3 Commits

Author	SHA1	Message	Date
drew	ffac6be326	fix(tui): count conversation separators for pruning	2026-06-17 00:57:52 -04:00
HAL9000	90d65d8a3e	fix(tui): fix prune_excess minimum and robot test iteration count ... ConversationSettings clamped prune_excess to minimum=100, so the helper's prune_excess=50 was silently raised to 100, making trigger_line_count=200 exactly equal to the 200 lines added in the robot test. The > check never fired, pruning never ran, and both prune-trigger and prune-note-inserted failed. Fixes: - Lower prune_excess minimum from 100 to 10 in _normalise_positive so values like 50 are accepted without clamping. - Change range(20) to range(21) in cmd_prune_trigger and cmd_prune_note_inserted: 21 blocks x 10 lines = 210 lines triggers pruning twice (at block 15 and block 20), leaving total_lines=91 after the final prune event which satisfies the <= 100 assertion. ISSUES CLOSED: #6350	2026-06-17 00:57:51 -04:00
HAL9000	c5df00c6d3	feat(tui): fix lint, add robot/benchmark/changelog for pruning ... - Fix ruff format violations in app.py, conversation.py, and tui_conversation_pruning_steps.py (3 files reformatted) - Fix ruff E501 line-too-long in helper_tui_conversation_pruning.py - Expand ConversationSettings docstring to explain hysteresis design - Add _note_active invariant comment in ConversationStream.__init__ - Add full docstring to load_conversation_settings (config path, keys, fallback behaviour) - Narrow bare except Exception to (OSError, json.JSONDecodeError) and (ValueError, TypeError) in load_conversation_settings - Add Robot Framework integration tests (tui_conversation_pruning.robot + helper) covering prune-trigger, note-inserted, protected-blocks, clear-resets-state, settings-defaults - Add ASV benchmarks (conversation_stream_bench.py) covering add_block baseline, heavy-load pruning, render after prune, clear+repopulate - Add CHANGELOG.md entry for feat(tui) conversation content pruning ISSUES CLOSED: #6350	2026-06-17 00:57:51 -04:00