Skip to content

feat(tui): modernize execution and focused workflows - #11

Merged
drewstone merged 3 commits into
mainfrom
feat/current-ux-audit
Aug 11, 2026
Merged

feat(tui): modernize execution and focused workflows#11
drewstone merged 3 commits into
mainfrom
feat/current-ux-audit

Conversation

@drewstone

Copy link
Copy Markdown
Contributor

Outcome

Braid now presents one stable execution identity above the transcript: AgentProfile, runner, model, connection, reasoning level, workspace, conversation, and branch.

Focused workflows use one divided full-screen surface.
/ask, comparisons, forks, approvals, and questions keep their consequence, response controls, and Escape or left-arrow path visible down to 40 by 12 cells.

Usage remains honest and separate for direct turns, analysis calls, and workers.
Estimated cost uses ~, observed floors use , and missing measurements remain unknown.

The former 477-line terminal coordinator is split into bounded identity, usage, focused-surface, and composition modules.

Real Pi product proof

This GIF is a clean install of the exact packed Braid candidate.
It configures a Product engineer AgentProfile, routes a coding task through Local CLI Bridge to Pi and GLM 5.2, verifies the changed workspace, then types /ask and displays the cited trace analysis.

Real Pi coding and analysis flow

Final analysis frame:

Real Pi analysis frame

  • Exact source commit: ab893af041e6d7d0efe670b9472af6a0d9bea651
  • Exact tarball SHA-256: 8dddc8d112dd142b8a11c5c08a4f7c59ed031fcae80a31081264de353988671a
  • Coding result: 8 passed, 0 failed; 12,804 input and 2,173 output tokens
  • /ask result: 3 cited findings; 7 of 7 calls visible across 2 pages
  • Analysis usage: 53,757 input and 8,153 output tokens; estimated cost $0.0501908
  • Packed visual coverage: 75 artifacts, 11 states, and 4 terminal sizes

Verification

pnpm check passed.

  • 615 runnable repository tests passed; 2 protected external checks skipped
  • 15 performance tests passed
  • 18 live-demo safety tests passed
  • CLI Bridge adversarial matrix passed
  • 457 modules and 2,336 edges have zero cycles
  • 68 production packages have no known vulnerability

External limits

Braid exposes two current runtime limits instead of simulating support:

Protected Tangle inference and sandbox checks still require their service credentials.

@drewstone
drewstone merged commit dba897b into main Aug 11, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant