fix(crm-starter-export): scope is the PIPELINE, not the roster — operator correction applied

The export's default flips to one row per pipeline row (96, tracker
order — multi-deal orgs like Accelerate the Future keep their three
rows), enrichment joined where a canonical org matched, blanks where
not. The event-based org creations stay in the DB as kept noise; the
all-roster shape survives behind --scope all.

People scope accordingly: --orgs-csv reads the Phase-1 file's
external_id column, exports only persons with an edge into the included
set, and guarantees the primary attach is an included org.

Second run: 96 org rows (67 matched: 29 exact + 38 fuzzy; 29 unmatched
to the sidecar = the to-capture list), 61 people across 64 distinct
pipeline orgs. Plan open decisions 2 and 3 recorded as settled.

Files changed:
- scripts/export-crm-orgs-csv.mjs
- scripts/export-crm-people-csv.mjs
- context-v/plans/CRM-Starter-Export-Orgs-Then-People.md

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RW28dw3kQAKXr2ZNefCukE

Author: mpstaton

context-v/plans/CRM-Starter-Export-Orgs-Then-People.md

@@ -201,12 +201,15 @@ external_id has somewhere to land.
    **opportunity** records per company (Twenty's native pipeline
    object)? Recommendation: opportunities — that's what they are — but
    the operator rules at dry-run review.
-2. **Org scope confirmation** — recommendation above is all-364 +
-   filter-in-sheet; alternative is pre-filtering to `funders/` bucket
-   (+ pipeline matches) if the CRM should only ever see funders.
-3. **Person floor** — export all 417, or only persons with a rated
-   edge (`relevance` set)? Recommendation: all, with `relevance` as a
-   sheet-filterable column.
+2. ~~Org scope~~ — SETTLED (operator ruling 2026-07-27, overriding the
+   plan's all-364 recommendation): **pipeline rows only**. The export is
+   one row per PIPELINE row (96 — multi-deal orgs stay multi-row, exactly
+   like the tracker), enrichment joined where a canonical org matched.
+   Event-based org creations are kept in the DB but are noise for the
+   CRM; the all-roster shape survives behind `--scope all`.
+3. ~~Person floor~~ — SETTLED by the same ruling: people scope to the
+   imported orgs (`--orgs-csv` filter; primary attach always an included
+   org). The all-persons export remains available by omitting the flag.
 
 ## Out of scope
 

scripts/export-crm-orgs-csv.mjs

@@ -37,12 +37,16 @@ const { Surreal } = requireScripts('surrealdb');
 const { connect } = requireServices('@nats-io/transport-node');
 
 // ---- args -------------------------------------------------------------------
-const args = { client: 'reach-edu', match: 'Master-Pipeline-Tracker' };
+const args = { client: 'reach-edu', match: 'Master-Pipeline-Tracker', scope: 'pipeline' };
 for (let i = 2; i < process.argv.length; i += 1) {
   const k = process.argv[i];
   if (k === '--client') args.client = process.argv[++i];
   else if (k === '--record-set-name-match') args.match = process.argv[++i];
   else if (k === '--out-dir') args.outDir = process.argv[++i];
+  // 'pipeline' (default, operator ruling 2026-07-27): one row per PIPELINE
+  // row — the tracker's shape, enriched. Multi-deal orgs stay multi-row.
+  // 'all': one row per canonical org (the event-based long tail included).
+  else if (k === '--scope') args.scope = process.argv[++i];
 }
 const today = new Date().toISOString().slice(0, 10);
 const OUT_DIR = resolve(args.outDir ?? `clients/${args.client}/outputs/${today}_crm-starter`);
@@ -161,47 +165,44 @@ console.log(`pipeline rows: ${pipelineRows.length}`);
 
 // 6. Join pipeline → orgs: exact by corpus_funder_slug, then name/alias.
 const orgBySlug = new Map(orgs.map((o) => [o.slug, o]));
-const byNorm = new Map(); // normalized name/alias → slug (first wins; collisions logged)
+const byNorm = new Map(); // normalized name/alias → slug (first wins)
 for (const o of orgs) {
   for (const cand of [o.complete_name, o.conventional_name, o.slug.replace(/-/g, ' '), ...(o.aliases ?? [])]) {
     const n = normName(cand);
     if (n && !byNorm.has(n)) byNorm.set(n, o.slug);
   }
 }
-const pipelineByOrg = new Map(); // slug → { row, matched }
-const unmatched = [];
-for (const row of pipelineRows) {
+// Per-PIPELINE-row resolution — multiple rows may share one org (multi-deal
+// orgs like Accelerate the Future); each keeps its own row, tracker-style.
+const resolved = pipelineRows.map((row) => {
   const name = row['Prospect / Organization'] ?? '';
   const exact = row.corpus_funder_slug && orgBySlug.has(row.corpus_funder_slug) ? row.corpus_funder_slug : null;
   const fuzzy = exact ? null : byNorm.get(normName(name));
-  const slug = exact ?? fuzzy;
-  if (!slug) {
-    unmatched.push(row);
-    continue;
-  }
-  if (!pipelineByOrg.has(slug)) {
-    pipelineByOrg.set(slug, { row, matched: exact ? 'exact' : 'fuzzy' });
-  } else {
-    console.warn(`  ⚠ second pipeline row also matches ${slug}: "${name}" — kept the first, this one goes to the sidecar`);
-    unmatched.push(row);
-  }
-}
-console.log(`pipeline matched: ${pipelineByOrg.size} (exact+fuzzy) · unmatched: ${unmatched.length}`);
+  return { row, slug: exact ?? fuzzy ?? null, matched: exact ? 'exact' : fuzzy ? 'fuzzy' : 'none' };
+});
+const matchedCount = resolved.filter((r) => r.slug).length;
+console.log(`pipeline rows matched to a canonical org: ${matchedCount}/${resolved.length}`);
 
-// 7. Shape org rows.
+// 7. Shape rows — canonical-enrichment column block for one org.
 const exported_at = new Date().toISOString();
-const outRows = orgs.map((o) => {
-  const links = o.org_links ?? [];
+const enrichmentFor = (o) => {
+  if (!o) {
+    return {
+      external_id: '', name: '', conventional_name: '', aliases: '', domains: '',
+      bucket: '', tags: '',
+      ...Object.fromEntries(Object.values(PROMOTED_KINDS).map((c) => [c, ''])),
+      other_links: '', streams: '', stream_count: '', related_orgs: '',
+    };
+  }
   const flat = {};
   const other = [];
-  for (const l of links) {
+  for (const l of o.org_links ?? []) {
     const col = PROMOTED_KINDS[l.kind];
     if (col && !flat[col]) flat[col] = l.url;
     else other.push(`${l.kind ?? 'other'}: ${l.url}`);
   }
   const streams = (o.media_streams ?? []).map((s) => `${s.name ? s.name + ' — ' : ''}${s.url}${s.kind ? ` (${s.kind})` : ''}`);
-  const p = pipelineByOrg.get(o.slug);
-  const row = {
+  return {
     external_id: o.slug,
     name: o.complete_name ?? o.conventional_name ?? o.slug,
     conventional_name: o.conventional_name ?? '',
@@ -214,17 +215,44 @@ const outRows = orgs.map((o) => {
     streams: streams.join('\n'),
     stream_count: streams.length,
     related_orgs: (relsByOrg.get(o.slug) ?? []).join('\n'),
-    pipeline_org_name: p?.row['Prospect / Organization'] ?? '',
-    pipeline_matched: p?.matched ?? 'none',
-    ...Object.fromEntries(PIPELINE_COLS.map((c) => [c, p?.row[c] ?? ''])),
-    exported_at,
   };
-  return row;
-});
+};
 
-// Pipeline-matched rows first (they're the review priority), then by name.
-outRows.sort((a, b) =>
-  (a.pipeline_matched === 'none') - (b.pipeline_matched === 'none') || a.name.localeCompare(b.name));
+let outRows;
+let unmatched = [];
+if (args.scope === 'pipeline') {
+  // The tracker's shape: one row per pipeline row, in tracker order,
+  // enrichment blank where no canonical org matched. Rows with no match
+  // ALSO land in the sidecar as the to-capture list.
+  outRows = resolved.map(({ row, slug, matched }) => ({
+    ...enrichmentFor(slug ? orgBySlug.get(slug) : null),
+    pipeline_org_name: row['Prospect / Organization'] ?? '',
+    pipeline_matched: matched,
+    ...Object.fromEntries(PIPELINE_COLS.map((c) => [c, row[c] ?? ''])),
+    exported_at,
+  }));
+  unmatched = resolved.filter((r) => !r.slug).map((r) => r.row);
+} else {
+  // --scope all: one row per canonical org (first matching pipeline row
+  // attached), the event-based long tail included.
+  const pipelineByOrg = new Map();
+  for (const r of resolved) {
+    if (r.slug && !pipelineByOrg.has(r.slug)) pipelineByOrg.set(r.slug, r);
+  }
+  outRows = orgs.map((o) => {
+    const p = pipelineByOrg.get(o.slug);
+    return {
+      ...enrichmentFor(o),
+      pipeline_org_name: p?.row['Prospect / Organization'] ?? '',
+      pipeline_matched: p?.matched ?? 'none',
+      ...Object.fromEntries(PIPELINE_COLS.map((c) => [c, p?.row[c] ?? ''])),
+      exported_at,
+    };
+  });
+  outRows.sort((a, b) =>
+    (a.pipeline_matched === 'none') - (b.pipeline_matched === 'none') || a.name.localeCompare(b.name));
+  unmatched = resolved.filter((r) => !r.slug).map((r) => r.row);
+}
 
 const HEADERS = [
   'external_id', 'name', 'conventional_name', 'aliases', 'domains', 'bucket', 'tags',

scripts/export-crm-people-csv.mjs

@@ -31,6 +31,12 @@ for (let i = 2; i < process.argv.length; i += 1) {
   const k = process.argv[i];
   if (k === '--client') args.client = process.argv[++i];
   else if (k === '--out-dir') args.outDir = process.argv[++i];
+  // Scope people to the orgs actually being imported (operator ruling
+  // 2026-07-27: pipeline orgs only; event-based creations are kept noise).
+  // Reads the external_id column of a Phase-1 orgs.csv; only persons with
+  // at least one edge to an included org export, and their PRIMARY attach
+  // is always an included org. Omit for the all-persons export.
+  else if (k === '--orgs-csv') args.orgsCsv = process.argv[++i];
 }
 const today = new Date().toISOString().slice(0, 10);
 const OUT_DIR = resolve(args.outDir ?? `clients/${args.client}/outputs/${today}_crm-starter`);
@@ -53,6 +59,30 @@ const RELEVANCE_RANK = {
 };
 const rankOf = (r) => RELEVANCE_RANK[r ?? ''] ?? 2;
 
+// Minimal CSV parse (quoted multiline cells) — only to pull external_id.
+let includedSlugs = null;
+if (args.orgsCsv) {
+  const { readFileSync } = await import('node:fs');
+  const text = readFileSync(resolve(args.orgsCsv), 'utf8');
+  const rows = [];
+  let row = [], cell = '', inQ = false;
+  for (let i = 0; i < text.length; i += 1) {
+    const c = text[i];
+    if (inQ) {
+      if (c === '"' && text[i + 1] === '"') { cell += '"'; i += 1; }
+      else if (c === '"') inQ = false;
+      else cell += c;
+    } else if (c === '"') inQ = true;
+    else if (c === ',') { row.push(cell); cell = ''; }
+    else if (c === '\n') { row.push(cell); rows.push(row); row = []; cell = ''; }
+    else if (c !== '\r') cell += c;
+  }
+  const [h, ...data] = rows;
+  const col = h.indexOf('external_id');
+  includedSlugs = new Set(data.map((r) => r[col]).filter(Boolean));
+  console.log(`scoping to ${includedSlugs.size} orgs from ${args.orgsCsv}`);
+}
+
 const db = new Surreal();
 await db.connect(process.env.SURREAL_URL);
 await db.signin({ username: process.env.SURREAL_USER, password: process.env.SURREAL_PASS });
@@ -85,10 +115,16 @@ console.log(`person→org edges: ${edges.length}`);
 
 // 3. Shape one row per person.
 const exported_at = new Date().toISOString();
-const rows = persons.map((p) => {
+const rows = persons.flatMap((p) => {
   const my = (edgesByPerson.get(String(p.person_uuid)) ?? [])
     .slice()
-    .sort((a, b) => rankOf(b.relevance) - rankOf(a.relevance));
+    // Included-org edges outrank everything (the primary attach must be an
+    // org that exists in the import), then relevance.
+    .sort((a, b) =>
+      (includedSlugs ? includedSlugs.has(b.org_slug) - includedSlugs.has(a.org_slug) : 0) ||
+      rankOf(b.relevance) - rankOf(a.relevance));
+  // Scoped run: skip persons with no edge into the included org set.
+  if (includedSlugs && !my.some((e) => includedSlugs.has(e.org_slug))) return [];
   const primary = my[0];
   const extras = my.slice(1);
   const links = p.personal_links ?? [];