feat: ship source-backed Gradio Q&A

README.md

@@ -17,10 +17,11 @@ tags:
 
 # CoastWise
 
-CoastWise is an offline-first California coastal species and harvest-safety
-lookup MVP. It helps users search curated San Francisco coast species and
-categories, read conservative rule cards, see cached CDFW/CDPH source links, and
-understand when to verify before harvesting.
+CoastWise is an offline-first California ocean regulation and harvest-safety
+Q&A MVP. It lets users ask practical questions such as "what's the min size of
+lingcod?" or "can I eat mussels from Pacifica today?" and answers from cached
+official CDFW/CDPH source snapshots with source links, retrieval time,
+freshness status, uncertainty, and a next safe action.
 
 The MVP is a decision-support field guide. It does not certify species
 identification, legality, harvest authorization, or food safety.
@@ -31,13 +32,13 @@ Target track: Backyard AI.
 
 Target badges and awards, pending final evidence:
 
-- Best Agent: active-session workflow trace from search/photo candidate to
-  deterministic rule/advisory result and constrained explanation.
+- Best Agent: source-backed question interpretation, evidence selection, and
+  constrained wording over deterministic CDFW/CDPH facts or snippets.
 - OpenBMB: MiniCPM5-1B is the planned small-model path for constrained
   safety-result explanation.
 - Tiny Titan: the submitted model inventory stays at or below 4B parameters.
-- Off the Grid: cached name lookup, rule cards, and advisory overlays work
-  without cloud APIs.
+- Off the Grid: bundled seed cache and last-good runtime cache can answer
+  supported questions without cloud APIs.
 - Off Brand: custom CoastWise Gradio interface.
 - Best Demo / Field Notes: public demo video, social post, and write-up links
   will be added before submission.
@@ -50,25 +51,35 @@ evidence.
 
 ## MVP Scope
 
-- Name search for curated San Francisco coast species, aliases, and grouped
-  categories.
-- Conservative rule cards backed by official CDFW/CDPH source metadata.
-- CDPH advisory and annual mussel quarantine overlays.
-- Stale-data handling that defaults harvest guidance to verification.
-- Candidate-only photo support using approved reference-image categories.
-- A constrained generated explanation over deterministic safety results.
-- Session-only handling for photos, query text, generated explanations, and
-  location text.
+- Natural-language Q&A for California ocean sport fishing regulations and CDPH
+  advisory/quarantine topics.
+- Statewide official source registry covering CDFW ocean regulations,
+  in-season changes, Northern, Mendocino, San Francisco, Central, Southern, SF
+  Bay, Groundfish Summary, Crabs, CDPH Shellfish Advisories, and CDPH Annual
+  Mussel Quarantine sources.
+- Structured answers for the reviewed validation set, including lingcod size,
+  lingcod bag limit, Dungeness crab guidance, CDPH shellfish advisories, and
+  annual mussel quarantine.
+- Cited official snippet fallback when cached text exists but structured
+  extraction is incomplete.
+- Startup/manual refresh with last-good cache preservation and visible
+  updated/unchanged/failed source status.
+- 24-hour stale-data handling that still answers from cache when offline while
+  making verification prominent.
+- Session-only handling for questions, optional location text, generated
+  wording, and refresh UI state.
+- Photo candidate support is not part of the core source-backed MVP flow.
 
 ## Model Inventory
 
 | Model | Provider | Source URL | Role | Parameters | Runtime mode | Required for core safety | Award relevance |
 |---|---|---|---|---:|---|---|---|
-| MiniCPM5-1B | OpenBMB | https://huggingface.co/openbmb/MiniCPM5-1B | Constrained safety-result explanation | 1.0B | Disabled/mocked fallback in current local MVP | No | OpenBMB, Tiny Titan |
+| MiniCPM5-1B | OpenBMB | https://huggingface.co/openbmb/MiniCPM5-1B | Optional constrained wording over deterministic answers | 1.0B | Disabled/mocked fallback in current local MVP | No | OpenBMB, Tiny Titan |
 
-Optional future path: MiniCPM-V 4.6 may assist or rerank photo candidates only
-inside the curated reference-image candidate set. No model fine-tuning is
-required for the MVP.
+No model fine-tuning is required for the MVP. Model output, when enabled, may
+only help with interpretation or wording and cannot create or override
+regulation facts, advisory facts, source links, dates, freshness, status, or
+next safe action.
 
 All listed models are under the hackathon 32B cap. Tiny Titan remains eligible
 only while every submitted model stays at or below 4B total parameters.
@@ -76,19 +87,31 @@ only while every submitted model stays at or below 4B total parameters.
 ## Official Sources And Freshness
 
 Cached official source metadata is bundled for offline lookup. CoastWise treats
-cached rule or advisory data older than 30 days as stale and caps harvest-minded
-guidance at `verify_before_harvest` unless a stricter `do_not_harvest` rule
-already applies.
+cached official source snapshots older than 24 hours as stale, but stale cache
+may still be used when offline with a visible verification warning.
 
-Cache date for the current curated source set: 2026-06-13.
+Seed cache retrieval time for the current reviewed source set:
+2026-06-14T12:00:00+00:00.
 
 | Source | Use | Cached URL |
 |---|---|---|
-| CDFW San Francisco Region Ocean Regulations | San Francisco coast species and rule-card baseline | https://wildlife.ca.gov/Fishing/Ocean/Regulations/Fishing-Map/San-Francisco |
-| CDFW Ocean Sport Fishing | General ocean sport fishing source for grouped cards | https://wildlife.ca.gov/Fishing/Ocean/Regulations |
+| CDFW Ocean Sport Fishing | Statewide ocean sport fishing source registry entry | https://wildlife.ca.gov/Fishing/Ocean/Regulations |
+| CDFW Ocean Sport Fishing In-Season Regulation Changes | Statewide in-season changes source registry entry | https://wildlife.ca.gov/Fishing/Ocean/Regulations/Inseason |
+| CDFW Northern Ocean Region | Northern California ocean region source registry entry | https://wildlife.ca.gov/Fishing/Ocean/Regulations/Fishing-Map/Northern |
+| CDFW Mendocino Ocean Region | Mendocino ocean region source registry entry | https://wildlife.ca.gov/Fishing/Ocean/Regulations/Fishing-Map/Mendocino |
+| CDFW San Francisco Ocean Region | San Francisco ocean region source registry entry | https://wildlife.ca.gov/Fishing/Ocean/Regulations/Fishing-Map/San-Francisco |
+| CDFW Central Ocean Region | Central California ocean region source registry entry | https://wildlife.ca.gov/Fishing/Ocean/Regulations/Fishing-Map/Central |
+| CDFW Southern Ocean Region | Southern California ocean region source registry entry | https://wildlife.ca.gov/Fishing/Ocean/Regulations/Fishing-Map/Southern |
+| CDFW San Francisco Bay Region | San Francisco Bay region source registry entry | https://wildlife.ca.gov/Fishing/Ocean/Regulations/Fishing-Map/SF-Bay |
+| CDFW Groundfish Summary | Lingcod validation-set size and bag facts | https://wildlife.ca.gov/Fishing/Ocean/Regulations/Groundfish-Summary |
+| CDFW Crabs | Dungeness crab validation-set guidance and crab source snippets | https://wildlife.ca.gov/Conservation/Marine/Invertebrates/Crabs |
 | CDPH Shellfish Advisories | Shellfish advisory overlay source | https://www.cdph.ca.gov/Programs/OPA/Pages/Shellfish-Advisories.aspx |
 | CDPH Annual Mussel Quarantine | Annual mussel quarantine overlay source | https://www.cdph.ca.gov/Programs/CEH/DRSEM/Pages/EMB/Shellfish/Annual-Mussel-Quarantine.aspx |
 
+Only public official source snapshots and refresh metadata are written to the
+runtime cache. Questions and optional beach/coastal-area text are not saved by
+CoastWise.
+
 ## Demo And Public Evidence
 
 - Demo video: pending, add a public link before final submission.
@@ -98,6 +121,19 @@ Cache date for the current curated source set: 2026-06-13.
 - Public Space history: pending, push meaningful Codex-attributed commits to the
   Hugging Face Space repository before claiming the OpenAI Codex award.
 
+## Release Checklist
+
+- README frontmatter includes Build Small hackathon tags.
+- Public Hugging Face Space URL is available and listed above.
+- Demo video and social post links are listed above.
+- Public Space history contains judge-visible Codex commit evidence for the
+  OpenAI Codex prize.
+- GitHub may remain private; do not rely on private GitHub commits as the only
+  Codex evidence.
+- Source freshness policy, seed/offline behavior, safety limits, privacy
+  limits, and model inventory are documented in this README.
+- Full pytest suite passes before final Space push.
+
 ## Local Run
 
 ```bash

app.py

@@ -2,6 +2,8 @@
 
 from __future__ import annotations
 
+from datetime import datetime, timezone
+
 import gradio as gr
 
 from coastwise.ai import (
@@ -12,23 +14,41 @@ from coastwise.ai import (
 )
 from coastwise.data import get_rule_card
 from coastwise.photo import candidate_from_photo, photo_result_to_dict
+from coastwise.qna import answer_question, answer_to_dict
 from coastwise.rules import compose_safety_result, default_context, result_to_dict
 from coastwise.schemas import CoastalRegion, UserIntent
 from coastwise.search import search_species
+from coastwise.source_cache import DEFAULT_SEED_CACHE_DIR, load_seed_cache
+from coastwise.source_refresh import refresh_official_sources
+from coastwise.source_registry import load_official_sources
 from coastwise.submission import default_model_inventory
 
 
 APP_TITLE = "CoastWise"
 
-INTRO = """
-Offline California coastal species lookup.
+SOURCE_BACKED_CONTROL_LABELS = (
+    "Question",
+    "Beach or coastal area",
+    "Ask",
+    "Refresh official sources",
+    "Source status",
+)
 
-SF coast harvest-safety guidance.
+INTRO = """
+California CDFW/CDPH Q&A.
 """
 
 APP_CSS = """
+html,
+body {
+    max-width: 100% !important;
+    overflow-x: hidden !important;
+}
+
 .gradio-container {
+    width: 100% !important;
     max-width: 100vw !important;
+    min-width: 0 !important;
     overflow-x: hidden !important;
 }
 
@@ -41,6 +61,7 @@ APP_CSS = """
 .gradio-container .block,
 .gradio-container .wrap,
 .gradio-container fieldset {
+    width: 100% !important;
     max-width: 100% !important;
     min-width: 0 !important;
 }
@@ -51,6 +72,14 @@ APP_CSS = """
 .gradio-container h2,
 .gradio-container h3 {
     overflow-wrap: anywhere;
+    word-break: break-word;
+    white-space: normal !important;
+}
+
+.gradio-container input,
+.gradio-container textarea,
+.gradio-container button {
+    max-width: 100% !important;
 }
 
 @media (max-width: 640px) {
@@ -58,6 +87,11 @@ APP_CSS = """
         padding-left: 16px !important;
         padding-right: 16px !important;
     }
+
+    .gradio-container h1 {
+        font-size: 28px !important;
+        line-height: 1.2 !important;
+    }
 }
 """
 
@@ -71,6 +105,15 @@ Does not authorize harvest or food safety.
 Verify CDFW/CDPH before harvest.
 """
 
+SOURCE_BACKED_PRODUCT_LIMITS = (
+    "CoastWise does not authorize harvest or act as legal/food-safety/species-ID "
+    "authority. Verify official CDFW/CDPH guidance before harvest or consumption."
+)
+
+SESSION_ONLY_NOTICE = (
+    "Session-only: questions and optional locations are not saved."
+)
+
 INTENT_LABELS = {
     "Just curious": UserIntent.JUST_CURIOUS,
     "Catch and release": UserIntent.CATCH_AND_RELEASE,
@@ -84,6 +127,97 @@ REGION_LABELS = {
 }
 
 
+def _active_cache():
+    return load_seed_cache(DEFAULT_SEED_CACHE_DIR)
+
+
+def ask_source_backed_question(question: str, beach_or_coastal_area: str = "") -> dict[str, object]:
+    answer = answer_question(
+        question,
+        beach_or_coastal_area,
+        _active_cache(),
+        datetime.now(timezone.utc),
+    )
+    return answer_to_dict(answer)
+
+
+def _render_source_backed_answer_payload(result: dict[str, object]) -> str:
+    source_lines = "\n".join(
+        f"- [{title}]({url})"
+        for title, url in zip(result["source_titles"], result["source_links"], strict=False)
+    ) or "- No source-backed evidence available."
+    snippet_lines = "\n".join(f"> {snippet}" for snippet in result["supporting_snippets"]) or "- No snippet shown."
+    uncertainty = "\n".join(f"- {note}" for note in result["uncertainty_notes"]) or "- None."
+    location_warning = result["location_warning"] or "Location recognized or not needed for this general answer."
+    retrieved = result["retrieved_at"] or "No cached retrieval time available"
+    return f"""
+## Direct answer
+{result["direct_answer"]}
+
+**Status:** `{result["status"]}`  
+**Evidence type:** `{result["evidence_type"]}`  
+**Freshness:** `{result["freshness_status"]}`  
+**Retrieved:** {retrieved}
+
+## Source evidence
+{source_lines}
+
+## Location context
+{result["matched_place"] or "General California ocean"}  
+{location_warning}
+
+## Supporting snippet
+{snippet_lines}
+
+## Uncertainty
+{uncertainty}
+
+## Next safe action
+{result["next_safe_action"]}
+
+## Product limits
+{SOURCE_BACKED_PRODUCT_LIMITS}
+
+{SESSION_ONLY_NOTICE}
+"""
+
+
+def render_source_backed_answer_from_cache(
+    question: str,
+    beach_or_coastal_area: str,
+    cache,
+    now: datetime,
+) -> str:
+    answer = answer_question(question, beach_or_coastal_area, cache, now)
+    return _render_source_backed_answer_payload(answer_to_dict(answer))
+
+
+def render_source_backed_answer(question: str, beach_or_coastal_area: str = "") -> str:
+    result = ask_source_backed_question(question, beach_or_coastal_area)
+    return _render_source_backed_answer_payload(result)
+
+
+def refresh_official_sources_status(fetcher=None) -> dict[str, object]:
+    result, _cache = refresh_official_sources(
+        load_official_sources(),
+        _active_cache(),
+        datetime.now(timezone.utc),
+        fetcher,
+    )
+    return {
+        "message": result.user_message,
+        "updated": result.updated_source_ids,
+        "unchanged": result.unchanged_source_ids,
+        "failed": result.failed_source_ids,
+        "cache_preserved": result.cache_preserved_source_ids,
+    }
+
+
+def render_refresh_status() -> str:
+    status = refresh_official_sources_status()
+    return f"**Refresh official sources:** {status['message']}"
+
+
 def lookup_species(
     query: str,
     intent_label: str,
@@ -195,30 +329,19 @@ def build_app() -> gr.Blocks:
         gr.Markdown(f"# {APP_TITLE}")
         gr.Markdown(INTRO)
         gr.Markdown(LIMITS)
-        query = gr.Textbox(label="Species or category", placeholder="Try mussel, lingcod, crab")
-        intent = gr.Radio(
-            ["Just curious", "Catch and release", "Thinking of harvest"],
-            value="Just curious",
-            label="Intent",
-        )
-        region = gr.Dropdown(
-            ["Point Arena to Pigeon Point", "San Francisco Bay", "Unknown"],
-            value="Point Arena to Pigeon Point",
-            label="Region",
-        )
-        county_or_beach = gr.Textbox(label="Optional county or beach", placeholder="Optional")
-        lookup = gr.Button("Look up")
-        output = gr.Markdown("Rule-card results will appear here.")
+        gr.Markdown(SESSION_ONLY_NOTICE)
+        gr.Markdown("**Source status:** seed cache loaded; refresh available.")
+        query = gr.Textbox(label="Question", placeholder="what's the min size of lingcod?")
+        county_or_beach = gr.Textbox(label="Beach or coastal area", placeholder="Optional, e.g. Pacifica or Monterey")
+        lookup = gr.Button("Ask")
+        refresh = gr.Button("Refresh official sources")
+        output = gr.Markdown("Source-backed answers will appear here.")
         lookup.click(
-            render_lookup_result,
-            inputs=[query, intent, region, county_or_beach],
+            render_source_backed_answer,
+            inputs=[query, county_or_beach],
             outputs=output,
         )
-        gr.Markdown("## Photo candidate support")
-        photo = gr.Image(label="Optional photo", type="pil")
-        photo_button = gr.Button("Compare photo to curated references")
-        photo_output = gr.Markdown("Photo candidates will appear here.")
-        photo_button.click(render_photo_candidate, inputs=photo, outputs=photo_output)
+        refresh.click(render_refresh_status, outputs=output)
     return demo
 
 

coastwise/advisories.py

@@ -5,11 +5,25 @@ from __future__ import annotations
 from datetime import date
 
 from coastwise.data import CDPH_QUARANTINE_URL, CDPH_SHELLFISH_URL, SOURCE_CACHE_DATE
-from coastwise.schemas import Advisory, CoastalRegion, HarvestStatus, QueryContext, RuleCard
+from coastwise.schemas import Advisory, AdvisoryFact, CoastalRegion, HarvestStatus, QueryContext, RuleCard
+from coastwise.source_cache import SourceCache
 
 
 SHELLFISH_CARD_KEYS = {"mussels", "clams"}
 SHELLFISH_CATEGORIES = {"shellfish"}
+SOURCE_BACKED_SHELLFISH_TERMS = {
+    "bivalve",
+    "bivalves",
+    "clam",
+    "clams",
+    "crab",
+    "crabs",
+    "mussel",
+    "mussels",
+    "oyster",
+    "oysters",
+    "shellfish",
+}
 
 ANNUAL_MUSSEL_QUARANTINE_WARNING = (
     "Annual CDPH mussel quarantine runs May 1 through October 31; do not harvest "
@@ -87,3 +101,38 @@ def matching_advisories(rule_card: RuleCard, context: QueryContext) -> list[Advi
 
     return matches
 
+
+def _normalize_source_backed_term(value: str | None) -> str:
+    return " ".join((value or "").replace("_", " ").casefold().split())
+
+
+def _advisory_fact_terms(fact: AdvisoryFact) -> set[str]:
+    affected = _normalize_source_backed_term(fact.affected_species_or_category)
+    terms = {affected}
+    if affected == "mussels":
+        terms.update({"mussel", "shellfish"})
+    if affected == "shellfish":
+        terms.update(SOURCE_BACKED_SHELLFISH_TERMS)
+    return {term for term in terms if term}
+
+
+def source_backed_advisory_facts(cache: SourceCache, species_or_category: str | None) -> list[AdvisoryFact]:
+    """Return cached CDPH advisory facts matching a shellfish species/category."""
+
+    normalized = _normalize_source_backed_term(species_or_category)
+    if not normalized:
+        return []
+
+    requested_terms = {normalized}
+    if normalized in SOURCE_BACKED_SHELLFISH_TERMS:
+        requested_terms.add("shellfish")
+    if normalized in {"mussel", "mussels"}:
+        requested_terms.update({"mussel", "mussels"})
+
+    matches = [
+        fact
+        for fact in cache.advisory_facts
+        if requested_terms & _advisory_fact_terms(fact)
+        or any(term in fact.supporting_snippet.casefold() for term in requested_terms)
+    ]
+    return sorted(matches, key=lambda fact: (fact.effective_start is None, fact.id))

coastwise/submission.py

@@ -22,7 +22,7 @@ def default_model_inventory() -> ModelInventory:
                 key="minicpm5_1b",
                 name="MiniCPM5-1B",
                 provider="OpenBMB",
-                source_url="https://huggingface.co/openbmb",
+                source_url="https://huggingface.co/openbmb/MiniCPM5-1B",
                 total_parameters_b=1.0,
                 roles=(ModelRole.SAFETY_EXPLANATION,),
                 runtime_mode=RuntimeMode.DISABLED_OR_MOCKED,
@@ -70,6 +70,14 @@ def validate_submission_package(package: SubmissionPackage) -> list[ValidationIs
         if not value:
             issues.append(ValidationIssue(field, "error", f"{field} is required."))
 
+    if package.space_url and "huggingface.co/spaces/" not in package.space_url:
+        issues.append(
+            ValidationIssue(
+                "space_url",
+                "error",
+                "Public Space URL must point to the Hugging Face Spaces submission.",
+            )
+        )
     if not package.readme_has_frontmatter_tags:
         issues.append(ValidationIssue("readme_has_frontmatter_tags", "error", "README frontmatter tags are required."))
     if not package.safety_limitations_documented:
@@ -84,6 +92,13 @@ def validate_submission_package(package: SubmissionPackage) -> list[ValidationIs
                 "OpenAI Codex claim requires public Hugging Face Space commit evidence.",
             )
         )
+    if package.codex_evidence and "huggingface.co/spaces/" not in package.codex_evidence:
+        issues.append(
+            ValidationIssue(
+                "codex_evidence",
+                "error",
+                "Codex evidence should be judge-visible in public Hugging Face Space history.",
+            )
+        )
     issues.extend(validate_model_inventory(package.model_inventory))
     return issues
-

tests/test_advisories.py

@@ -3,9 +3,11 @@
 from datetime import date
 
 from coastwise.advisories import annual_mussel_quarantine_applies, matching_advisories
+from coastwise.advisories import source_backed_advisory_facts
 from coastwise.data import get_rule_card
 from coastwise.rules import default_context
 from coastwise.schemas import CoastalRegion, HarvestStatus, UserIntent
+from coastwise.source_cache import DEFAULT_SEED_CACHE_DIR, load_seed_cache
 
 
 def test_annual_mussel_quarantine_applies_may_through_october():
@@ -32,3 +34,12 @@ def test_matching_cdph_advisories_include_region_level_shellfish_warning():
         and "CDPH" in advisory.source_name
         for advisory in advisories
     )
+
+
+def test_source_backed_cdph_advisory_facts_include_quarantine_and_cooking_warning():
+    facts = source_backed_advisory_facts(load_seed_cache(DEFAULT_SEED_CACHE_DIR), "mussels")
+
+    assert facts
+    assert any("May 1" in fact.advisory_status for fact in facts)
+    assert any("Cooking does not reliably destroy" in fact.warning for fact in facts)
+    assert all("cdph.ca.gov" in fact.source_url for fact in facts)

tests/test_app.py

@@ -1,9 +1,14 @@
 """Gradio app integration tests."""
 
+import time
+from dataclasses import replace
+from datetime import datetime, timedelta, timezone
+
 import gradio as gr
 from PIL import Image
 
 import app
+from coastwise.source_cache import DEFAULT_SEED_CACHE_DIR, SourceCache, load_seed_cache
 
 
 def test_app_import_exposes_coastwise_identity_without_launching():
@@ -18,6 +23,118 @@ def test_build_app_returns_blocks_with_lookup_shell():
     assert demo.title == "CoastWise"
 
 
+def test_source_backed_app_shell_exposes_primary_controls_without_legacy_inputs():
+    assert "Question" in app.SOURCE_BACKED_CONTROL_LABELS
+    assert "Beach or coastal area" in app.SOURCE_BACKED_CONTROL_LABELS
+    assert "Refresh official sources" in app.SOURCE_BACKED_CONTROL_LABELS
+    assert "Intent" not in app.SOURCE_BACKED_CONTROL_LABELS
+    assert "Region" not in app.SOURCE_BACKED_CONTROL_LABELS
+    assert "Photo candidate" not in app.SOURCE_BACKED_CONTROL_LABELS
+
+
+def test_source_backed_question_journey_renders_evidence_and_next_action():
+    result = app.ask_source_backed_question("what's the min size of lingcod?", "")
+    markdown = app.render_source_backed_answer("what's the min size of lingcod?", "")
+
+    assert result["direct_answer"]
+    assert result["source_links"]
+    assert result["freshness_status"] == "current"
+    assert "22" in markdown
+    assert "Source evidence" in markdown
+    assert "Retrieved" in markdown
+    assert "Next safe action" in markdown
+
+
+def test_manual_refresh_status_uses_nontechnical_source_group_language():
+    status = app.refresh_official_sources_status(fetcher=lambda _source: "unchanged text")
+
+    assert "CDFW" in status["message"]
+    assert "CDPH" in status["message"]
+    assert "checked" in status["message"].lower()
+    assert "traceback" not in status["message"].lower()
+
+
+def test_app_renders_stale_and_no_cache_source_backed_states():
+    now = datetime(2026, 6, 14, 13, tzinfo=timezone.utc)
+    seed = load_seed_cache(DEFAULT_SEED_CACHE_DIR)
+    stale = SourceCache(
+        sources=seed.sources,
+        snapshots=tuple(replace(snapshot, retrieved_at=now - timedelta(hours=25)) for snapshot in seed.snapshots),
+        chunks=seed.chunks,
+        facts=seed.facts,
+        advisory_facts=seed.advisory_facts,
+    )
+
+    stale_markdown = app.render_source_backed_answer_from_cache(
+        "what's the min size of lingcod?",
+        "",
+        stale,
+        now,
+    )
+    no_cache_markdown = app.render_source_backed_answer_from_cache(
+        "what's the min size of lingcod?",
+        "",
+        SourceCache((), (), (), ()),
+        now,
+    )
+
+    assert "stale" in stale_markdown.lower()
+    assert "refresh official" in no_cache_markdown.lower()
+
+
+def test_app_renders_cdph_warning_and_verification_for_mussels():
+    markdown = app.render_source_backed_answer("can I eat mussels from Pacifica today?", "")
+    lowered = markdown.lower()
+
+    assert "cdph" in lowered
+    assert "cooking does not reliably" in lowered
+    assert "verify" in lowered
+    assert "safe to eat" not in lowered
+
+
+def test_app_renders_source_evidence_limits_and_partial_unsupported_states():
+    now = datetime(2026, 6, 14, 13, tzinfo=timezone.utc)
+    partial = app.render_source_backed_answer("what should I check for ocean rules near Monterey?", "Monterey")
+    unsupported = app.render_source_backed_answer("what is the trout limit in Lake Tahoe?", "")
+    no_cache = app.render_source_backed_answer_from_cache(
+        "what's the min size of lingcod?",
+        "",
+        SourceCache((), (), (), ()),
+        now,
+    )
+    combined = "\n".join((partial, unsupported, no_cache)).casefold()
+
+    assert "source evidence" in partial.casefold()
+    assert "uncertainty" in partial.casefold()
+    assert "partial" in partial.casefold()
+    assert "outside this mvp" in unsupported.casefold()
+    assert "refresh official" in no_cache.casefold()
+    assert "does not authorize harvest" in combined
+    assert "session-only" in combined
+    assert "source-backed evidence" in combined
+
+
+def test_cached_answer_app_build_and_refresh_feedback_fit_latency_budgets():
+    start = time.perf_counter()
+    markdown = app.render_source_backed_answer("what's the min size of lingcod?", "")
+    answer_seconds = time.perf_counter() - start
+
+    start = time.perf_counter()
+    demo = app.build_app()
+    build_seconds = time.perf_counter() - start
+
+    start = time.perf_counter()
+    status = app.refresh_official_sources_status(fetcher=lambda _source: "unchanged official source text")
+    refresh_seconds = time.perf_counter() - start
+
+    assert "22" in markdown
+    assert isinstance(demo, gr.Blocks)
+    assert "checked" in status["message"].casefold()
+    assert answer_seconds < 3
+    assert build_seconds < 3
+    assert refresh_seconds < 3
+
+
 def test_lookup_species_returns_source_backed_result_dict():
     result = app.lookup_species("lingcod", "Just curious", "Point Arena to Pigeon Point", "")
 

tests/test_privacy.py

@@ -38,3 +38,22 @@ def test_rendered_ui_copy_does_not_offer_save_export_or_history_actions():
     assert "save" not in markdown
     assert "export" not in markdown
     assert "history" not in markdown
+
+
+def test_source_backed_question_location_and_wording_are_session_only():
+    result = app.ask_source_backed_question("what's the min size of lingcod?", "Pacifica")
+    markdown = app.render_source_backed_answer("what's the min size of lingcod?", "Pacifica").casefold()
+
+    assert result["matched_place"] == "Pacifica"
+    assert "session-only" in markdown
+    assert "questions and optional locations are not saved" in markdown
+    assert not hasattr(app, "SOURCE_BACKED_QUESTION_HISTORY")
+    assert not hasattr(app, "SOURCE_BACKED_LOCATION_HISTORY")
+    assert not hasattr(app, "GENERATED_ANSWER_HISTORY")
+
+
+def test_source_backed_controls_do_not_offer_account_save_export_history_or_photo_core():
+    labels = " ".join(app.SOURCE_BACKED_CONTROL_LABELS).casefold()
+
+    for forbidden in ("account", "save", "export", "history", "photo", "camera"):
+        assert forbidden not in labels

tests/test_safety_copy.py

@@ -6,8 +6,11 @@ import app
 FORBIDDEN_PHRASES = (
     "safe to eat",
     "definitely legal",
+    "legal to harvest",
     "authorized to harvest",
+    "certifies",
     "confirmed identification",
+    "you may harvest",
 )
 
 
@@ -32,3 +35,16 @@ def test_photo_candidate_copy_never_claims_confirmed_identification():
     assert "harvest guidance" not in lowered
     for phrase in FORBIDDEN_PHRASES:
         assert phrase not in lowered
+
+
+def test_source_backed_answers_show_limits_without_forbidden_claims():
+    markdown = app.render_source_backed_answer("can I eat mussels from Pacifica today?", "")
+    lowered = markdown.casefold()
+
+    assert "does not authorize harvest" in lowered
+    assert "food-safety" in lowered
+    assert "session-only" in lowered
+    assert "source evidence" in lowered
+    assert "retrieved" in lowered
+    for phrase in FORBIDDEN_PHRASES:
+        assert phrase not in lowered

tests/test_submission.py

@@ -1,6 +1,10 @@
 """Submission-readiness tests."""
 
+import json
+
 from coastwise.schemas import AwardTag, SubmissionPackage
+from coastwise.source_cache import DEFAULT_SEED_CACHE_DIR, DEFAULT_SOURCE_CACHE_ROOT, load_seed_cache
+from coastwise.source_registry import load_official_sources, statewide_source_coverage
 from coastwise.submission import (
     default_model_inventory,
     validate_model_inventory,
@@ -37,3 +41,29 @@ def test_submission_package_validator_flags_missing_public_evidence():
     fields = {issue.field for issue in issues}
 
     assert {"space_url", "demo_video_url", "social_post_url", "codex_evidence"} <= fields
+
+
+def test_readme_documents_statewide_source_cache_policy_and_validation_limits():
+    readme = open("README.md", encoding="utf-8").read().casefold()
+    coverage = statewide_source_coverage()
+
+    assert "24 hours" in readme
+    assert "statewide official source registry" in readme
+    assert "structured answers for the reviewed validation set" in readme
+    assert "snippet fallback" in readme
+    assert "questions and optional beach/coastal-area text are not saved" in readme
+    for source in load_official_sources():
+        if source.id in coverage.required_source_ids:
+            assert source.url.casefold() in readme
+
+
+def test_seed_manifest_documents_statewide_coverage_and_24_hour_stale_policy():
+    manifest = json.loads((DEFAULT_SOURCE_CACHE_ROOT / "seed-manifest.json").read_text(encoding="utf-8"))
+    cache = load_seed_cache(DEFAULT_SEED_CACHE_DIR)
+    source_ids = set(cache.snapshots_by_source_id)
+
+    assert manifest["coverage_id"] == statewide_source_coverage().id
+    assert manifest["stale_after_hours"] == 24
+    assert set(statewide_source_coverage().required_source_ids) <= source_ids
+    assert {"lingcod_min_size", "dungeness_crab_bag_limit"} <= set(cache.facts_by_id)
+    assert {"cdph_annual_mussel_quarantine", "cdph_shellfish_verify"} <= set(cache.advisory_facts_by_id)