Commit Graph

110 Commits

Author SHA1 Message Date
lilleman bfde3ffecb Answer prose round 1: 0 🔴, 6 🟡 (0 misread); the comments, the presentation and the routing decision state the new steps
CI / tests (pull_request) Failing after 3m51s
2026-11-17 08:41:04 +01:00
lilleman 64ab847b0a Bind a project's rung to its ladder in ProjectRung, and step an L 3 worker one rung up and an L 4 worker two up
CI / tests (pull_request) Failing after 3m48s
2026-11-17 08:12:16 +01:00
lilleman 2600fbb254 Answer stability round 1: 1 🔴, 5 🟡 (0 misread); imports sorted, the replay cuts the task and the rules from its grounds and fails on a missing heading, outcome parameters keyword-only, the mtime order dropped 2026-11-17 01:13:39 +01:00
lilleman e1d2eae7f4 Archive each judge run's prompt and json reply beside the verdict it gave, and replay from the archived prompt
CI / tests (pull_request) Failing after 2m45s
2026-11-16 23:42:36 +01:00
lilleman f58250d0ce Answer architecture round 1: 4 🔴, 5 🟡 (0 misread); one-way plan packages with a plan core and an Atlassian client, one parse and print of an item's cells, a pure chunk-start decision, a settings table, one plan read and one standing per pick, one config load in chunk-pr, notifier backends in notifier, and the bin regroup filed as item 392 2026-11-16 14:47:46 +01:00
lilleman eecb0387ca Run each chunk in its own tmux window, up to parallel-chunks at once, and mark a window that waits
CI / tests (pull_request) Failing after 2m56s
2026-11-16 11:59:45 +01:00
lilleman bedd304479 Merge remote-tracking branch 'origin/main' into patch-0.10.1
# Conflicts:
#	CHANGELOG.md
#	README.md
#	bin/loop_claude/loop/chunk_start.py
#	bin/loop_claude/loop/close.py
#	bin/loop_claude/loop/main.py
#	bin/loop_claude/routing.py
#	bin/loop_claude/test_loop_claude.py
#	bin/project_config/main.py
#	bin/project_config/test_project_config.py
#	docs/decisions.md
2026-11-15 19:54:42 +01:00
lilleman 0a20b6d4f6 Read the routing ladder from routing-ladder.json, which a project replaces with a file of its own in its config's folder
CI / tests (pull_request) Successful in 2m37s
2026-11-15 18:51:59 +01:00
lilleman 0ac8cad97f Route every read and write of the plan through plan_store and todo-item, which gains list, edit and move
CI / tests (pull_request) Failing after 2m13s
2026-11-14 15:15:25 +01:00
lilleman c1924e51ef Remove loop-inbox, so an interrupt is answered at loop-claude's terminal only
CI / tests (pull_request) Successful in 2m18s
2026-11-13 23:22:49 +01:00
lilleman eb61043c12 Read a project's goals and audience from the file its config names, in place of README.md
CI / tests (pull_request) Successful in 2m17s
2026-11-13 00:06:11 +01:00
lilleman 5c37027046 Answer architecture round 1: 1 🔴, 8 🟡 (0 misread); one reader per message, prices and Spent block moved, one table of larv-review commands 2026-11-12 20:05:52 +01:00
lilleman fa54ed46b1 Answer prose round 1: 0 🔴, 7 🟡 (0 misread); rename Record to Annotation, and reword the decision, the help and the changelog
CI / tests (pull_request) Successful in 2m37s
2026-11-12 10:25:30 +01:00
lilleman 3c4cedd389 Record with each archived verdict the setup that ran it, what each model spent by kind, and the human's reversals
CI / tests (pull_request) Failing after 2m12s
2026-11-12 09:57:56 +01:00
lilleman c61997f7b7 Run the judge of an Opus 5.5 project on Opus 5.5 at xhigh in every chunk
CI / tests (pull_request) Successful in 2m35s
2026-11-12 02:54:03 +01:00
lilleman e7128656b2 Answer stability round 1: 0 🔴, 4 🟡 (0 misread); ask for the Chunk items: line on resume too, split it on "and", and say on stderr why todo.md scores no item 2026-11-11 18:16:14 +01:00
lilleman 7507b3b670 Say in rule's docstring that code stamps the verdict with the chunk's items 2026-11-11 17:44:42 +01:00
lilleman daa3a54b90 Have loop-claude read the items a chunk takes from the worker's Chunk items: line, and keep the judge's escalation and release-ready fields, which code misread in replays
CI / tests (pull_request) Successful in 2m7s
2026-11-11 17:43:40 +01:00
lilleman 12d70586ed Answer principles round 1: 0 🔴, 1 🟡 (0 misread); test the trials' imports from bin, pass one loop value through loop/, type the Exempt reasons
CI / tests (pull_request) Successful in 2m20s
2026-11-10 22:20:48 +01:00
lilleman 70a1b95086 Answer architecture round 1: 0 🔴, 4 🟡 (0 misread); move the shared loop state out of loop_claude, split ranking and routing, give the reviewer rung one rule
CI / tests (pull_request) Successful in 2m14s
2026-11-10 22:03:16 +01:00
lilleman 268200de23 Pass the judge a Brief, decide read_seats from its inputs and return a failed bench run as a line 2026-11-08 22:51:27 +01:00
lilleman 9fea1d8286 Answer prose round 1: 0 🔴, 9 🟡 (0 misread); split the interrupt entry, unstack the decisions' lists, say why Cycle's fields and the package's modules stay at one level, name the import cycle, and say how the judge's quotes are checked 2026-11-04 23:35:35 +01:00
lilleman 953d0c4176 Answer principles round 1: 1 🔴, 5 🟡 (0 misread); archive a ruling as the judge rules so the states carry only whether the worker stepped, hold every judge quote in one table, give each launch sentence its own function, pass the questions folder in place of rebinding paths, open routing calls in the development clone, return Failure from the file readers, read unreadable tags as Unknown, fold judge costs into costed, and tidy the 🟢s 2026-11-04 23:23:41 +01:00
lilleman 452a0c4271 Answer architecture round 1: 1 🔴, 12 🟡 (0 misread); route every verdict through next_route with one step-up writer, give a release one home in release.py and rename chunk_pick to release_standing, split loop/main.py, delete the unreachable lowered-pass code, render agent-followed Markdown from one list, refuse a round title naming no pass, and rename the clashing names 2026-11-04 22:57:30 +01:00
lilleman 2f6ff46f99 Count the judge's cost in the routing figures 2026-11-04 20:32:23 +01:00
lilleman ac9d107066 Answer prose round 1: 0 🔴, 8 🟡 (0 misread); reword the running loops' docstrings, help and decision, widen the decision's title, keep filed items out of items_taken_in 2026-10-29 20:42:37 +01:00
lilleman dcdb8b5868 Answer architecture round 1: 0 🔴, 7 🟡 (0 misread); hold one lock per loop for its whole run, group the judge's readings of the items taken in, list them as given 2026-10-29 20:23:07 +01:00
lilleman 9bcfaebb83 List every running loop in loop-inbox, with the ID and tagline of each item its chunk takes 2026-10-29 19:57:11 +01:00
lilleman 9ebd06e748 Restructure the interrupt so one function owns the ask and loop-inbox imports nothing from cycle/ 2026-10-29 13:11:04 +01:00
lilleman 3409ab3d07 Answer architecture round 1: 1 🔴, 6 🟡 (0 misread); open claude through one hand_over, route loop-inbox's /talk through the Talking state, give Ask one state 2026-10-28 23:57:33 +01:00
lilleman a7fe878631 Show every waiting loop-claude in loop-inbox and take each one's answer there 2026-10-28 23:24:29 +01:00
lilleman c583900199 Answer product-owner round 1: 0 🔴, 1 🟡 (0 misread); stamp the transcript and trust lines through say 2026-10-28 18:47:39 +01:00
lilleman 6e171fe2c0 Answer principles round 1: 0 🔴, 2 🟡 (0 misread); move say beside Effects and drop a clock comparison from the black-box test 2026-10-28 18:40:24 +01:00
lilleman 43e7e96b41 Open every line loop-claude prints with the machine's local time 2026-10-28 18:30:24 +01:00
lilleman f41b549ff7 Merge origin/main into item-59 2026-10-28 02:24:52 +01:00
lilleman 14736e8e6e Answer principles round 1: 0 🔴, 9 🟡 (0 misread); name the 200k window's compaction point, read a compaction summary as Claude Code's, remove item 231, refresh items 234, 235, 52 and 102, file item 237 2026-10-27 00:32:10 +01:00
lilleman f2e239d111 Compact the worker's context above 200k tokens, from item 231's evaluation of context cost, and file items 235 and 236 2026-10-27 00:13:08 +01:00
lilleman 2135afe8d7 Answer architecture round 1: 1 🔴, 4 🟡 (0 misread); share the stop archiving, fix the limit at 6, reword the exit-2 contract, close the cycle in one step 2026-10-23 15:09:46 +02:00
lilleman 387a5ce506 Answer architecture round 1: 0 🔴, 2 🟡; finish "record", record the three meanings as a decision, name rule-hook's check once 2026-10-23 10:54:24 +02:00
lilleman 6f98a1855b Answer prose round 1: 0 🔴, 5 🟡; tighten the escalation mark's comments, item 221's detail, and fold three decision entries into item 190 2026-10-23 05:22:15 +02:00
lilleman f781ac611b Answer principles round 1: 0 🔴, 1 🟡; rename the module escalation_mark.py so it no longer shares the escalation field's name 2026-10-23 05:07:30 +02:00
lilleman bf8b642ead Answer architecture round 1: 0 🔴, 2 🟡; rename Step to StepUp and let escalation.py own the judge's step-line sentence 2026-10-23 04:54:45 +02:00
lilleman f0e14f3fdf Hold the escalation mark in one module, and carry a ruling's step through the cycle's states 2026-10-23 04:45:03 +02:00
lilleman 3b69ced357 State each routing rule once in the docstring of the code that enforces it, cut loop-claude --help's routing paragraph, and name the project's rung project in bin/loop_claude 2026-10-22 22:42:01 +02:00
lilleman cb48719493 Answer architecture round 1: 0 🔴, 4 🟡; move verdicts out of cycle, derive RECORDED, type the archived worker as Record, drop stale help names 2026-10-22 22:00:09 +02:00
lilleman 851d7d7c79 Archive the route as a JSON record, hand routing the verdict, and name after_unfinished and the judge's highest levels taken in by what they do 2026-10-22 21:46:36 +02:00
lilleman 2bba66ac0a Answer architecture round 1: 0 🔴, 6 🟡; show the judge the mark, carry the entries read on the verdict, name the checked texts 2026-10-18 13:50:06 +02:00
lilleman 3a8e4e6f75 Accept an escalation quote only from transcript entries after the worker's last step-up 2026-10-18 13:24:05 +02:00
lilleman 4b2d4a1734 Answer principles round 1: 0 🔴, 6 🟡; correct the interrupt's definition, show the judge the question beside the typed answer, ask in one call returning Talk, add a Talking state, give the worker its own question past the cap, assert the prompts' facts, and keep stepped_from as the step's one owner 2026-10-17 23:32:21 +02:00
lilleman 96834412c7 Answer architecture round 1: 0 🔴, 2 🟡; keep stepped_from through the typed answer, name the events HumanTyped and HumanTalked, parse the answer from the prompt's start, and say why the judge decides release_ready 2026-10-17 23:16:34 +02:00