diff --git a/helper-scripts/fetch-source-data.sh b/helper-scripts/fetch-source-data.sh
index cbca9b9..d29fee2 100755
--- a/helper-scripts/fetch-source-data.sh
+++ b/helper-scripts/fetch-source-data.sh
@@ -16,7 +16,9 @@
 #     --dest DIR        library directory          (default: ./public/sources)
 #     --seed DIR        copy DIR/* into the library first (no clobber)
 #     --category CAT    only fetch one category    (e.g. 200_Medical)
+#     --collection ID   only fetch one collection  (e.g. austere-medicine)
 #     --list            print the manifest and exit
+#     --list-collections print the declared collections and exit
 #     --validate        lint the manifest (no network) and exit
 #     --force           re-download files that already exist
 #     --strict          exit non-zero if any download failed
@@ -31,6 +33,7 @@ MANIFEST="sources.yaml"
 DEST="public/sources"
 SEED=""
 ONLY_CATEGORY=""
+ONLY_COLLECTION=""
 MODE="fetch"
 FORCE=0
 STRICT=0
@@ -41,7 +44,9 @@ while [[ $# -gt 0 ]]; do
         --dest)     DEST="$2";     shift 2 ;;
         --seed)     SEED="$2";     shift 2 ;;
         --category) ONLY_CATEGORY="$2"; shift 2 ;;
+        --collection) ONLY_COLLECTION="$2"; shift 2 ;;
         --list)     MODE="list";     shift ;;
+        --list-collections) MODE="list-collections"; shift ;;
         --validate) MODE="validate"; shift ;;
         --force)    FORCE=1;  shift ;;
         --strict)   STRICT=1; shift ;;
@@ -73,24 +78,52 @@ parse_manifest() {
         }
         function emit() {
             if (url != "")
-                printf "%s\037%s\037%s\037%s\037%s\n", url, cat, fn, sha, title
-            url = cat = fn = sha = title = ""
+                printf "%s\037%s\037%s\037%s\037%s\037%s\n", url, cat, fn, sha, title, coll
+            url = cat = fn = sha = title = coll = ""
         }
         /^[[:space:]]*#/ { next }
         /^[[:space:]]*-[[:space:]]*url:/      { emit(); url   = val($0); next }
         /^[[:space:]]+filename:/              { fn    = val($0); next }
         /^[[:space:]]+category:/              { cat   = val($0); next }
+        /^[[:space:]]+collection:/            { coll  = val($0); next }
         /^[[:space:]]+sha256:/                { sha   = val($0); next }
         /^[[:space:]]+title:/                 { title = val($0); next }
         END { emit() }
     ' "$MANIFEST"
 }
 
+# Emits one record per collection definition:  id \037 name \037 description
+parse_collections() {
+    awk '
+        function val(line) {
+            sub(/^[^:]*:[[:space:]]*/, "", line)
+            sub(/[[:space:]]+#.*$/, "", line)
+            gsub(/^[[:space:]]+|[[:space:]]+$/, "", line)
+            gsub(/^"|"$/, "", line)
+            return line
+        }
+        function emit() {
+            if (id != "") printf "%s\037%s\037%s\n", id, name, desc
+            id = name = desc = ""
+        }
+        /^collections:[[:space:]]*$/          { incoll = 1; next }
+        /^sources:[[:space:]]*$/              { if (incoll) { emit(); incoll = 0 } }
+        incoll && /^[[:space:]]*-[[:space:]]*id:/ { emit(); id = val($0); next }
+        incoll && /^[[:space:]]+name:/            { name = val($0); next }
+        incoll && /^[[:space:]]+description:/     { desc = val($0); next }
+        END { emit() }
+    ' "$MANIFEST"
+}
+
 # ── Validate ─────────────────────────────────────────────────────────
 validate_manifest() {
     local errors=0 count=0
-    declare -A seen
-    while IFS="$US" read -r url cat fn sha title; do
+    declare -A seen defined
+    # Collections declared in the top-level `collections:` block
+    while IFS="$US" read -r id name desc; do
+        [[ -n "$id" ]] && defined[$id]=1
+    done < <(parse_collections)
+    while IFS="$US" read -r url cat fn sha title coll; do
         count=$((count + 1))
         local where="entry #$count (${fn:-$url})"
         if [[ ! "$url" =~ ^https?:// ]]; then
@@ -111,6 +144,11 @@ validate_manifest() {
         if [[ -z "$title" ]]; then
             echo "  ✗ $where: missing title"; errors=$((errors+1))
         fi
+        if [[ -z "$coll" ]]; then
+            echo "  ✗ $where: missing collection"; errors=$((errors+1))
+        elif [[ -z "${defined[$coll]:-}" ]]; then
+            echo "  ✗ $where: collection '$coll' not declared in the collections: block"; errors=$((errors+1))
+        fi
         if [[ -n "$sha" && ! "$sha" =~ ^[0-9a-fA-F]{64}$ ]]; then
             echo "  ✗ $where: sha256 must be 64 hex chars"; errors=$((errors+1))
         fi
@@ -128,15 +166,28 @@ validate_manifest() {
 
 # ── List ─────────────────────────────────────────────────────────────
 list_manifest() {
-    printf '%-18s %-52s %s\n' "CATEGORY" "FILENAME" "TITLE"
-    while IFS="$US" read -r url cat fn sha title; do
-        printf '%-18s %-52s %s\n' "$cat" "$fn" "$title"
+    printf '%-22s %-18s %-48s %s\n' "COLLECTION" "CATEGORY" "FILENAME" "TITLE"
+    while IFS="$US" read -r url cat fn sha title coll; do
+        printf '%-22s %-18s %-48s %s\n' "$coll" "$cat" "$fn" "$title"
     done < <(parse_manifest | sort)
 }
 
+# Print the declared collections with a live item count for each.
+list_collections() {
+    declare -A cnt
+    while IFS="$US" read -r url cat fn sha title coll; do
+        [[ -n "$coll" ]] && cnt[$coll]=$(( ${cnt[$coll]:-0} + 1 ))
+    done < <(parse_manifest)
+    printf '%-24s %-6s %s\n' "ID" "ITEMS" "NAME"
+    while IFS="$US" read -r id name desc; do
+        printf '%-24s %-6s %s\n' "$id" "${cnt[$id]:-0}" "$name"
+    done < <(parse_collections)
+}
+
 case "$MODE" in
-    validate) echo "Validating $MANIFEST ..."; validate_manifest; exit $? ;;
-    list)     list_manifest; exit 0 ;;
+    validate)         echo "Validating $MANIFEST ..."; validate_manifest; exit $? ;;
+    list)             list_manifest; exit 0 ;;
+    list-collections) list_collections; exit 0 ;;
 esac
 
 # ── Fetch ────────────────────────────────────────────────────────────
@@ -163,8 +214,9 @@ echo "── Fetching manifest: $MANIFEST → $DEST"
 ok=0; skipped=0; failed=0
 failed_list=""
 
-while IFS="$US" read -r url cat fn sha title; do
-    [[ -n "$ONLY_CATEGORY" && "$cat" != "$ONLY_CATEGORY" ]] && continue
+while IFS="$US" read -r url cat fn sha title coll; do
+    [[ -n "$ONLY_CATEGORY"   && "$cat"  != "$ONLY_CATEGORY"   ]] && continue
+    [[ -n "$ONLY_COLLECTION" && "$coll" != "$ONLY_COLLECTION" ]] && continue
 
     dir="$DEST/$cat"
     out="$dir/$fn"
diff --git a/public/app.js b/public/app.js
index 7a39024..fb8813d 100644
--- a/public/app.js
+++ b/public/app.js
@@ -294,6 +294,152 @@
     }
   }
 
+  // ── Add-content modal (catalog import + upload) ─────────────────────
+
+  let importPoll = null;
+
+  function openLibraryModal() {
+    $('library-modal').hidden = false;
+    populateUploadCategories();
+    loadCatalog();
+  }
+
+  function closeLibraryModal() {
+    $('library-modal').hidden = true;
+    if (importPoll) { clearInterval(importPoll); importPoll = null; }  // stop background polling
+  }
+
+  async function loadCatalog() {
+    const list = $('catalog-list');
+    list.textContent = 'Loading…';
+    try {
+      const res = await fetch('/api/catalog');
+      if (!res.ok) throw new Error('HTTP ' + res.status);
+      const data = await res.json();
+      renderCatalog(data);
+      if (data.importing) pollImport();   // resume progress polling if an import is already running
+    } catch (e) {
+      list.textContent = 'Could not load the catalog.';
+    }
+  }
+
+  function renderCatalog(data) {
+    const list = $('catalog-list');
+    list.innerHTML = '';
+    (data.collections || []).forEach((c) => {
+      const done = c.count > 0 && c.installed >= c.count;
+      const row = document.createElement('div');
+      row.className = 'catalog-row';
+
+      const meta = document.createElement('div');
+      meta.className = 'catalog-meta';
+      meta.innerHTML =
+        `<span class="catalog-name">${escapeHtml(c.name)}</span>` +
+        `<span class="catalog-desc">${escapeHtml(c.description)}</span>` +
+        `<span class="catalog-count">${c.installed}/${c.count} installed</span>`;
+
+      const btn = document.createElement('button');
+      btn.type = 'button';
+      btn.className = 'ghost-btn catalog-install' + (done ? ' is-done' : '');
+      btn.textContent = done ? 'Installed' : (data.importing ? 'Working…' : 'Install');
+      btn.disabled = done || data.importing || !data.import_enabled;
+      btn.addEventListener('click', () => importCollection(c.id, btn));
+
+      row.appendChild(meta);
+      row.appendChild(btn);
+      list.appendChild(row);
+    });
+
+    if (!data.import_enabled) {
+      const note = document.createElement('p');
+      note.className = 'modal-note';
+      note.textContent = 'Network import is disabled on this deployment — you can still upload your own files below.';
+      list.appendChild(note);
+    }
+  }
+
+  async function importCollection(id, btn) {
+    btn.disabled = true;
+    btn.textContent = 'Starting…';
+    try {
+      const res = await fetch('/api/import', {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({ collection: id }),
+      });
+      if (res.status === 202) {
+        pollImport();
+      } else {
+        const d = await res.json().catch(() => ({}));
+        btn.textContent = 'Retry';
+        btn.disabled = false;
+        if (res.status === 409) pollImport();
+        console.warn('import:', d.error || res.status);
+      }
+    } catch (e) {
+      btn.textContent = 'Retry';
+      btn.disabled = false;
+    }
+  }
+
+  // While an import runs, refresh the catalog (install progress) and the
+  // sidebar library (docs appear as they download + index), until done.
+  function pollImport() {
+    if (importPoll) return;
+    importPoll = setInterval(async () => {
+      try {
+        const res = await fetch('/api/catalog');
+        if (res.ok) {
+          const data = await res.json();
+          renderCatalog(data);
+          refreshLibrary();
+          if (!data.importing) { clearInterval(importPoll); importPoll = null; }
+        }
+      } catch (e) { /* keep polling */ }
+    }, 3000);
+  }
+
+  function populateUploadCategories() {
+    const sel = $('upload-category');
+    if (sel.options.length) return;
+    Object.keys(CATEGORY_LABELS).forEach((cat) => {
+      const o = document.createElement('option');
+      o.value = cat;
+      o.textContent = CATEGORY_LABELS[cat];
+      sel.appendChild(o);
+    });
+  }
+
+  async function uploadFile(file) {
+    const status = $('upload-status');
+    status.hidden = false;
+    status.className = 'modal-status';
+    status.textContent = 'Uploading ' + file.name + '…';
+    const fd = new FormData();
+    fd.append('category', $('upload-category').value);
+    fd.append('file', file);
+    try {
+      const res = await fetch('/api/upload', { method: 'POST', body: fd });
+      const d = await res.json().catch(() => ({}));
+      if (res.ok) {
+        status.textContent = 'Uploaded ' + (d.filename || file.name) + ' — indexing shortly.';
+        resetUploadPicker();
+        refreshLibrary();
+      } else {
+        status.className = 'modal-status is-error';
+        status.textContent = 'Upload failed: ' + (d.error || ('HTTP ' + res.status));
+      }
+    } catch (e) {
+      status.className = 'modal-status is-error';
+      status.textContent = 'Upload failed.';
+    }
+  }
+
+  function resetUploadPicker() {
+    $('upload-file').value = '';
+    $('upload-drop-text').textContent = 'Choose a PDF or TXT file, or drop it here…';
+  }
+
   // ── Query flow ─────────────────────────────────────────────────────
 
   function setBusy(busy) {
@@ -410,6 +556,44 @@
     $('sidebar-toggle').addEventListener('click', () => toggleSidebar());
     $('backdrop').addEventListener('click', () => toggleSidebar(false));
 
+    // ── Add-content modal wiring ──────────────────────────────────
+    $('library-add-btn').addEventListener('click', openLibraryModal);
+    $('library-modal-close').addEventListener('click', closeLibraryModal);
+    $('library-modal-backdrop').addEventListener('click', closeLibraryModal);
+    document.addEventListener('keydown', (e) => {
+      if (e.key === 'Escape' && !$('library-modal').hidden) closeLibraryModal();
+    });
+
+    const fileInput = $('upload-file');
+    const drop = $('upload-drop');
+    drop.addEventListener('click', () => fileInput.click());
+    fileInput.addEventListener('change', () => {
+      if (fileInput.files.length) $('upload-drop-text').textContent = fileInput.files[0].name;
+    });
+    ['dragover', 'dragenter'].forEach((ev) =>
+      drop.addEventListener(ev, (e) => { e.preventDefault(); drop.classList.add('dragover'); }));
+    ['dragleave', 'dragend'].forEach((ev) =>
+      drop.addEventListener(ev, () => drop.classList.remove('dragover')));
+    drop.addEventListener('drop', (e) => {
+      e.preventDefault();
+      drop.classList.remove('dragover');
+      if (e.dataTransfer.files.length) {
+        fileInput.files = e.dataTransfer.files;
+        $('upload-drop-text').textContent = e.dataTransfer.files[0].name;
+      }
+    });
+    $('upload-form').addEventListener('submit', (e) => {
+      e.preventDefault();
+      if (fileInput.files.length) {
+        uploadFile(fileInput.files[0]);
+      } else {
+        const s = $('upload-status');
+        s.hidden = false;
+        s.className = 'modal-status';
+        s.textContent = 'Pick a file first.';
+      }
+    });
+
     $('user-input').focus();
 
     refreshStatus();
diff --git a/public/index.html b/public/index.html
index b0443f7..2e0b8c8 100644
--- a/public/index.html
+++ b/public/index.html
@@ -39,7 +39,10 @@ <h2 class="panel-title">System</h2>
       </section>
 
       <section class="panel panel-grow" aria-label="Field library">
-        <h2 class="panel-title">Field library <span class="badge" id="library-count">0</span></h2>
+        <h2 class="panel-title">Field library <span class="badge" id="library-count">0</span>
+          <button class="ghost-btn panel-action" id="library-add-btn" type="button"
+                  aria-haspopup="dialog">+ Add</button>
+        </h2>
         <div class="library" id="library-list">
           <p class="library-empty">No documents indexed yet. Run the content
           fetcher or drop PDFs into the sources volume — they are indexed
@@ -96,6 +99,42 @@ <h2 class="panel-title">Field library <span class="badge" id="library-count">0</
     </main>
   </div>
 
+  <!-- ── Add-content modal ─────────────────────────────────────── -->
+  <div class="modal" id="library-modal" hidden role="dialog" aria-modal="true"
+       aria-labelledby="library-modal-title">
+    <div class="modal-backdrop" id="library-modal-backdrop"></div>
+    <div class="modal-card" role="document">
+      <header class="modal-head">
+        <h2 id="library-modal-title">Add to the field library</h2>
+        <button class="icon-btn" id="library-modal-close" type="button" aria-label="Close">
+          <svg viewBox="0 0 24 24" aria-hidden="true"><path d="M6 6l12 12M18 6L6 18" stroke="currentColor" stroke-width="2" stroke-linecap="round" fill="none"/></svg>
+        </button>
+      </header>
+      <p class="modal-sub">Import curated public-domain collections, or upload your own
+        PDF/TXT documents. New files are indexed automatically — no restart needed.</p>
+
+      <section class="modal-section">
+        <h3 class="modal-h3">Curated collections</h3>
+        <div id="catalog-list" class="catalog-list">Loading…</div>
+      </section>
+
+      <section class="modal-section">
+        <h3 class="modal-h3">Upload your own</h3>
+        <form id="upload-form" class="upload-form">
+          <label class="upload-cat">Category
+            <select id="upload-category"></select>
+          </label>
+          <button type="button" class="upload-drop" id="upload-drop">
+            <input type="file" id="upload-file" accept=".pdf,.txt" hidden>
+            <span id="upload-drop-text">Choose a PDF or TXT file, or drop it here…</span>
+          </button>
+          <button type="submit" class="ghost-btn" id="upload-submit">Upload</button>
+        </form>
+        <p class="modal-status" id="upload-status" hidden></p>
+      </section>
+    </div>
+  </div>
+
   <script src="app.js"></script>
 </body>
 </html>
diff --git a/public/style.css b/public/style.css
index 700b308..276fbf6 100644
--- a/public/style.css
+++ b/public/style.css
@@ -765,3 +765,121 @@ body {
 
   .toggle span { display: none; }
 }
+
+/* ══════════════════════════════════════════════════════════════════════
+   Add-content: Field-library action + import/upload modal
+   ══════════════════════════════════════════════════════════════════════ */
+
+.panel-action {
+  margin-left: auto;
+  padding: 2px 8px;
+  font-size: 0.68rem;
+  letter-spacing: 0.04em;
+}
+
+.modal[hidden] { display: none; }
+.modal {
+  position: fixed;
+  inset: 0;
+  z-index: 50;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  padding: 20px;
+}
+.modal-backdrop {
+  position: absolute;
+  inset: 0;
+  background: rgba(2, 16, 17, 0.72);
+  backdrop-filter: blur(2px);
+}
+.modal-card {
+  position: relative;
+  width: min(640px, 100%);
+  max-height: 88vh;
+  overflow-y: auto;
+  background: var(--bg1);
+  border: 1px solid var(--line);
+  border-radius: var(--radius-lg);
+  box-shadow: var(--shadow-lg);
+  padding: 20px 22px 24px;
+}
+.modal-head {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 12px;
+}
+.modal-head h2 { margin: 0; font-size: 1.05rem; }
+.modal-sub {
+  margin: 8px 0 4px;
+  color: var(--fg-dim);
+  font-size: 0.85rem;
+  line-height: 1.5;
+}
+.modal-section { margin-top: 20px; }
+.modal-h3 {
+  margin: 0 0 10px;
+  font-family: var(--mono);
+  font-size: 0.7rem;
+  font-weight: 600;
+  letter-spacing: 0.14em;
+  text-transform: uppercase;
+  color: var(--fg-dim);
+}
+.modal-note   { margin: 10px 2px 0; font-size: 0.78rem; color: var(--warning); }
+.modal-status { margin: 10px 2px 0; font-size: 0.82rem; color: var(--fg-dim); }
+.modal-status.is-error { color: var(--err); }
+
+.catalog-list { display: flex; flex-direction: column; gap: 8px; }
+.catalog-row {
+  display: flex;
+  align-items: center;
+  gap: 12px;
+  background: var(--bg2);
+  border: 1px solid var(--line);
+  border-radius: var(--radius-md);
+  padding: 10px 12px;
+}
+.catalog-meta { display: flex; flex-direction: column; gap: 2px; flex: 1; min-width: 0; }
+.catalog-name  { font-weight: 600; font-size: 0.9rem; }
+.catalog-desc  { color: var(--fg-dim); font-size: 0.78rem; line-height: 1.4; }
+.catalog-count { color: var(--fg-faint); font-size: 0.72rem; font-family: var(--mono); }
+.catalog-install { flex-shrink: 0; }
+.catalog-install.is-done { color: var(--success); border-color: transparent; }
+
+.upload-form { display: grid; grid-template-columns: 1fr; gap: 10px; }
+.upload-cat {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+  font-size: 0.8rem;
+  color: var(--fg-dim);
+}
+.upload-cat select {
+  flex: 1;
+  background: var(--bg2);
+  color: var(--fg);
+  border: 1px solid var(--line);
+  border-radius: var(--radius-sm);
+  padding: 6px 8px;
+  font-family: inherit;
+  font-size: 0.82rem;
+}
+.upload-drop {
+  display: block;
+  width: 100%;
+  text-align: center;
+  padding: 18px;
+  background: var(--bg2);
+  border: 1px dashed var(--fg-faint);
+  border-radius: var(--radius-md);
+  color: var(--fg-dim);
+  font-size: 0.85rem;
+  cursor: pointer;
+}
+.upload-drop.dragover {
+  border-color: var(--accent);
+  background: var(--accent-soft);
+  color: var(--fg);
+}
diff --git a/sources.yaml b/sources.yaml
index b04762e..d68f72a 100644
--- a/sources.yaml
+++ b/sources.yaml
@@ -24,12 +24,43 @@
 # Validate after editing:  helper-scripts/fetch-source-data.sh --validate
 # ══════════════════════════════════════════════════════════════════════
 
+# ── Collections ───────────────────────────────────────────────────────
+# Named bundles a user can import as a unit (see GET /api/catalog and
+# `fetch-source-data.sh --collection <id>`). Every item below tags exactly
+# one collection via its `collection:` field.
+collections:
+  - id: core-survival
+    name: "Core Survival & Preparedness"
+    description: "Field survival, shelter, navigation, weather, and disaster-readiness checklists."
+  - id: austere-medicine
+    name: "Austere & Emergency Medicine"
+    description: "First aid, field surgery, dentistry, and health care where there is no doctor."
+  - id: food-and-water
+    name: "Food & Water"
+    description: "Growing, canning, and preserving food; safe water, gravity-flow systems, solar disinfection."
+  - id: power-and-engineering
+    name: "Power & Engineering"
+    description: "Off-grid solar power and appropriate-technology village engineering."
+  - id: comms-and-radio
+    name: "Communications & Radio"
+    description: "Amateur/emergency radio and interoperable field communications."
+  - id: education
+    name: "Education & Textbooks"
+    description: "Open textbooks and foundational learning material."
+  - id: civics-and-culture
+    name: "Civics & Culture"
+    description: "Founding civic documents and enduring public-domain texts."
+  - id: software-and-tech
+    name: "Software & Technical Reference"
+    description: "Programming, operating systems, and version control references."
+
 sources:
 
   # ── 100 Survival & Preparedness ───────────────────────────────────
   - url: https://www.ready.gov/sites/default/files/2021-02/ready_checklist.pdf
     filename: FEMA-Emergency-Checklist.pdf
     category: 100_Survival
+    collection: core-survival
     title: "FEMA — Emergency Supply Checklist"
     license: PD (US Government)
     size: ~150 KB
@@ -37,6 +68,7 @@ sources:
   - url: https://www.ready.gov/sites/default/files/2020-03/ready_emergency-financial-first-aid-toolkit.pdf
     filename: FEMA-Financial-First-Aid-Kit.pdf
     category: 100_Survival
+    collection: core-survival
     title: "FEMA — Emergency Financial First Aid Kit"
     license: PD (US Government)
     size: ~1.7 MB
@@ -44,6 +76,7 @@ sources:
   - url: https://archive.org/download/Fm21-76SurvivalManual/FM21-76_SurvivalManual.pdf
     filename: FM21-76-US-Army-Survival.pdf
     category: 100_Survival
+    collection: core-survival
     title: "US Army Survival Manual FM 21-76 (shelter, water, foraging, trapping)"
     license: PD (US Government)
     size: ~3 MB
@@ -51,6 +84,7 @@ sources:
   - url: https://archive.org/download/milmanual-fm-3-25.26-map-reading-and-land-navigation/fm_3-25.26_map_reading_and_land_navigation.pdf
     filename: FM3-25-26-Map-Reading-Land-Navigation.pdf
     category: 100_Survival
+    collection: core-survival
     title: "US Army FM 3-25.26 — Map Reading and Land Navigation"
     license: PD (US Government)
     size: ~24 MB
@@ -58,6 +92,7 @@ sources:
   - url: https://archive.org/download/pdfy-ULdTVDxCZR1mXh25/Nuclear%20War%20Survival%20Skills%20-%20Cresson%20H%20Kearney.pdf
     filename: NuclearWarSurvivalSkills-CressonHKearney.pdf
     category: 100_Survival
+    collection: core-survival
     title: "Nuclear War Survival Skills (Kearney, Oak Ridge National Laboratory)"
     license: PD (US Government research; author authorized free reproduction)
     size: ~6.4 MB
@@ -65,6 +100,7 @@ sources:
   - url: https://www.weather.gov/media/owlie/SGJune6-11.pdf
     filename: NWS-Weather-Spotters-Field-Guide.pdf
     category: 100_Survival
+    collection: core-survival
     title: "NOAA/NWS — Weather Spotter's Field Guide"
     license: PD (US Government)
     size: ~28 MB
@@ -72,21 +108,24 @@ sources:
   - url: https://www.gutenberg.org/cache/epub/17093/pg17093.txt
     filename: Camp-Life-In-The-Woods-Trapping.txt
     category: 100_Survival
+    collection: core-survival
     title: "Camp Life in the Woods and the Tricks of Trapping (Gibson, 1881)"
-    license: PD (Project Gutenberg #17093)
+    license: "PD (Project Gutenberg #17093)"
     size: ~500 KB
 
   - url: https://www.gutenberg.org/cache/epub/28255/pg28255.txt
     filename: Shelters-Shacks-And-Shanties.txt
     category: 100_Survival
+    collection: core-survival
     title: "Shelters, Shacks and Shanties (D.C. Beard, 1914)"
-    license: PD (Project Gutenberg #28255)
+    license: "PD (Project Gutenberg #28255)"
     size: ~400 KB
 
   # ── 200 Medical ───────────────────────────────────────────────────
   - url: https://archive.org/download/where-there-is-no-doctor/Where%20There%20Is%20No%20Doctor.pdf
     filename: Where-There-Is-No-Doctor.pdf
     category: 200_Medical
+    collection: austere-medicine
     title: "Where There Is No Doctor (Hesperian)"
     license: Hesperian open copyright (free non-commercial reproduction)
     size: ~8.7 MB
@@ -94,6 +133,7 @@ sources:
   - url: https://archive.org/download/where-there-is-no-dentist-hesperian-2010/Where_there_is_no_dentist_-_Hesperian_2010.pdf
     filename: Where-There-Is-No-Dentist.pdf
     category: 200_Medical
+    collection: austere-medicine
     title: "Where There Is No Dentist (Hesperian)"
     license: Hesperian open copyright (free non-commercial reproduction)
     size: ~6.8 MB
@@ -101,6 +141,7 @@ sources:
   - url: https://archive.org/download/WhereWomenHaveNoDoctor/18.HersperianFoundation-WhereWomenHaveNoDoctor.pdf
     filename: Where-Women-Have-No-Doctor.pdf
     category: 200_Medical
+    collection: austere-medicine
     title: "Where Women Have No Doctor (Hesperian)"
     license: Hesperian open copyright (free non-commercial reproduction)
     size: ~12 MB
@@ -108,6 +149,7 @@ sources:
   - url: https://archive.org/download/UsArmyFirstAidManualFm4-25.11/UsArmyFirstAidManual.pdf
     filename: FM4-25-11-First-Aid.pdf
     category: 200_Medical
+    collection: austere-medicine
     title: "US Army FM 4-25.11 — First Aid"
     license: PD (US Government)
     size: ~2.6 MB
@@ -115,6 +157,7 @@ sources:
   - url: https://archive.org/download/emergency-war-surgery-5th-edition/Emergency%20War%20Surgery%205th%20Edition.pdf
     filename: Emergency-War-Surgery-5th-Edition.pdf
     category: 200_Medical
+    collection: austere-medicine
     title: "Emergency War Surgery, 5th Edition (Borden Institute)"
     license: PD (US Government)
     size: ~30 MB
@@ -122,6 +165,7 @@ sources:
   - url: https://archive.org/download/survival-and-austere-medicine-3rd-ed-2017/Survival%20and%20Austere%20Medicine%203rd%20Ed%20%282017%29.pdf
     filename: Survival-And-Austere-Medicine-3rd-Ed.pdf
     category: 200_Medical
+    collection: austere-medicine
     title: "Survival and Austere Medicine: An Introduction (3rd ed., 2017)"
     license: Free distribution (authors permit non-commercial copying)
     size: ~23 MB
@@ -129,6 +173,7 @@ sources:
   - url: https://www.nctsn.org/sites/default/files/resources/pfa_field_operations_guide.pdf
     filename: PFA-Field-Operations-Guide.pdf
     category: 200_Medical
+    collection: austere-medicine
     title: "Psychological First Aid — Field Operations Guide (NCTSN/NCPTSD)"
     license: Free distribution (US Government funded)
     size: ~3.4 MB
@@ -137,6 +182,7 @@ sources:
   - url: https://archive.org/download/usda-guide-home-canning/USDA-Complete-Guide-to-Home-Canning-2015-revision.pdf
     filename: USDA-Complete-Guide-Home-Canning.pdf
     category: 300_Food
+    collection: food-and-water
     title: "USDA Complete Guide to Home Canning (2015 revision, all 7 guides)"
     license: PD (US Government)
     size: ~17 MB
@@ -144,14 +190,16 @@ sources:
   - url: https://www.gutenberg.org/cache/epub/9550/pg9550.txt
     filename: Manual-Of-Gardening-Bailey.txt
     category: 300_Food
+    collection: food-and-water
     title: "Manual of Gardening, 2nd ed. (L.H. Bailey)"
-    license: PD (Project Gutenberg #9550)
+    license: "PD (Project Gutenberg #9550)"
     size: ~900 KB
 
   # ── 400 Engineering, Water & Power ────────────────────────────────
   - url: https://www.nrel.gov/docs/fy04osti/36178.pdf
     filename: NREL-PV-Solar-Systems.pdf
     category: 400_Engineering
+    collection: power-and-engineering
     title: "NREL — Photovoltaic Solar Resource of the United States"
     license: PD (US Government)
     size: ~1 MB
@@ -159,6 +207,7 @@ sources:
   - url: https://archive.org/download/VillageTechnologyHandbook/Village%20Technology%20Handbook.pdf
     filename: VITA-Village-Technology-Handbook.pdf
     category: 400_Engineering
+    collection: power-and-engineering
     title: "VITA — Village Technology Handbook (water, agriculture, construction)"
     license: PD / freely distributed (USAID-funded)
     size: ~15 MB
@@ -166,6 +215,7 @@ sources:
   - url: https://archive.org/download/fa_Handbook_of_Gravity-Flow_Water_Systems/Handbook_of_Gravity-Flow_Water_Systems.pdf
     filename: Handbook-Of-Gravity-Flow-Water-Systems.pdf
     category: 400_Engineering
+    collection: food-and-water
     title: "A Handbook of Gravity-Flow Water Systems (Jordan)"
     license: Freely distributed (Appropriate Technology Library)
     size: ~18 MB
@@ -173,6 +223,7 @@ sources:
   - url: https://archive.org/download/fa_Solar_Disinfection_of_Water/Solar_Disinfection_of_Water.pdf
     filename: Solar-Disinfection-Of-Water.pdf
     category: 400_Engineering
+    collection: food-and-water
     title: "Solar Disinfection of Water (SODIS)"
     license: Freely distributed (Appropriate Technology Library)
     size: ~13 MB
@@ -181,6 +232,7 @@ sources:
   - url: https://www.arrl.org/files/file/Technology/tis/info/pdf/0209038.pdf
     filename: ARRL-Emergency-Comms.pdf
     category: 500_Comms
+    collection: comms-and-radio
     title: "ARRL — Emergency Communications"
     license: Freely distributed (ARRL)
     size: ~540 KB
@@ -188,6 +240,7 @@ sources:
   - url: https://archive.org/download/national-interoperability-field-operations-guide-nifog-version-2.02/National%20Interoperability%20Field%20Operations%20Guide%20%28NIFOG%29%20Version%202.02.pdf
     filename: CISA-NIFOG-2.02.pdf
     category: 500_Comms
+    collection: comms-and-radio
     title: "CISA — National Interoperability Field Operations Guide (NIFOG) v2.02"
     license: PD (US Government)
     size: ~6.7 MB
@@ -196,6 +249,7 @@ sources:
   - url: https://assets.openstax.org/oscms-prodcms/media/documents/ElementaryAlgebra2e-WEB.pdf
     filename: OpenStax-Elementary-Algebra-2e.pdf
     category: 600_Education
+    collection: education
     title: "OpenStax — Elementary Algebra 2e"
     license: CC BY 4.0
     size: ~45 MB
@@ -204,6 +258,7 @@ sources:
   - url: https://www.un.org/sites/un2.un.org/files/2021/03/udhr.pdf
     filename: UN-Universal-Declaration-Human-Rights.pdf
     category: 700_Social
+    collection: civics-and-culture
     title: "Universal Declaration of Human Rights"
     license: PD (United Nations)
     size: ~100 KB
@@ -211,6 +266,7 @@ sources:
   - url: https://www.govinfo.gov/content/pkg/CDOC-110hdoc50/pdf/CDOC-110hdoc50.pdf
     filename: US-Constitution.pdf
     category: 700_Social
+    collection: civics-and-culture
     title: "The Constitution of the United States (pocket edition)"
     license: PD (US Government)
     size: ~4 MB
@@ -218,21 +274,24 @@ sources:
   - url: https://www.gutenberg.org/cache/epub/1404/pg1404.txt
     filename: Federalist-Papers.txt
     category: 700_Social
+    collection: civics-and-culture
     title: "The Federalist Papers"
-    license: PD (Project Gutenberg #1404)
+    license: "PD (Project Gutenberg #1404)"
     size: ~1.2 MB
 
   - url: https://www.gutenberg.org/cache/epub/10/pg10.txt
     filename: KJV-Bible.txt
     category: 700_Social
+    collection: civics-and-culture
     title: "King James Bible"
-    license: PD (Project Gutenberg #10)
+    license: "PD (Project Gutenberg #10)"
     size: ~4.4 MB
 
   # ── 800 Software / Technical Reference ────────────────────────────
   - url: https://greenteapress.com/thinkpython2/thinkpython2.pdf
     filename: ThinkPython2.pdf
     category: 800_Software
+    collection: software-and-tech
     title: "Think Python, 2nd Edition (Downey)"
     license: CC BY-NC
     size: ~920 KB
@@ -240,6 +299,7 @@ sources:
   - url: https://greenteapress.com/thinkos/thinkos.pdf
     filename: ThinkOS.pdf
     category: 800_Software
+    collection: software-and-tech
     title: "Think OS: A Brief Introduction to Operating Systems (Downey)"
     license: CC BY-NC
     size: ~370 KB
@@ -247,6 +307,7 @@ sources:
   - url: https://eloquentjavascript.net/Eloquent_JavaScript.pdf
     filename: Eloquent-JavaScript.pdf
     category: 800_Software
+    collection: software-and-tech
     title: "Eloquent JavaScript, 3rd Edition (Haverbeke)"
     license: CC BY-NC
     size: ~2 MB
@@ -254,6 +315,7 @@ sources:
   - url: https://github.com/progit/progit2/releases/download/2.1.426/progit.pdf
     filename: Pro-Git-2.pdf
     category: 800_Software
+    collection: software-and-tech
     title: "Pro Git, 2nd Edition (Chacon & Straub)"
     license: CC BY-NC-SA
     size: ~19 MB
diff --git a/src/catalog.h b/src/catalog.h
new file mode 100644
index 0000000..6ea7298
--- /dev/null
+++ b/src/catalog.h
@@ -0,0 +1,115 @@
+#pragma once
+
+// ── Catalog parser ───────────────────────────────────────────────────
+// Reads the curated source manifest (sources.yaml) into memory. This is a
+// deliberately minimal line parser for the flat schema that manifest uses —
+// NOT a general YAML parser — and mirrors the awk parser in
+// helper-scripts/fetch-source-data.sh so the two never diverge. The manifest
+// is a trusted, image-baked file, so the parser optimises for clarity.
+
+#include <string>
+#include <vector>
+#include <fstream>
+#include "config.h"
+
+struct CatalogItem {
+    std::string url, filename, category, collection, title, license, size, sha256;
+};
+
+struct CatalogCollection {
+    std::string id, name, description;
+};
+
+struct Catalog {
+    std::vector<CatalogCollection> collections;
+    std::vector<CatalogItem>       items;
+
+    const CatalogCollection* find_collection(const std::string& id) const {
+        for (const auto& c : collections) if (c.id == id) return &c;
+        return nullptr;
+    }
+};
+
+// Clean a raw value that follows the first ':' on a line. A double-quoted
+// value is returned verbatim (so a legitimate '#' inside it — e.g. a license
+// like "PD (Project Gutenberg #17093)" — is preserved). An unquoted value has
+// a trailing " # comment" stripped and is trimmed.
+inline std::string catalog_clean_value(std::string v) {
+    size_t b = v.find_first_not_of(" \t");
+    if (b == std::string::npos) return "";
+    v = v.substr(b);
+
+    if (v.front() == '"') {
+        size_t end = v.find('"', 1);
+        if (end != std::string::npos) return v.substr(1, end - 1);
+        // Unterminated quote — fall through to best-effort unquoted handling.
+    }
+
+    size_t h = v.find(" #");
+    if (h != std::string::npos) v = v.substr(0, h);
+    size_t e = v.find_last_not_of(" \t\r");
+    return (e == std::string::npos) ? "" : v.substr(0, e + 1);
+}
+
+inline Catalog parse_catalog(const std::string& path) {
+    Catalog cat;
+    std::ifstream in(path);
+    if (!in) return cat;
+
+    enum Section { NONE, COLLECTIONS, SOURCES } sec = NONE;
+    CatalogItem item;       bool have_item = false;
+    CatalogCollection col;  bool have_col  = false;
+
+    auto flush_item = [&] {
+        if (have_item) cat.items.push_back(item);
+        item = CatalogItem(); have_item = false;
+    };
+    auto flush_col = [&] {
+        if (have_col) cat.collections.push_back(col);
+        col = CatalogCollection(); have_col = false;
+    };
+
+    std::string line;
+    while (std::getline(in, line)) {
+        if (!line.empty() && line.back() == '\r') line.pop_back();
+        size_t b = line.find_first_not_of(" \t");
+        if (b == std::string::npos) continue;             // blank
+        std::string t = line.substr(b);
+        if (t[0] == '#') continue;                        // comment
+
+        // Top-level keys (no indentation) switch sections.
+        if (b == 0) {
+            if (t.rfind("collections:", 0) == 0) { flush_col(); flush_item(); sec = COLLECTIONS; }
+            else if (t.rfind("sources:", 0) == 0) { flush_col(); flush_item(); sec = SOURCES; }
+            else { flush_col(); flush_item(); sec = NONE; }
+            continue;
+        }
+
+        // Indented "key: value" (optionally list-prefixed with "- ").
+        bool dash = false;
+        std::string s = t;
+        if (s.rfind("- ", 0) == 0) { dash = true; s = s.substr(2); }
+        size_t colon = s.find(':');
+        if (colon == std::string::npos) continue;
+        std::string key = s.substr(0, colon);
+        std::string val = catalog_clean_value(s.substr(colon + 1));
+
+        if (sec == COLLECTIONS) {
+            if (dash && key == "id")        { flush_col(); col.id = val; have_col = true; }
+            else if (key == "name")         col.name        = val;
+            else if (key == "description")  col.description = val;
+        } else if (sec == SOURCES) {
+            if (dash && key == "url")       { flush_item(); item.url = val; have_item = true; }
+            else if (key == "filename")     item.filename   = val;
+            else if (key == "category")     item.category   = val;
+            else if (key == "collection")   item.collection = val;
+            else if (key == "title")        item.title      = val;
+            else if (key == "license")      item.license    = val;
+            else if (key == "size")         item.size       = val;
+            else if (key == "sha256")       item.sha256     = val;
+        }
+    }
+    flush_col();
+    flush_item();
+    return cat;
+}
diff --git a/src/config.h b/src/config.h
index 9d5fd5e..0ded8e4 100644
--- a/src/config.h
+++ b/src/config.h
@@ -34,6 +34,14 @@ const size_t MAX_QUERY_CHARS   = 8000;            // longest accepted question
 const size_t MAX_CONV_ID_CHARS = 128;
 const size_t MAX_CONVERSATIONS = 200;             // bounded history map
 
+// Largest single file accepted by POST /api/upload. cpp-httplib's payload
+// limit is global, so this also caps how much ANY request can buffer — kept
+// modest to bound that exposure. Large curated documents are pulled by the
+// server-side importer (curl), not through an upload body, so this only needs
+// to fit a user's own document; the /query and /api/import handlers separately
+// reject bodies over MAX_REQUEST_BODY.
+const size_t MAX_UPLOAD_BYTES  = 32u * 1024u * 1024u; // 32 MB per uploaded file
+
 // ── Ingestion ────────────────────────────────────────────────────────
 const size_t MAX_DOCUMENT_CHARS = 8u * 1000u * 1000u; // per-document text cap
 const int    FILE_SETTLE_SECONDS = 10; // skip files modified more recently
@@ -167,3 +175,22 @@ inline std::string get_llm_model_name() {
 inline std::string get_embedding_model_name() {
     return env_or("EMBEDDING_MODEL", "nomic-embed-text");
 }
+
+// ── Content import (catalog + upload) ─────────────────────────────────
+// The curated manifest the /api/catalog + /api/import endpoints read, and the
+// fetcher script they exec to download vetted catalog items. Both are baked
+// into the image (see Dockerfile); overridable for local runs and tests.
+inline std::string get_catalog_path() {
+    return env_or("JIC_CATALOG_PATH", "sources.yaml");
+}
+
+inline std::string get_fetcher_path() {
+    return env_or("JIC_FETCHER_PATH", "/app/bin/fetch-sources");
+}
+
+// Catalog import reaches out to the network. Allow locking it down for
+// air-gapped deployments (uploading local files stays available regardless).
+inline bool network_import_enabled() {
+    const char* v = getenv("JIC_DISABLE_NETWORK_IMPORT");
+    return !(v && *v);
+}
diff --git a/src/server.cpp b/src/server.cpp
index 8eef77e..950a103 100644
--- a/src/server.cpp
+++ b/src/server.cpp
@@ -18,6 +18,10 @@
 #include <atomic>
 #include <csignal>
 #include <cctype>
+#include <cerrno>
+#include <spawn.h>
+#include <sys/wait.h>
+#include <unistd.h>
 
 #include "httplib.h"
 #include "llama.h"
@@ -28,6 +32,7 @@
 #include "embeddings.h"
 #include "llm.h"
 #include "sqlite_vec_index.h"
+#include "catalog.h"
 
 namespace fs = std::filesystem;
 using json = nlohmann::json;
@@ -127,6 +132,10 @@ static void handle_query(const httplib::Request& req, httplib::Response& res) {
     }
 
     // ── Parse & validate input (client errors → 400, not 500) ───────
+    if (req.body.size() > MAX_REQUEST_BODY) {   // /query is small JSON, not an upload
+        send_error(res, 413, "Request body too large");
+        return;
+    }
     json rj;
     try {
         rj = json::parse(req.body);
@@ -326,6 +335,287 @@ static void handle_library(const httplib::Request&, httplib::Response& res) {
     res.set_content(out.dump(), "application/json");
 }
 
+// ═════════════════════════════════════════════════════════════════════
+// Content import & upload
+//
+// Two ways to grow the library at runtime, both writing into the sources
+// volume where the ingestion service auto-indexes new files:
+//   • POST /api/import — download a vetted catalog collection. The server
+//     execs the bundled fetcher, which only ever contacts the fixed URLs in
+//     the image-baked manifest, so there is no arbitrary-URL / SSRF surface.
+//   • POST /api/upload — accept a user's own PDF/TXT (works fully offline).
+// GET /api/catalog lists what is available and what is already installed.
+// ═════════════════════════════════════════════════════════════════════
+
+extern char** environ;
+
+// One import at a time — a collection download can take minutes; this guards
+// against a client spawning many concurrent fetchers.
+static std::atomic<bool> g_import_running{false};
+
+static bool str_ends_with(const std::string& s, const std::string& suf) {
+    return s.size() >= suf.size() &&
+           s.compare(s.size() - suf.size(), suf.size(), suf) == 0;
+}
+
+// Reduce a browser-supplied filename to a safe basename: no path components,
+// no traversal, a conservative character set. Returns "" if nothing usable.
+static std::string sanitize_upload_filename(const std::string& raw) {
+    std::string name = fs::path(raw).filename().string();   // strip any path
+    std::string out;
+    for (char c : name) {
+        unsigned char u = static_cast<unsigned char>(c);
+        if (std::isalnum(u) || c == '.' || c == '_' || c == '-' ||
+            c == ' ' || c == '(' || c == ')')
+            out += c;
+    }
+    while (!out.empty() && (out.front() == ' ' || out.front() == '.'))
+        out.erase(out.begin());
+    while (!out.empty() && out.back() == ' ')
+        out.pop_back();
+    if (out.empty() || out.size() > 200) return "";
+    if (out.find("..") != std::string::npos) return "";
+    return out;
+}
+
+// Category must match the taxonomy pattern NNN_Alpha (e.g. 200_Medical); this
+// alone forbids path separators and traversal.
+static bool valid_upload_category(const std::string& c) {
+    if (c.size() < 5 || c.size() > 40) return false;
+    if (!std::isdigit((unsigned char)c[0]) || !std::isdigit((unsigned char)c[1]) ||
+        !std::isdigit((unsigned char)c[2]) || c[3] != '_') return false;
+    for (size_t i = 4; i < c.size(); i++)
+        if (!std::isalpha((unsigned char)c[i])) return false;
+    return true;
+}
+
+// Download one catalog collection by exec-ing the bundled fetcher. Uses
+// posix_spawn with an explicit argv (never a shell), so the collection id —
+// already validated against the catalog — cannot be interpreted as a command.
+// Runs on a detached worker thread; clears the single-flight flag when done.
+static void run_import_collection(std::string collection) {
+    std::string fetcher  = get_fetcher_path();
+    std::string manifest = get_catalog_path();
+    std::string dest     = get_sources_dir();
+
+    std::vector<std::string> args = {
+        fetcher, "--manifest", manifest, "--dest", dest,
+        "--collection", collection
+    };
+    std::vector<char*> argv;
+    for (auto& a : args) argv.push_back(const_cast<char*>(a.c_str()));
+    argv.push_back(nullptr);
+
+    std::cout << "Import: fetching collection '" << collection << "'" << std::endl;
+    pid_t pid = 0;
+    int rc = posix_spawn(&pid, fetcher.c_str(), nullptr, nullptr,
+                         argv.data(), environ);
+    if (rc == 0) {
+        int status = 0;
+        waitpid(pid, &status, 0);
+        std::cout << "Import: collection '" << collection << "' done (exit "
+                  << (WIFEXITED(status) ? WEXITSTATUS(status) : -1) << ")" << std::endl;
+    } else {
+        std::cerr << "Import: failed to spawn fetcher (" << get_fetcher_path()
+                  << "): " << std::strerror(rc) << std::endl;
+    }
+    g_import_running.store(false);
+}
+
+// GET /api/catalog — the importable catalog, grouped by collection, with an
+// "installed" flag per item (file already present in the sources volume).
+static void handle_catalog(const httplib::Request&, httplib::Response& res) {
+    Catalog cat = parse_catalog(get_catalog_path());
+    const fs::path sources_root = get_sources_dir();
+
+    json cols = json::array();
+    for (const auto& c : cat.collections) {
+        json items = json::array();
+        int installed = 0;
+        for (const auto& it : cat.items) {
+            if (it.collection != c.id) continue;
+            std::error_code ec;
+            bool present = fs::exists(sources_root / it.category / it.filename, ec) && !ec;
+            if (present) installed++;
+            items.push_back({
+                {"filename",  it.filename},
+                {"category",  it.category},
+                {"title",     it.title},
+                {"license",   it.license},
+                {"size",      it.size},
+                {"installed", present}
+            });
+        }
+        cols.push_back({
+            {"id",          c.id},
+            {"name",        c.name},
+            {"description", c.description},
+            {"count",       items.size()},
+            {"installed",   installed},
+            {"items",       items}
+        });
+    }
+    json out;
+    out["collections"]    = cols;
+    out["import_enabled"] = network_import_enabled();
+    out["importing"]      = g_import_running.load();
+    // error_handler_t::replace: never throw out of the handler if a manifest
+    // string contains an invalid UTF-8 byte — emit a replacement char instead.
+    res.set_content(out.dump(-1, ' ', false, json::error_handler_t::replace),
+                    "application/json");
+}
+
+// POST /api/import  { "collection": "<id>" } — start downloading a collection.
+static void handle_import(const httplib::Request& req, httplib::Response& res) {
+    if (!network_import_enabled()) {
+        send_error(res, 403, "Catalog import over the network is disabled on this deployment.");
+        return;
+    }
+    if (req.body.size() > MAX_REQUEST_BODY) {   // this endpoint needs only tiny JSON
+        send_error(res, 413, "Request body too large");
+        return;
+    }
+    json rj;
+    try { rj = json::parse(req.body); }
+    catch (const std::exception&) { send_error(res, 400, "Request body must be valid JSON"); return; }
+
+    if (!rj.contains("collection") || !rj["collection"].is_string()) {
+        send_error(res, 400, "Missing required string field: collection");
+        return;
+    }
+    std::string collection = rj["collection"];
+
+    // Validate against the catalog (a closed set) AND as a strict slug —
+    // defence in depth, even though posix_spawn never involves a shell.
+    Catalog cat = parse_catalog(get_catalog_path());
+    if (!cat.find_collection(collection)) {
+        send_error(res, 404, "Unknown collection id");
+        return;
+    }
+    for (char c : collection)
+        if (!(std::islower((unsigned char)c) || std::isdigit((unsigned char)c) || c == '-')) {
+            send_error(res, 400, "Invalid collection id");
+            return;
+        }
+
+    bool expected = false;
+    if (!g_import_running.compare_exchange_strong(expected, true)) {
+        send_error(res, 409, "An import is already in progress.");
+        return;
+    }
+    try {
+        std::thread(run_import_collection, collection).detach();
+    } catch (const std::exception& e) {
+        g_import_running.store(false);   // never leave the flag wedged
+        std::cerr << "Import: could not start worker: " << e.what() << std::endl;
+        send_error(res, 503, "Could not start the import worker; please retry.");
+        return;
+    }
+
+    res.status = 202;
+    json out;
+    out["status"]     = "started";
+    out["collection"] = collection;
+    res.set_content(out.dump(), "application/json");
+}
+
+// POST /api/upload  (multipart: file=<pdf|txt>, category=NNN_Name)
+static void handle_upload(const httplib::Request& req, httplib::Response& res) {
+    if (!req.is_multipart_form_data()) {
+        send_error(res, 400, "Expected multipart/form-data");
+        return;
+    }
+    if (!req.has_file("file")) {   // clean 400 rather than risk a throw on a missing field
+        send_error(res, 400, "Missing form field: file");
+        return;
+    }
+    auto file = req.get_file_value("file");
+    std::string category = req.has_file("category")
+                         ? req.get_file_value("category").content
+                         : (req.has_param("category") ? req.get_param_value("category") : "");
+
+    if (!valid_upload_category(category)) {
+        send_error(res, 400, "Invalid category (expected NNN_Name, e.g. 200_Medical)");
+        return;
+    }
+    std::string fn = sanitize_upload_filename(file.filename);
+    if (fn.empty()) { send_error(res, 400, "Invalid or missing filename"); return; }
+
+    // Validate the extension case-insensitively (ingestion lowercases it, so
+    // e.g. FOO.PDF is a valid PDF and must pass the same gate here).
+    std::string ext_lc = fn;
+    for (char& c : ext_lc) c = static_cast<char>(std::tolower((unsigned char)c));
+    const bool is_pdf = str_ends_with(ext_lc, ".pdf");
+    const bool is_txt = str_ends_with(ext_lc, ".txt");
+    if (!is_pdf && !is_txt) {
+        send_error(res, 400, "Only .pdf or .txt files can be indexed");
+        return;
+    }
+    if (file.content.empty()) { send_error(res, 400, "Empty file"); return; }
+    if (file.content.size() > MAX_UPLOAD_BYTES) {
+        send_error(res, 413, "File exceeds the size limit");
+        return;
+    }
+    if (is_pdf && file.content.compare(0, 4, "%PDF") != 0) {
+        send_error(res, 400, "File does not look like a PDF");
+        return;
+    }
+
+    std::error_code ec;
+    fs::path dir = fs::path(get_sources_dir()) / category;
+    fs::create_directories(dir, ec);
+    if (ec) { send_error(res, 500, "Could not create the category directory"); return; }
+
+    // Refuse when the volume is nearly full, so uploads can't fill the disk and
+    // wedge the search index / DB. Fails CLOSED: if free space can't even be
+    // queried, reject rather than write blindly.
+    std::error_code space_ec;
+    auto space = fs::space(dir, space_ec);
+    if (space_ec || space.available < file.content.size() + (64ull << 20)) {
+        send_error(res, 507, "Not enough free space in the library volume");
+        return;
+    }
+
+    fs::path target = dir / fn;
+    if (fs::exists(target, ec)) {
+        send_error(res, 409, "A file with that name already exists in this category");
+        return;
+    }
+    // Unique temp name so two concurrent uploads never write the same .part
+    // (the ".part" suffix is not an ingestible extension, so a transient temp
+    // is never indexed).
+    static std::atomic<uint64_t> upload_seq{0};
+    fs::path part = dir / (fn + "." + std::to_string(upload_seq.fetch_add(1)) + ".part");
+    {
+        std::ofstream o(part, std::ios::binary | std::ios::trunc);
+        if (!o) { send_error(res, 500, "Could not write the file"); return; }
+        o.write(file.content.data(), static_cast<std::streamsize>(file.content.size()));
+        if (!o.good()) { o.close(); fs::remove(part, ec); send_error(res, 500, "Write failed"); return; }
+    }
+    // Finalise atomically WITHOUT clobbering: link() fails with EEXIST if the
+    // target appeared meanwhile, closing the check-then-rename race. The
+    // ingester never sees the .part (not an ingestible extension).
+    if (link(part.c_str(), target.c_str()) != 0) {
+        int e = errno;
+        fs::remove(part, ec);
+        if (e == EEXIST) {
+            send_error(res, 409, "A file with that name already exists in this category");
+        } else {
+            send_error(res, 500, "Could not finalise the file");
+        }
+        return;
+    }
+    fs::remove(part, ec);   // drop the temporary hard link; target remains
+
+    std::cout << "Upload: stored " << category << "/" << fn
+              << " (" << file.content.size() << " bytes)" << std::endl;
+    json out;
+    out["status"]   = "ok";
+    out["category"] = category;
+    out["filename"] = fn;
+    res.set_content(out.dump(), "application/json");
+}
+
 // ═════════════════════════════════════════════════════════════════════
 // Background thread: refresh cached counts every 10 s
 // ═════════════════════════════════════════════════════════════════════
@@ -503,8 +793,10 @@ int main() {
     httplib::Server svr;
     g_server = &svr;
 
-    // Reject oversized request bodies before buffering them
-    svr.set_payload_max_length(MAX_REQUEST_BODY);
+    // Body cap. Raised to MAX_UPLOAD_BYTES so /api/upload can accept real
+    // documents; cpp-httplib's limit is global, and on this single-user
+    // appliance that upper bound on buffered request size is acceptable.
+    svr.set_payload_max_length(MAX_UPLOAD_BYTES);
     svr.set_read_timeout(15, 0);
     svr.set_write_timeout(60, 0);
 
@@ -546,6 +838,9 @@ int main() {
     svr.Post("/query",       handle_query);
     svr.Get ("/status",      handle_status);
     svr.Get ("/api/library", handle_library);
+    svr.Get ("/api/catalog", handle_catalog);
+    svr.Post("/api/import",  handle_import);
+    svr.Post("/api/upload",  handle_upload);
     svr.Options(".*", [&cors_origin](const httplib::Request&, httplib::Response& res) {
         res.status = cors_origin.empty() ? 405 : 204;
     });