scripts: chicago analyzer field-name fixes + vectorize sanitizer hardening

Two small fixes surfaced during smoke testing:

analyze_chicago_contracts.ts: permit field is contact_1_name not
contact_1; reported_cost is integer-string. Fixed filter (was rejecting
all 2853 permits) and contractor extraction (was empty).

vectorize_raw_corpus.ts: sanitize() expanded to strip control chars +
ALL backslashes (kills incomplete \uXXXX escapes) + UTF-16 surrogates
(unpaired surrogates from emoji split by truncate boundary). Llm_team
response cache had docs with all three pollution shapes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

scripts/vectorize_raw_corpus.ts

@@ -46,6 +46,18 @@ function truncate(s: string, n = 4000): string {
   return s == null ? "" : (s.length > n ? s.slice(0, n) : s);
 }
 
+// Sanitize text before posting as JSON. Strips control chars and
+// drops incomplete \uXXXX escape sequences which break Rust's
+// serde JSON parser at the gateway. Llm_team response cache had
+// rows with truncated \u escapes that 400'd the whole batch.
+function sanitize(s: string): string {
+  if (!s) return "";
+  return s
+    .replace(/[\x00-\x08\x0B\x0C\x0E-\x1F\x7F]/g, "") // strip control chars
+    .replace(/\\/g, "") // strip ALL backslashes — kills any malformed \uXXXX in source data
+    .replace(/[\uD800-\uDFFF]/g, ""); // strip UTF-16 surrogates (lone ones from emoji split by truncate)
+}
+
 // ─── EXTRACTORS — one per source ───
 // Each shapes raw rows into {id, text} for the gateway's chunker.
 
@@ -129,13 +141,13 @@ function extractLlmTeamResponseCache(raw: string): Doc[] {
       const text = [
         `Cached response ${r.cache_key ?? r.id ?? i}`,
         `Created: ${r.created_at ?? "?"}`,
-        r.prompt ? `Prompt: ${truncate(r.prompt, 1500)}` : "",
-        r.response ? `Response: ${truncate(r.response, 2500)}` : "",
+        r.prompt ? `Prompt: ${sanitize(truncate(r.prompt, 1500))}` : "",
+        r.response ? `Response: ${sanitize(truncate(r.response, 2500))}` : "",
         r.model ? `Model: ${r.model}` : "",
       ].filter(Boolean).join("\n");
       return { id: `resp_${r.cache_key ?? r.id ?? i}`, text };
     } catch {
-      return { id: `resp_${i}`, text: line.slice(0, 2000) };
+      return { id: `resp_${i}`, text: sanitize(line.slice(0, 2000)) };
     }
   });
 }