[Atlas] Per-disease landing page synthesizes all artifacts (auto-updating) done

Effort: thorough

Goal

Today a researcher who lands on /disease/alzheimers is bounced through a
301 to a wiki page (see api.py:61436) that shows static prose. SciDEX
actually has, for each disease, hundreds of artifacts spread across every
layer — top hypotheses, debates, open questions, prediction markets,
notebooks, citations, challenges, funding flows. None of this is surfaced
in one place. Build a real per-disease landing page that pulls EVERY
artifact type linked to the disease and renders a synthesized
"state of the field" view, auto-refreshing as new artifacts arrive.

This is the canonical demonstration of the artifact-unification spine
(wave-1 Q-DSC/Q-OPENQ/Q-LIVE/Q-PROP/Q-PERC) paying off: cross-layer
synthesis no human or single agent can produce by hand.

Acceptance Criteria

/disease/{name} 301 redirect (api.py:61436) is updated so any
disease wiki page that has an entity_id pointing at a known
neurodegenerative disease ALSO gets a "View live landing page"
banner linking to /disease-landing/{slug}. register_dashboard (scidex/atlas/artifact_registry.py:2318)
with slug pattern disease-landing-<slug>. One artifact per
disease — registered eagerly for the top-20 ND diseases by a
one-shot script scripts/seed_disease_landing_dashboards.py,
lazily on first hit otherwise. against scidex/senate/dashboard_engine.py:_validate_query:
1. top_hypotheses — hypotheses joined to hypothesis_entities
(or entity_links) on disease entity_id, ordered by
composite_score DESC, top 10.
2. open_debates — debate_sessions joined to those hypothesis
ids, status not in (closed, archived), most recent 10.
3. unanswered_questions — Q-OPENQ artifacts (artifact_type =
open_question) tagged with the disease, ranked by score, top 10.
4. live_markets — markets joined on the hypothesis ids, status
= active, ordered by liquidity_usd DESC, top 10.
5. recent_notebooks — artifacts WHERE artifact_type IN
('notebook','analysis','figure','dataset') linked via
artifact_links, last 30 days.
6. key_citations — top 10 papers by citation_count whose
mesh_terms or wiki_entities link includes the disease.
7. funding_flows — challenges (Q-CW bounties) for the disease,
total dollars on the table.
8. recent_proposals — Q-PROP proposed_artifact rows from
eb6ff576-9d5 continuous-proposal pipeline tagged with
disease, last 14 days.
9. whats_changed — count of every artifact type from #1-8
created in the last 7 days vs. the prior 7 days, for a
"momentum" header strip. dashboard_engine._TEMPLATES. Layout:
- Hero strip: disease name, top mechanisms, momentum strip
(rising/falling arrows for each artifact-type count).
- Three-column "What we know / What we're debating / What we don't
know" grid mapping #1 / #2+#4 / #3+#5.
- Bottom row: funding flows, recent notebooks (with thumbnails),
key papers. /api/dashboard/{id}/refresh. 2 debates + 4 questions + 1 market + 2 notebooks + 1 paper +
1 challenge and asserts each panel renders the correct counts. /disease-landing/parkinsons-disease against production DB and
verify ≥1 row in every panel (or an explicit empty-state block
with a "no debates yet — be the first" CTA pointing at
/agora/new).

Approach

Read the disease wiki redirect at api.py:61436 to know the slug

convention (diseases-<name>).

Audit which entity-link table the codebase uses to tie artifacts to

diseases (likely entity_links, wiki_entities, or
canonical_entity_links — see
scidex/atlas/canonical_entity_links.py). One join per data_source
will cover almost all of #1-7.

Reuse the _pc() ILIKE-escape helper landed for

q-live-ad-therapeutic-landscape (work log notes the bug it fixed).

The momentum strip is a single CTE with WHERE created_at >

NOW() - INTERVAL '14 days' and a CASE WHEN created_at >
NOW() - INTERVAL '7 days' bucketing — keep all math in SQL so
render stays stateless.

Add new template; mirror style of landscape_heatmap.html for visual

consistency.

Dependencies

• e352460b-2d76 — view_spec_json DSL (done)
• Wave-1 Q-OPENQ — supplies open-question source
• Wave-1 Q-PROP — supplies proposed-artifact source
• 41620b88-115d — seed-dashboard pattern (done)

Work Log

2026-04-27 — Implementation (task:d00d502f)

Approach: Implemented two complementary paths:

GET /disease/{name:path} (api.py:61748) — direct synthesis page using the same

science_field_page pattern. Queries live DB on every request (no cache). Uses
translate(LOWER(col), '''', '') normalization to match apostrophe-containing disease
names (e.g. "Alzheimer's disease") from slugs like "alzheimers-disease". Synthesizes:
hypotheses + market prices, debates, knowledge gaps, notebooks/artifacts, challenges,
allocation proposals. Two-column layout, stats bar, graceful empty-section handling.

GET /disease-landing/{slug:path} (api.py:62180) — 302 redirect to /disease/{slug};

exists as the spec-named canonical URL while reusing the above page.

GET /diseases/{name:path} — 301 redirect to /disease/{name} (was going to wiki).

scidex/senate/dashboard_engine.py:

- Added challenges, prediction_markets, allocation_proposals to ALLOWED_TABLES
- Added disease_landing.html Jinja2 template to _TEMPLATES (three-column synthesis
layout with momentum strip, What We Know / Debating / Don't Know grid, bottom row
of challenges + proposals + papers)

scripts/seed_disease_landing_dashboards.py — registers the generic "disease-landing"

dashboard artifact with 9 parameterized data sources (:disease_pat, :disease_stem
runtime params). Run once to seed the artifact, then use
/api/dashboard/disease-landing/render?params={"disease_pat":"%alzheimers%",...}.

tests/test_disease_landing.py — 9 pytest tests: template registration, ALLOWED_TABLES,

200 responses for Alzheimer's and ALS, section header presence, redirect behaviors,
graceful degradation for unknown diseases. All pass.

Acceptance criteria status: