Synthesizing your architecture

Synthesizing lives on the Architecture tab of Discovery. It combines your mapped repositories with any confirmed infrastructure resources into two things: a narrative and a relationship graph.

The tab explains: "How the system connects — synthesized from the codebase map and the confirmed infrastructure. The narrative is also written to .config/project/architecture.md and auto-injected into agents." Click "Synthesize" (or "Re-synthesize" if a narrative already exists) to run it, and it typically takes under a minute. Until you run it at least once, the tab shows: "No architecture model yet — run 'Synthesize' once the inventory has confirmed resources."

You do not strictly need any cloud infrastructure to get value from this; running it right after mapping code still produces a narrative from your codebase alone. It is simply richer once you have also discovered and confirmed infrastructure.

What do I get?

• A narrative: a written description of how the system fits together, rendered as formatted text, with a synthesized {time ago} timestamp underneath.
• A relationships table, one row per inferred edge: From (source, with its kind), Relationship (a badge, one of deploys to, uses storage, uses db, calls, consumes api, or relates to), To (destination, with its kind), Confidence (high/medium/low), and Why (a one-line rationale).

The same relationships also surface in a "Relationships" panel on the Infrastructure tab, and any new edges an Investigate run finds get merged in here too.

Who can do this?

Requires discovery:manage (and a mutable project) to run Synthesize; discovery:read is enough to read whatever narrative and relationships already exist.