Repository navigation
T27 Model Control: MODEL panel (epic #13172) - #13173
Merged
Merged
Conversation
Adds the MODEL panel to the T27 OS right-hand strip beside KEYS, DOCTOR and PROFILE, answering the four questions: what is running, why this model, whose credentials, how much spent (known/estimated/reserved/ unknown). Unknown cost is never shown as zero. Specs: - model_trace.t27: per-attempt trace record, classification, redaction, versioned price table, golden incident line (11 tests) - model_budget.t27: gate/reservation/retry/fallback/period logic (10 tests) - model_control_panel.t27: panel rows, totals, four answers, geometry (6 tests) - model_log.t27: parses deployed ladder.log into trace rows, unknown fields marked unknown not guessed (7 tests) - model_trace_main.t27: recorder that writes ~/.t27os/model-trace.tsv from the ladder log tail (1 test) UI wiring: - statusbar.t27: MODEL link between KEYS and DOCTOR at the right end - glass_window.t27: mp flag, draw_model_panel, click handling, panel exclusion, all 101 glass_window tests pass Sealed in .trinity/seals/ (owner rule DSH_MCP_VISIBLE_EXECUTION separately in gHashTag/t27). Closes #13172
Contributor
This was referenced Nov 4, 2026
Merged
Merged
This was referenced Nov 9, 2026
Co-authored-by: Dmitrii Vasilev <gHashTag@users.noreply.github.com>
This was referenced Nov 9, 2026
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using high effort and found 5 potential issues.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 9e747f2. Configure here.
| w = put_cells(rows[at * TR_CELLS .. at * TR_CELLS + TR_CELLS], out[w..]); | ||
| w = put_bytes(out, w, "\n"); | ||
| i = i + 1; | ||
| } |
Contributor
There was a problem hiding this comment.
Trace file rows clobber header
High Severity
put_cells writes from the start of the slice it is given and returns that relative length. Assigning that return value back to w then passing it to put_bytes on the full out buffer overwrites the header and earlier rows. The MODEL panel's model-trace.tsv comes out truncated and corrupt, so the panel cannot show the recorded attempts.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 9e747f2. Configure here.
| } | ||
|
|
||
| pub fn main() -> void { | ||
| _ = sys_write(2, "A\n", 2); |
Contributor
There was a problem hiding this comment.
Debug breadcrumbs left in recorder
Medium Severity
main writes A through E to stderr around each stage of the recorder. The file header says a successful run prints one summary line, so these leftover breadcrumbs leak into every normal invocation.
Additional Locations (2)
Reviewed by Cursor Bugbot for commit 9e747f2. Configure here.
| _ = sys_write(2, "C\n", 2); | ||
| var rows : [CAP * TR_CELLS]u64 = undefined; | ||
| var t : [ML_T_CELLS]u64 = undefined; | ||
| scan_log(ML_BUF[0..BUF_BYTES], n, rows[0..CAP * TR_CELLS], CAP, t[0..ML_T_CELLS]); |
Contributor
There was a problem hiding this comment.
Recorder totals cover only tail
High Severity
BUF_BYTES is 64 KiB and main seeks to the last 64 KiB of ladder.log before a single scan_log. Header comments promise a 16 MB tail, and scan_log/acc_init say totals cover the whole log via 64 KiB chunks. On the documented 13 MB incident log the panel's total/unfunded/wasted_ms only reflect the newest slice, which understates the damage the epic exists to show.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 9e747f2. Configure here.
| qn = put_bytes(line[0..128], qn, " -> "); | ||
| qn = put_bytes(line[0..128], qn, tp[0..tn]); | ||
| qn = put_bytes(line[0..128], qn, "\n"); | ||
| _ = sys_write(1, line[0..128], @as(u64, @intCast(qn))); |
Contributor
There was a problem hiding this comment.
Success line overflows path buffer
Medium Severity
The success line is assembled in a 128-byte line buffer, then put_bytes appends tp[0..tn]. T27_MODEL_TRACE and the default HOME path are allowed up to a few hundred bytes, and put_bytes does not clip. A long path traps or corrupts after the trace file has already been written.
Reviewed by Cursor Bugbot for commit 9e747f2. Configure here.
| } | ||
| pub fn on_model(x : f64, y : f64, sw : f64, sh : f64) -> bool { | ||
| return in_bar(y, sh) and x >= model_left(sw) and x < model_left(sw) + SB_MODEL_W; | ||
| } |
Contributor
There was a problem hiding this comment.
MODEL slot shifts KEYS with no draw
Medium Severity
right_w and keys_left now reserve a 48 px MODEL slot between KEYS and DOCTOR, and IT_COUNT is 12. glass_window still draws only KEYS and DOCTOR, still sizes footer slots/lens/widths for 11 items, and never calls on_model or PG_MODEL. KEYS and OPEN SOURCE shift left into a dead gap, and enabling IT_MODEL in present would index past those arrays.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 9e747f2. Configure here.
This was referenced Nov 9, 2026
Contributor
PR DashboardGenerated at: 2026-11-09 22:00:06 UTC
Summary
Seal Status
|
This was referenced Nov 9, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #13172
Note
Low Risk
Spec-only additions with embedded tests; no production ladder or window wiring in this diff beyond statusbar constants.
Overview
Adds T27 specs for epic #13172 so the OS can observe model routing, spend, and the 402→503 retry incident without reading secrets or faking zero cost.
model_trace.t27defines per-attempt records: separate upstream vs gateway status, body-vs-envelope ambiguity, credential naming (no key material), known/estimated/unknown cost, and rules that UNFUNDED must not retry on the same account.model_log.t27parses existing[ladder]log lines into trace cells (including 402 upstream + 503 gateway), scans logs for totals, and serializes rows for the panel.model_budget.t27specifies send gating (reservations, limits, attempt/time caps, paid fallback policy) and post-refusal retry behavior aligned with trace classes.model_control_panel.t27turns trace rows into the four owner questions (running, why, whose, spend), panel geometry, and row/total formatting.model_trace_main.t27is a small recorder: tail-readladder.log, keep the last 64 requests, write~/.t27os/model-trace.tsvfor the window.statusbar.t27adds a MODEL footer link (PG_MODEL) between KEYS and OPEN SOURCE T27, with layout/click tests updated.Reviewed by Cursor Bugbot for commit 9e747f2. Bugbot is set up for automated code reviews on this repo. Configure here.