Fix auto-restart watchdog (strptime parsing log filenames with new _<sessioncode> suffix), SyntaxWarning (\s in JS-inside-Python string), and pin python_version to 3.11 (avoids Python 3.14 asyncio cleanup noise with Gradio 6.x).

Fix auto-restart watchdog strptime: log filenames now include _<sessioncode> suffix; parse only the YYYYMMDD_HHMMSS prefix so idle check succeeds.

revise-gate modal: reword body to "Please revise at least 10% ... before submitting your final version."

revise-gate modal: single-line body, no title, button -> Go back to argument

revise-gate modal: quantified 90%/10% wording

switch hf_dataset to AE-Talk-bot/log-C-Jun-18-2026 (split per-condition to relieve HF commit rate limit)

recovery: rename localStorage keys to current condition names + JS namespace __SOC_RECOVERY -> __RECOVERY (clean up no_ai / socratic_frame legacy)

key-pool rotation: per-session OpenAI key assignment from OPENAI_KEY_01..10 (core/__init__.py)

key-pool rotation: per-session OpenAI key assignment from OPENAI_KEY_01..10 (core/logger.py)

key-pool rotation: per-session OpenAI key assignment from OPENAI_KEY_01..10 (core/perfect_answer.py)

key-pool rotation: per-session OpenAI key assignment from OPENAI_KEY_01..10 (core/config_loader.py)

key-pool rotation: per-session OpenAI key assignment from OPENAI_KEY_01..10 (app.py)

switch to log-Jun-17-2026 + C-non

Update README.md

Initial deploy of C (non-ca): no post-essay chat. Stripped chat additions from answer-ca base.

Rename condition + mode_prefix to *-ca to match local folder rename. hf_dataset unchanged (legacy data stays in same dataset under new subfolder).

Auto-grow chat input: lines=1, max_lines=6; CSS splits chat-input from buttons (textarea grows 42-160px; buttons stay 42px); chat-row aligns flex-end so Send stays at the bottom of the input.

general_chat_prompt: strip to vanilla — just case + initial_argument + perfect_answer, no behavior steering

general_chat_prompt: drop defensive "do not reveal discussion plan / probe / framework" rule since none are provided to the bot

B condition: add post-essay general chat. After perfect_answer is shown, msg_input + send_btn become interactive; on_send builds messages from case + initial_argument + perfect_answer (NOT discussion_plan) and calls LLM. Logs each chat turn via SessionLogger.add_turn(manager_verdict=None).

B condition: add post-essay general chat. After perfect_answer is shown, msg_input + send_btn become interactive; on_send builds messages from case + initial_argument + perfect_answer (NOT discussion_plan) and calls LLM. Logs each chat turn via SessionLogger.add_turn(manager_verdict=None).

B condition: add post-essay general chat. After perfect_answer is shown, msg_input + send_btn become interactive; on_send builds messages from case + initial_argument + perfect_answer (NOT discussion_plan) and calls LLM. Logs each chat turn via SessionLogger.add_turn(manager_verdict=None).

INTRO_MESSAGE: B condition reframed for AI-assistant essay + optional chat

case + placeholder: B condition reframed as 'AI assistant generates + optional chat'

case + placeholder: B condition reframed as 'AI assistant generates + optional chat'

STEP 1 framework detection: full reasoning-mode diagnosis with refined mixed-drafts rule. STEP 1 replaces keyword indicator lists with (a) identify load-bearing reasoning move, (b) 3 diagnostic questions in order (predicted outcomes / principles regardless of outcomes / character of an actor), (c) topic-vs-load decoupling table, (d) 'remove the lens' verification test. Mixed drafts rule: NOT a hard consequence default. Instead, pick the participant's CORE argument by explicit emphasis markers ('the most important reason' / 'perhaps for me' / 'the thing that is most ethically questionable is X') (highest priority), argumentative weight, then 'remove the lens' decisive test. Consequence is last-resort fallback only. Tested against pilot Kantian argument 5/5 correctly → duty_rights. Other downstream sections (essay, plan, hooks) unchanged.

STEP 1 framework detection rewritten from keyword-matching to reasoning-mode diagnosis. Previous prompt used indicator keyword lists ('right to X', 'consent', 'privacy' → duty_rights, etc.), which would misclassify drafts that mention a topic but argue through a different framework. New STEP 1 asks the LLM to (a) identify the load-bearing reasoning move, (b) ask 3 diagnostic questions in order (predicted outcomes / principles regardless of outcomes / character of an actor), (c) apply a topic-vs-load decoupling table, (d) verify via 'remove the lens' test. Mixed drafts hard-default to consequence. Regression tested in socratic_frame (8/8 cases including 1 synthetic privacy+consequence, 4 pilot consequence+privacy, 3 pilot duty_rights+privacy). Other downstream sections (essay generation, discussion plan, hooks) unchanged.

Session-recovery propagated from socratic: hidden gr.Textbox (#recovery-payload) + Gradio .then(js=...) chains write/clear localStorage; on_submit_initial / on_end_session yield JSON events on their CONFIRM yields. RECOVERY_HEAD adds modal styling + window.__SOC_RECOVERY helpers; storage key no_ai_session_v1; touch-observer adapted (no chat input — watches chatbot clicks + arg_input edits). on_recover_session rebuilds chatbot as [case_display, user_arg, INTRO_MESSAGE, perfect_answer.text] (mirrors on_submit_initial yield 2) and re-activates the 60s snapshot_timer. SessionLogger.from_dict restores dominant_framework (no_ai's logger lacks argument_submitted_at, so that field is skipped).

UI parity with socratic/socratic-frame: (1) hide Gradio Chatbot top chrome via show_label=False + CSS for share/clear/delete/trash; (2) strip the '# Case: ...' H1 from the initial chatbot message via display-only CASE_TEXT_DISPLAY; (3) hide the left yellow feedback-hint column (guide-col visible=False) so chatbot expands left, and drop the bottom-row mirror column to keep action buttons aligned under right-col.

UI parity with socratic/socratic-frame: (1) hide Gradio Chatbot top chrome via show_label=False + CSS for share/clear/delete/trash; (2) strip the '# Case: ...' H1 from the initial chatbot message via display-only CASE_TEXT_DISPLAY; (3) hide the left yellow feedback-hint column (guide-col visible=False) so chatbot expands left, and drop the bottom-row mirror column to keep action buttons aligned under right-col.

UI parity with socratic/socratic-frame: (1) hide Gradio Chatbot top chrome via show_label=False + CSS for share/clear/delete/trash; (2) strip the '# Case: ...' H1 from the initial chatbot message via display-only CASE_TEXT_DISPLAY; (3) hide the left yellow feedback-hint column (guide-col visible=False) so chatbot expands left, and drop the bottom-row mirror column to keep action buttons aligned under right-col.

UI parity with socratic/socratic-frame: (1) hide Gradio Chatbot top chrome via show_label=False + CSS for share/clear/delete/trash; (2) strip the '# Case: ...' H1 from the initial chatbot message via display-only CASE_TEXT_DISPLAY; (3) hide the left yellow feedback-hint column (guide-col visible=False) so chatbot expands left, and drop the bottom-row mirror column to keep action buttons aligned under right-col.

UI parity with socratic/socratic-frame: (1) hide Gradio Chatbot top chrome via show_label=False + CSS for share/clear/delete/trash; (2) strip the '# Case: ...' H1 from the initial chatbot message via display-only CASE_TEXT_DISPLAY; (3) hide the left yellow feedback-hint column (guide-col visible=False) so chatbot expands left, and drop the bottom-row mirror column to keep action buttons aligned under right-col.

UI parity with socratic/socratic-frame: (1) hide Gradio Chatbot top chrome via show_label=False + CSS for share/clear/delete/trash; (2) strip the '# Case: ...' H1 from the initial chatbot message via display-only CASE_TEXT_DISPLAY; (3) hide the left yellow feedback-hint column (guide-col visible=False) so chatbot expands left, and drop the bottom-row mirror column to keep action buttons aligned under right-col.

Record walkthrough_started_at client-side, pass to on_submit_initial via js= wrapper at submit

Move walkthrough_started_at + start_session to walkthrough END Start click (with JS retry)

Format walkthrough_started_at + argument_submitted_at as ISO 8601 in payload

Add walkthrough_started_at field to logger (top of payload, preserved across start_session)

Record walkthrough_started_at on Start walkthrough click (first click only)

Bump chatbot to 65vh and argument textbox to 55vh

Add welcome overlay before walkthrough; make image walkthrough full-screen; rename Done -> Start

Tutorial image 04.webp

Tutorial image 03.webp

Tutorial image 02.webp

Tutorial image 01.webp

Remove old PNG tutorial samples

Remove old PNG tutorial samples

Remove old PNG tutorial samples

Add tutorial image 03.png

Add tutorial image 02.png