Build log: M13 scale

July 20, 2026 · Milestone M13 scope

Phase C of the galaxy UX overhaul made Codemble usable on projects roughly three times larger, without changing one byte of parser output.

What changed

• Parsing moved off the request thread. POST /api/picker/select now answers 202 {"state": "parsing"} immediately and a worker thread parses. GET /api/picker/progress reports one of idle, parsing, ready, or error, with the current stage and a real file count.
• Five honest stages, in order: discovering, parsing, resolving, checks, layout. The file counter moves during parsing only, because that is the only stage measured per file. resolving — cross-file import and call resolution, the slowest stage — instead narrates its own real sub-steps (“Resolving imports”, “Resolving calls”, “Building the galaxy map”, “Composing your project”) so the screen keeps moving without inventing a denominator it does not have.
• Cancellation. Resetting the picker during a parse sets a flag that is checked between files, so the parse stops at the next file boundary and the picker re-arms. A crashed parse thread becomes an error state with the parser’s own message, surfaced back on the picker with a one-click retry — never a hung server.
• The scale cap moved from 300 to 1,000 supported source files. Above it, the picker offers the busiest scopes as buttons and accepts a typed path, both still jailed to your home directory. A piped, non-interactive CLI run now prints those same scopes instead of a bare refusal.
• Two measured cliffs removed. Check generation used to walk every graph edge up to four times per region; it now builds one index in a single pass (measured ~16x faster at 1,000 files, generated suites byte-identical). GET /api/graph and GET /api/map used to re-hydrate and re-sort the whole graph on every request; the serialized document is now cached and dropped only when a region lights up, Home changes, or the project is rebound (measured ~25-26x faster warm).
• A parser hotspot in the Python adapter’s own module-resolution lookup was also fixed — an O(definitions × modules) scan replaced with an O(depth) walk — for a further 1.56x on total parse wall-clock, again byte-identical.
• Clear this project’s progress is available from the star chart, behind a confirmation, scoped to the open project.

What did not change

Check suites and answers are byte-identical to before the index change; a committed golden fixture proves it for both the Python and the mixed JavaScript/TypeScript fixtures. Progress reporting cannot influence parser output: the same source produces the same graph JSON with or without a reporter attached, and a test pins that. The LanguageAdapter seam is unchanged.