apply_scene_close_summary fed detect_threads the chat-wide last-50
turns. When a chat has accumulated multiple scenes' worth of dialogue,
that bleeds prior-scene turns into the second close's classifier prompt
and risks mis-attributing threads (closing one that opened earlier,
re-opening one that already closed).
Add an optional ``since_event_id`` kwarg to ``_read_recent_dialogue``
that lower-bounds by event_log id, plus a ``_scene_opened_event_id``
helper that resolves the scene-open event for a given scene_id. Wire
both into the thread-detection call site so its scene_transcript
holds only the closing scene's turns. The per-POV summarizer keeps the
chat-wide approximation it had before — that's intentional.
Adds test_thread_detection_uses_scene_scoped_transcript.
Re-running apply_scene_close_summary on the same scene previously caused
recursive bloat: _build_key_quotes_suffix sourced quote text from
memories.pov_summary, which after the first close already carried a
"Key quotes:" suffix. The next close would then quote the quotes,
nesting deeper each time.
Strip any existing suffix from candidate text before truncating to
200 chars in the suffix builder, and from the fresh classifier output
before composing the new value in _summarize_and_apply_for_witness so
the rewrite replaces rather than stacks.
Adds test_scene_close_re_run_does_not_double_suffix.
Extend ParsedTurn with intent/landing_state_hint so the classifier can
flag skip-elision and skip-jump prose. The post_turn handler short-
circuits the regular narrative path when intent != "narrative":
elision runs through the shared controller in chat/web/skip.py;
jump returns 422 directing the user to the drawer's structured form
(simpler Phase 3 path — natural-language fiction-time delta parsing
is too fragile for v1 without a structured surface).
Extract the elision/jump logic that previously lived in drawer.py
into chat/web/skip.py so both the drawer T59 routes and the new
natural-language path share one canonical implementation. The drawer
routes become thin HTTP wrappers that translate ValueError to 400
and refresh the drawer partial; the existing drawer skip tests pass
unchanged.
The new natural-language elision derives ``new_time`` by bumping the
chat clock by 1 hour (Phase 3 stub) — the drawer's structured form
remains the path for picking a specific landing time.
Phase 2 T44 review noted that scene close still runs when a primary
turn is cancelled mid-stream and asked the implementer to review.
Review finding: the existing behavior is correct, not a bug. The
close-detection branch in post_turn consumes ONLY the user's prose
(fully appended to the event_log BEFORE streaming starts) and the
current container name. It does NOT consume the bot's output. A user
who types "we're done here, fade out" and then hits Stop mid-stream
still meant to close — the cancelled bot beat doesn't invalidate
that intent.
- Document the rationale with an inline comment near the
close-detection branch in chat/web/turns.py.
- Add regression test
test_cancelled_turn_still_closes_scene_when_user_prose_signals_close
that drives a stream raising CancelledError on first iteration and
asserts the scene_closed event still lands.
Phase 2 T44 deferred interjection regenerate — when the original turn
group included a follow-on interjection beat we left it untouched. Now
regenerate redoes BOTH halves:
- Detect a sibling interjection by looking up assistant_turn events
pinned to the same user_turn_id with `interjection_of` set.
- After streaming the new primary, run `detect_interjection` against
the new primary text.
- If True: stream a new interjection from the silent witness, append
with `interjection_of=<new primary speaker_id>`, supersede the
original interjection, and re-run memory + state-update for the new
beat.
- If False: supersede the original interjection without a replacement
(back-pointer goes to the new primary so the row stays consistently
hidden).
Also broadcast a `turn_html_replace` event for the new interjection so
the front-end can swap the prior interjection node in place (mirrors
T73.1's primary swap).
Tests:
- `test_regenerate_with_interjection_redoes_both_turns`: classifier
returns True; assert two new assistant_turns land for the same
user_turn, second carries `interjection_of`, originals superseded.
- `test_regenerate_drops_interjection_when_classifier_returns_false`:
classifier returns False; assert one new assistant_turn (primary
only) and the original interjection is superseded with no
replacement.
`interjection_of` carries the primary's *speaker_id* (matching the
existing convention in chat/web/turns.py) rather than the event_id.
T44's interjection branch wrote interjection memories via
record_turn_memory_for_present but never enqueued a SignificanceJob,
so the interjection beat could land in memory but never be scored —
which meant it could never auto-pin even when it carried a pivotal
moment.
- Capture the host-POV memory id from the interjection's memory write
result and enqueue a SignificanceJob mirroring the primary turn's
pattern. One enqueue per beat (host id; guest POV piggybacks on the
same score since the prose is identical for v2 — per-POV rewrite
happens at scene close in T45).
- New test test_interjection_enqueues_significance_job pins the
contract by intercepting worker.enqueue and asserting two distinct
jobs land per 3-entity turn that fires an interjection.
Replace the substring _detect_addressee_id helper with a classifier
call for the multi-entity case. The substring helper is kept as a
fast-path for the no-guest case (no LLM round-trip needed when only
one bot is present, preserves throughput).
- New service chat/services/addressee.py wrapping the existing
classifier wrapper. AddresseeDecision carries addressee_id +
confidence + reason; classifier failure falls back to the host with
reason="fallback" (graceful-degradation, matches the relationship_seed
/ interjection pattern).
- chat/web/turns.py post_turn now calls detect_addressee in the
multi-entity branch; 1:1 keeps the substring path.
- tests/test_addressee.py: 3 new tests (guest pick, host pick,
classifier-failure fallback).
- tests/test_turn_flow.py: existing multi-entity tests now feed a
canned addressee response in the queue. The addressee-routing test
is updated to assert classifier-driven routing rather than substring.
After the new assistant_turn lands, publish a `turn_html_replace` SSE
event carrying the rendered HTML, the new turn_id, and the original
assistant_turn id as `supersedes_id` so connected tabs can swap the
prior DOM node in-place. Phase 1 T29 deferred this — page had to refresh
to see the regenerated turn.
Uses a new event name (not the existing `turn_html`) because the HTMX
`sse-swap="turn_html"` consumer expects raw HTML and an *append*
semantic; regenerate is a *replace*. The new event ships as JSON
(supersedes_id forces sse.py's JSON branch) so the front-end JS can
read the swap target from the payload.
Test: `test_regenerate_broadcasts_turn_html_over_sse` patches the
`publish` reference inside the regenerate module and asserts the
event shape.
Memories grow per-flag witness checkboxes (you / host / guest) that
auto-submit on change via HTMX. The new POST route emits a manual_edit
event with target_kind=memory_witness and a {flag, value} payload;
prior_value mirrors the same shape so an inverse edit restores the
flag. The drawer's recent-memories query now selects the three
witness columns alongside the existing fields so the template can
render checkbox state without a second query per row.
When a host->candidate edge already exists from a prior chat, the
Add-guest form renders the prose textarea disabled with an "already
know each other" note. Submission without the explicit "re-seed
anyway" toggle skips seed_inter_bot_edges so existing edge content
(affinity, trust, knowledge, summaries) survives — guest_added and
group_node_initialized still fire. A small inline script enables /
disables the textarea per-option based on a pre-computed
existing_guest_edges dict surfaced by the GET handler.
Adds the four POST routes whose state-layer support was already
dispatched by the manual_edit projector (edge_trust, edge_summary,
memory_pov_summary) plus a new edge_knowledge_fact dispatch branch for
add/remove fact list manipulation. Drawer template gains editable
textareas, sliders, and add/remove fact controls. Remove semantics on
knowledge_fact match by string (not index) so concurrent edge_update
events appending facts between drawer renders don't desync the form.
T18 review (Phase 1) noted the NICE-tier trim drops previous-scene
FIRST while §6.3 spec lists previous-scene LAST in the NICE tier
group. Decision: keep the existing greedy order (previous-scene
first), and document why.
Rationale (now in code at the trim ladder):
1. Cheapest-impact-first — a per-POV previous-scene summary loses
less narrative continuity than the older dialogue turns or
memory hits it competes with.
2. Greedy lookahead is more expensive than the marginal narrative
loss. Dropping previous-scene typically clears the soft-budget
slack in one step.
Test added: test_nice_trim_order_documented pins the observed order
(previous-scene -> memories -> dialogue) so a future refactor can't
silently invert it. Sized so that all-NICE config overflows soft but
dropping just previous-scene fits — proves memories and older
dialogue turns survive while previous-scene is the FIRST drop.
Phase 2 T43 added a SECOND ACTIVITIES: block to render guest activity
separately from you+speaker. Two consecutive ACTIVITIES: headers can
read as a duplicate-section bug to the LLM and bloat the prompt.
Consolidate to a single ACTIVITIES: block whose body is composed from
up to three bullets (you, speaker, guest). The block itself is
MUST-tier (always renders); bullet-level trim drops bullets in the
order guest -> group node -> you -> other edges, with the speaker
bullet as the MUST-tier floor (the speaker's own current activity is
the load-bearing slice).
Implementation chose Option B from the polish plan: pre-truncate the
bullets list at trim time before _build_activity_block runs, rather
than introduce a granular tier mode in the trim machinery. Rationale
documented in code; the existing block-level trim ladder gains a
single new toggle (include_you_activity) and the SHOULD-tier
guest_activity_block is gone.
Tests:
- test_single_activities_block_with_three_bullets_when_3_entities:
exactly one ACTIVITIES: header with all three entity bullets.
- test_tight_budget_drops_guest_activity_bullet_first: speaker bullet
survives, guest bullet absent under tight budget.
- Existing test_assemble_with_tight_budget_drops_guest_activity_first
still passes (asserts on bullet absence, not block-header absence).
Phase 2 T46 pinned the witness mask contract on search_memories with a
witness_role parameter (host/guest/you). The prompt-assembly call site
in assemble_narrative_prompt was hardcoded to "host", which silently
returned the wrong rows when the speaker was the guest bot.
Derive the witness role from chat membership via a new private helper
_witness_role_for(speaker_bot_id, host_bot_id), and apply it at the
search_memories call. Behaviour is identical when the speaker is the
host (or when no guest is present); the fix is load-bearing only when
the guest bot is the speaker — exactly the scenario Phase 2 T43 added
support for.
Tests: pin both directions (host-as-speaker and guest-as-speaker) by
patching the imported search_memories reference and asserting the
witness_role argument the call site emits.
Joseph Doherty