demo: P1 — search filter now actually filters by state and role

The Co-Pilot search box read state and role from the dropdowns (#sst, #srl)
but appended them to the message string as ' in '+st. The server's NL
parser then matched the literal preposition "in" against the case-insensitive
regex /\b(IL|IN|...)\b/i and assigned state IN (Indiana) to every search.
Result: typing "forklift in IL" returned Indiana workers. Same for WI, TX,
any state — all silently became Indiana. That was the "cached/generic
response" the legacy staffing client was seeing.

Two prongs:

1. search.html doSearch() now passes structured fields:
     {message, state, role}
   instead of munging into the message text. Dropdown selections bypass
   NL parsing entirely.

2. /intelligence/chat smart_search route accepts those structured fields
   and prefers them over regex archaeology. Falls back to NL parsing only
   when fields aren't provided. Fixed the regex too: the prepositional
   form (?:in|from)\s+(STATE) wins, the standalone form requires uppercase
   (drops /i flag) so the lowercase preposition "in" can no longer match.

Verified live:
- POST /intelligence/chat {"message":"forklift","state":"IL"}
    → 167 IL forklift operators (Galesburg, Joliet, ...)
- POST /intelligence/chat {"message":"forklift","state":"WI","role":"Forklift Operator"}
    → 16 WI Forklift Operators (Milwaukee, Madison, ...)
- POST /intelligence/chat {"message":"forklift in IL"} (NL fallback)
    → 167 IL workers (regex now correctly distinguishes preposition from state code)

Playwright drove the live UI through devop.live/lakehouse and confirmed the
front-end posts the structured body and the result panel renders the right
state. Restart sequence: kill old bun :3700, bun run mcp-server/index.ts.

mcp-server/index.ts

@@ -1698,7 +1698,14 @@ async function main() {
             const filters: string[] = ["CAST(reliability AS DOUBLE) >= 0.5"];
             const understood: string[] = [];
 
-            // Extract role keywords
+            // Structured input from the search-form dropdowns. When set,
+            // these win over NL parsing — typing "forklift in IL" used to
+            // misparse the preposition "in" as state IN (Indiana). Trust
+            // explicit user selection over regex archaeology.
+            const explicitState = String(b.state || "").trim().toUpperCase();
+            const explicitRole = String(b.role || "").trim();
+
+            // Extract role keywords (skip if dropdown picked one)
             const roleKeywords: Record<string, string> = {
               "warehouse": "warehouse", "forklift": "forklift", "welder": "weld", "assembler": "assembl",
               "loader": "loader", "machine operator": "machine operator", "shipping": "shipping",
@@ -1707,8 +1714,13 @@ async function main() {
               "line lead": "line lead", "electrician": "electric", "packaging": "packaging",
               "tool and die": "tool", "logistics": "logistics", "safety": "safety", "cnc": "cnc",
             };
-            for (const [kw, sqlPart] of Object.entries(roleKeywords)) {
-              if (lower.includes(kw)) { filters.push(`LOWER(role) LIKE '%${sqlPart}%'`); understood.push(`role: ${kw}`); break; }
+            if (explicitRole) {
+              filters.push(`LOWER(role) LIKE '%${explicitRole.toLowerCase().replace(/'/g, "''")}%'`);
+              understood.push(`role: ${explicitRole}`);
+            } else {
+              for (const [kw, sqlPart] of Object.entries(roleKeywords)) {
+                if (lower.includes(kw)) { filters.push(`LOWER(role) LIKE '%${sqlPart}%'`); understood.push(`role: ${kw}`); break; }
+              }
             }
 
             // Extract city
@@ -1726,18 +1738,33 @@ async function main() {
               }
             }
 
-            // Extract state
+            // Extract state — dropdown wins; otherwise NL parse, but
+            // require either an explicit "in/from <STATE>" preposition
+            // OR an UPPERCASE 2-letter code, never a bare lowercase
+            // 2-letter token. Old regex matched "in" (preposition) as
+            // state IN (Indiana) because the /i flag made the standalone
+            // pattern case-insensitive — "forklift in IL" always returned
+            // Indiana workers.
             const stateNames: Record<string, string> = {
               "illinois":"IL","indiana":"IN","ohio":"OH","missouri":"MO","tennessee":"TN",
               "kentucky":"KY","wisconsin":"WI","michigan":"MI","iowa":"IA","minnesota":"MN"
             };
-            const stateMatch = lower.match(/\b(IL|IN|OH|MO|TN|KY|WI|MI|IA|MN)\b/i);
-            if (stateMatch && !understood.some(u => u.startsWith('city'))) {
-              filters.push(`state = '${stateMatch[1].toUpperCase()}'`);
-              understood.push(`state: ${stateMatch[1].toUpperCase()}`);
+            if (explicitState) {
+              if (!understood.some(u => u.startsWith('city'))) {
+                filters.push(`state = '${explicitState.replace(/'/g, "''")}'`);
+                understood.push(`state: ${explicitState}`);
+              }
             } else {
-              for (const [name, abbr] of Object.entries(stateNames)) {
-                if (lower.includes(name)) { filters.push(`state = '${abbr}'`); understood.push(`state: ${abbr}`); break; }
+              const prepMatch = q.match(/\b(?:in|from)\s+(IL|IN|OH|MO|TN|KY|WI|MI|IA|MN)\b/i);
+              const upperMatch = q.match(/\b(IL|IN|OH|MO|TN|KY|WI|MI|IA|MN)\b/);  // no /i — must be uppercase
+              const stateMatch = prepMatch || upperMatch;
+              if (stateMatch && !understood.some(u => u.startsWith('city'))) {
+                filters.push(`state = '${stateMatch[1].toUpperCase()}'`);
+                understood.push(`state: ${stateMatch[1].toUpperCase()}`);
+              } else {
+                for (const [name, abbr] of Object.entries(stateNames)) {
+                  if (lower.includes(name)) { filters.push(`state = '${abbr}'`); understood.push(`state: ${abbr}`); break; }
+                }
               }
             }
 

mcp-server/search.html

@@ -2274,13 +2274,15 @@ function doSearch(){
   var q=document.getElementById('sq').value.trim();if(!q)return;
   lastQuery=q;
   var st=document.getElementById('sst').value,rl=document.getElementById('srl').value;
-  // Append dropdown filters to the query so the smart parser picks them up
-  var fullQ=q;
-  if(st&&q.indexOf(st)<0)fullQ+=' in '+st;
-  if(rl&&q.toLowerCase().indexOf(rl.toLowerCase())<0)fullQ+=' '+rl;
+  // Pass dropdown filters as structured fields. Old code appended
+  // ' in '+st to the message, which the server misparsed: the
+  // preposition "in" matched the regex for state code "IN" (Indiana)
+  // and every search returned Indiana workers regardless of dropdown.
+  // Sending structured state/role lets the server skip NL parsing
+  // for those fields entirely.
   var out=document.getElementById('sresults');out.textContent='Finding the best matches...';
   fetch(A+'/intelligence/chat',{method:'POST',headers:{'Content-Type':'application/json'},
-  body:JSON.stringify({message:fullQ})
+  body:JSON.stringify({message:q,state:st||undefined,role:rl||undefined})
   }).then(function(r){return r.json()}).then(function(d){
     out.textContent='';
     // Show what the system understood