You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
feat: enhance scientific writing tests with page budget management
- Added tests to validate the use of target_main_pages and minimum_main_pages in the page budget manager.
- Updated existing tests to reflect changes in the scientific writing policy.
feat: improve state graph runtime tests for usage tracking and budget management
- Introduced tests to accumulate successful node usage and persist it.
- Added tests for handling failed nodes and budget constraints during execution.
- Enhanced existing tests to ensure proper rollback behavior and budget guard functionality.
feat: extend terminal app plan execution tests for feedback handling
- Added tests to verify the correct handling of feedback and critique during plan execution.
feat: enhance write paper PDF build tests for page budget validation
- Updated tests to include validation for minimum and target main pages in compiled page validation.
- Added new tests to ensure proper handling of PDF compilation exceeding target page budgets.
feat: implement model pricing logic for LLM clients
- Introduced model pricing logic to compute usage costs based on token consumption.
- Added tests to validate the pricing propagation for OpenAI and Codex models.
feat: create event logging and replay functionality
- Implemented a persisted event stream for logging run events and replaying them.
test: add comprehensive tests for model pricing and usage computation
- Created tests to ensure correct resolution of model billing and computation of usage costs.
When `## Manuscript Format` is used, `main_body_pages` is treated as the nominal main-body page target, not a hard upper cap. In config, prefer `paper_profile.target_main_pages` and `paper_profile.minimum_main_pages`; the legacy `paper_profile.main_page_limit` field remains as a compatibility alias during migration.
Copy file name to clipboardExpand all lines: docs/architecture.md
+6Lines changed: 6 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -70,6 +70,12 @@ Top-level progression to paper-writing behavior should preserve the distinction
70
70
71
71
A paper-scale outcome requires evidence beyond successful orchestration, including baseline/comparator presence, real experiment execution, quantitative comparison, and claim-to-evidence linkage.
72
72
73
+
Page-budget semantics should also remain explicit:
Copy file name to clipboardExpand all lines: docs/paper-quality-bar.md
+7Lines changed: 7 additions & 0 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -54,6 +54,13 @@ After drafting, `write_paper` emits a post-draft critique that can:
54
54
55
55
In that case the correct action is upstream repair/backtrack, not spending drafting or PDF-compilation effort on a manuscript that should still be blocked.
56
56
57
+
### Page-budget semantics
58
+
`write_paper` should treat page budgets as explicit targets/floors, not as an implicit upper cap:
59
+
60
+
-`paper_profile.target_main_pages` is the nominal main-body target used for writing budgets
61
+
-`paper_profile.minimum_main_pages` is the compiled-PDF floor checked after LaTeX build
62
+
- legacy `paper_profile.main_page_limit` remains a compatibility alias and should not be interpreted as a maximum-page constraint
63
+
57
64
## 4) Venue-style targeting
58
65
Users can select a target venue style via `paper_profile.target_venue_style` in config.
0 commit comments