect carries a sensible default. The same who, what, when, where, why and how can be a briefing, a timeline, a map, or a graph, whichever fits the question. These are illustrative; open the demo to use them.
The first FrameGraph does the expensive work: it builds the topic graph for the subject and pulls its entities and relationships into the knowledge cluster. Adding another framework, or a few extra aspects, after that is not a fresh start. The supplement draws on the same shared cluster, so anything already covered appears at once and only genuinely new angles are retrieved.
Because frameworks overlap, a second one inherits most of the first. A 5W1H reading and a PESTEL reading of the same subject share the bulk of their entities: the people, organisations, places and events that answer who and where are the same ones that populate the political, economic and social aspects. The supplement takes them from the cluster rather than fetching them again, and adds only the angles the cluster has not yet covered.
This is already how the engine behaves. Switch on a supplemental aspect and the ones the cluster covers render immediately, while the ones it does not stream in through the same retrieval the rest of the FrameGraph uses, ticking the cluster as they arrive. The left rail marks each aspect as available or needing retrieval, so it is always clear what comes free and what is being fetched. The first framing is the investment; every framing after it is incremental, and the cluster grows richer with each pass rather than being rebuilt.
A FrameGraph is not a prettier answer. It changes what you can do with an answer.
Every angle on the subject stays separate and drillable, so a rich topic is never flattened into one paragraph you cannot take apart.
The same subject can be read through 5W1H, PESTEL, SWOT, the ecosystem, jobs to be done, or a decision frame, and reframed in a click.
Each aspect and entity carries provenance and a trust tier, so an answer built on it can show its premises and how firmly each rests.
Save a Snap, export a report, hand off a view. A FrameGraph is a working artifact, not a one-off reply that disappears.
Drilling and retrieving accumulate a knowledge cluster that you can extend across sessions and reuse as the seed of something larger.
Extract turns the prose into typed statements, facts, claims, beliefs, hypotheses and predictions, ready to govern as a Knowledge Substrate.
FrameGraph is not a separate tool bolted on the side. It is the engine behind the Canvas and the Browse (Topics) surfaces of the Corvair application, and it can be opened from anywhere a user already is, so understanding a subject in depth is one click from the answer that prompted the question. See how the Canvas works.
Ask a question and an Open in Canvas action turns the answer into a navigable FrameGraph. Run a Search and any result row can be framed the same way. Finish a Deep Research report and frame its subject to walk it from every angle. Once open, a FrameGraph behaves like the rest of the application: the same source modes and the same trust tiers as Ask and Search, the same drill and citation behaviour, and the same governance behind it. What you find can be saved to your Vault as a Snap, turned into a deliverable, watched through News, or raised as a Case when the view shows a Corvair knowledge base is thin on a subject. Suggest to Steward routes a statement or entity into the same curation queue the steward works from, and a built substrate seed flows into the same governed knowledge base every answer is drawn from.
Frame an answer to see the who, what, when, where, why and how behind it, each sourced and drillable.
Turn a result row into a structured view instead of opening and reconciling documents by hand.
Frame the subject of a report to navigate its actors, timeline and relationships interactively.
FrameGraph is the Topics surface itself: browse what a Corvair knowledge base holds and where it is thin.
Save a Snap and reopen it later; a FrameGraph is a working artifact, not a reply that disappears.
Suggest statements and entities for curation, so exploring a subject helps grow a Corvair knowledge base.
FrameGraph sits above any retrieval service and beneath the Corvair Knowledge Substrate. A list of links or a single generated answer cannot do what it does, because neither rests on a governed, trust-tiered topic graph with a warrant behind every claim.
A Corvair knowledge base is not assembled by hand. FrameGraph is the discovery and seeding front-end: from any topic, link, document, or prompt it produces a navigable view, and the extract converts that view into typed statements with provenance and a trust tier. A steward reviews the seed, promotes what belongs, and a Corvair knowledge base keeps it current under its recipe. The same governance that protects every Knowledge Substrate answer, trust tiers and a validity warrant, applies to knowledge that began as a FrameGraph. Start a new knowledge base from nothing, or extend an existing one, in the same motion.
One Corvair knowledge base can be read through 5W1H, PESTEL, SWOT, the ecosystem and more, which a fixed schema cannot offer.
It shows not only what a Corvair knowledge base knows but where it is thin, turning gaps into Cases and curation rather than silent blind spots.
Understanding a subject and seeding a Corvair knowledge base are one motion, so every framed view can become governed, warranted knowledge.
| FrameGraphing | The method: applying a chosen framework over a topic graph to produce a navigable, sourced view. |
| FrameGraph | The artifact the method produces, the navigable view itself. |
| Topic graph | The layer of entities and relationships beneath the subject: people, organisations, places, events, concepts. |
| Framework | A selectable set of aspects to read the subject through, such as 5W1H, PESTEL, SWOT, ecosystem, jobs to be done, or a decision frame. |
| Aspect | One angle within a framework, for example who or why, each generated, sourced, and given a default visualisation. |
| Knowledge cluster | The working set of entities and relationships that accumulates as you drill and retrieve. |
| Snap | A saved FrameGraph view, the subject, framework, aspects, drill state, and visualisation, that can be reopened later. |
| Reframegraph | Re-reading the same subject through a different framework. |
| Extract | Lifting typed statements, facts, claims, beliefs, hypotheses and predictions, from the view to seed a substrate. |
