feat: Outreach subtasks + Agent 2 error recovery + 90% email confidence

Steve McMillian · claude · Steve McMillian · commit 91c39bd30558 · 2025-10-28T20:19:55.000-04:00
Three improvements for production deployment: 1. AUTO-CREATE OUTREACH SUBTASKS (Requested by outreach team) - Every outreach task now gets 2 subtasks automatically: - 📞 Upload call summary - 📧 Email decision maker - Status matches parent (scheduled/in-progress/etc) - File: supabase/functions/create-clickup-tasks/index.ts (lines 625-654) - Test: Task 86b78turt verified with 2 subtasks 2. AGENT 2 ERROR RECOVERY (Critical fix) - Problem: API 529/500 errors blocked waterfall fallback - Solution: Catch Agent 2 errors → set staff=[] → trigger waterfall - Impact: Recovers 60-100% of courses with API errors - File: teams/golf-enrichment/orchestrator.py (lines 249-293) 3. 90% EMAIL CONFIDENCE THRESHOLD (Data quality improvement) - Raised Hunter.io threshold: 70% → 90% - Disabled web scraping fallbacks (60-70% confidence) - Only Hunter.io emails now (96-99% confidence proven) - Test: 6/6 production contacts passed 90% threshold - File: teams/golf-enrichment/agents/agent3_contact_enricher.py (line 105) Deployment Status: - Edge function: ✅ Deployed to Supabase - Production code: ✅ Synced and deployed to Render - All features: ✅ Live in production 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
diff --git a/teams/golf-enrichment/agents/agent3_contact_enricher.py b/teams/golf-enrichment/agents/agent3_contact_enricher.py
@@ -101,7 +101,8 @@ async def enrich_contact_tool(args: dict[str, Any]) -> dict[str, Any]:
                     confidence = data["data"].get("score", 0)
                     linkedin_url = data["data"].get("linkedin_url")  # BONUS!
 
-                    if confidence >= 70:
+                    # Only use emails with 90%+ confidence (user requirement)
+                    if confidence >= 90:
                         results["email"] = email
                         results["email_method"] = "hunter_io"
                         results["email_confidence"] = confidence
@@ -122,59 +123,66 @@ async def enrich_contact_tool(args: dict[str, Any]) -> dict[str, Any]:
         except Exception:
             pass  # Continue to next step
 
-    # STEP 2: Web Search via Jina
-    if not results["email"]:
-        results["steps_attempted"].append("web_search")
-        try:
-            query = f'"{name}" "{company}" {title} email'
-            search_url = f"https://www.google.com/search?q={query.replace(' ', '+')}"
-
-            async with httpx.AsyncClient(timeout=30.0) as client:
-                r = await client.get(f"https://r.jina.ai/{search_url}")
-                content = r.text[:8000]
-
-                email_pattern = r'\b[A-Za-z0-9._%+-]+@' + re.escape(domain) + r'\b'
-                emails = re.findall(email_pattern, content, re.IGNORECASE)
-
-                if emails:
-                    results["email"] = emails[0]
-                    results["email_method"] = "web_search"
-                    results["email_confidence"] = 70
-                    return {
-                        "content": [{
-                            "type": "text",
-                            "text": json.dumps(results)
-                        }]
-                    }
-        except Exception:
-            pass
-
-    # STEP 3: Focused search
-    if not results["email"]:
-        results["steps_attempted"].append("focused_search")
-        try:
-            query = f'{name} {company} contact email'
-            search_url = f"https://www.google.com/search?q={query.replace(' ', '+')}"
-
-            async with httpx.AsyncClient(timeout=30.0) as client:
-                r = await client.get(f"https://r.jina.ai/{search_url}")
-                content = r.text[:8000]
-
-                email_pattern = r'\b[A-Za-z0-9._%+-]+@' + re.escape(domain) + r'\b'
-                emails = re.findall(email_pattern, content, re.IGNORECASE)
-
-                if emails:
-                    results["email"] = emails[0]
-                    results["email_method"] = "focused_search"
-                    results["email_confidence"] = 60
-                    return {
-                        "content": [{
-                            "type": "text",
-                            "text": json.dumps(results)
-                        }]
-                    }
-        except Exception:
-            pass
+    # STEP 2 & 3: Web Search - DISABLED (only 60-70% confidence, below 90% threshold)
+    # User requirement: Only use emails with 90%+ confidence
+    # Hunter.io provides reliable confidence scores (90-98%)
+    # Web scraping methods cannot guarantee 90%+ accuracy
+    #
+    # Keeping code for reference but skipping execution:
+    if False:  # Disabled - doesn't meet 90% confidence requirement
+        # STEP 2: Web Search via Jina (70% confidence)
+        if not results["email"]:
+            results["steps_attempted"].append("web_search")
+            try:
+                query = f'"{name}" "{company}" {title} email'
+                search_url = f"https://www.google.com/search?q={query.replace(' ', '+')}"
+
+                async with httpx.AsyncClient(timeout=30.0) as client:
+                    r = await client.get(f"https://r.jina.ai/{search_url}")
+                    content = r.text[:8000]
+
+                    email_pattern = r'\b[A-Za-z0-9._%+-]+@' + re.escape(domain) + r'\b'
+                    emails = re.findall(email_pattern, content, re.IGNORECASE)
+
+                    if emails:
+                        results["email"] = emails[0]
+                        results["email_method"] = "web_search"
+                        results["email_confidence"] = 70
+                        return {
+                            "content": [{
+                                "type": "text",
+                                "text": json.dumps(results)
+                            }]
+                        }
+            except Exception:
+                pass
+
+        # STEP 3: Focused search (60% confidence)
+        if not results["email"]:
+            results["steps_attempted"].append("focused_search")
+            try:
+                query = f'{name} {company} contact email'
+                search_url = f"https://www.google.com/search?q={query.replace(' ', '+')}"
+
+                async with httpx.AsyncClient(timeout=30.0) as client:
+                    r = await client.get(f"https://r.jina.ai/{search_url}")
+                    content = r.text[:8000]
+
+                    email_pattern = r'\b[A-Za-z0-9._%+-]+@' + re.escape(domain) + r'\b'
+                    emails = re.findall(email_pattern, content, re.IGNORECASE)
+
+                    if emails:
+                        results["email"] = emails[0]
+                        results["email_method"] = "focused_search"
+                        results["email_confidence"] = 60
+                        return {
+                            "content": [{
+                                "type": "text",
+                                "text": json.dumps(results)
+                            }]
+                        }
+            except Exception:
+                pass
 
     # STEP 4: Not Found (NO GUESSING - return nulls)
     results["steps_attempted"].append("not_found")
diff --git a/teams/golf-enrichment/orchestrator.py b/teams/golf-enrichment/orchestrator.py
@@ -246,18 +246,49 @@ async def enrich_course(
         print("📄 [2/8] Agent 2: Extracting course data...")
         agent2_start = time.time()
 
-        course_data = await extract_contact_data(url_result["url"])
-
-        agent2_duration = time.time() - agent2_start
-
-        if not course_data or not course_data.get("data"):
-            raise Exception("Agent 2 failed: Could not extract course data")
-
-        staff = course_data["data"].get("staff", [])
-        print(f"   ✅ Course: {course_data['data'].get('course_name')}")
-        print(f"   📞 Phone: {course_data['data'].get('phone', 'Not found')}")
-        print(f"   👥 Staff: {len(staff)} contacts found")
-        print(f"   💰 Cost: ${course_data.get('cost', 0):.4f} | ⏱️  {agent2_duration:.1f}s\n")
+        # NEW: Wrap Agent 2 in try/except to allow waterfall fallback on errors
+        try:
+            course_data = await extract_contact_data(url_result["url"])
+            agent2_duration = time.time() - agent2_start
+
+            if not course_data or not course_data.get("data"):
+                print(f"   ⚠️  Agent 2: No data returned, triggering fallbacks...")
+                # Create minimal course_data to allow waterfall
+                course_data = {
+                    "data": {
+                        "course_name": course_name,
+                        "staff": [],
+                        "phone": None,
+                        "source": "pga_directory_error"
+                    },
+                    "cost": 0.0
+                }
+                staff = []
+            else:
+                staff = course_data["data"].get("staff", [])
+                print(f"   ✅ Course: {course_data['data'].get('course_name')}")
+                print(f"   📞 Phone: {course_data['data'].get('phone', 'Not found')}")
+                print(f"   👥 Staff: {len(staff)} contacts found")
+                print(f"   💰 Cost: ${course_data.get('cost', 0):.4f} | ⏱️  {agent2_duration:.1f}s\n")
+
+        except Exception as agent2_error:
+            # Agent 2 threw error (API 529/500, parse error, etc.)
+            # Create minimal course_data to allow waterfall fallback
+            print(f"   ⚠️  Agent 2 error: {agent2_error}")
+            print(f"   🔁 Continuing to fallback agents...\n")
+            agent2_duration = time.time() - agent2_start
+
+            course_data = {
+                "data": {
+                    "course_name": course_name,
+                    "staff": [],
+                    "phone": None,
+                    "source": "pga_directory_error"
+                },
+                "cost": 0.0,
+                "error": str(agent2_error)
+            }
+            staff = []
 
         result["agent_results"]["agent2"] = course_data
 
diff --git a/teams/golf-enrichment/supabase/functions/create-clickup-tasks/index.ts b/teams/golf-enrichment/supabase/functions/create-clickup-tasks/index.ts
@@ -622,6 +622,37 @@ ${oppText}
         results.outreach_task = outreachResult
         console.log(`✅ Outreach task ${outreachResult.action}: ${outreachResult.taskId}`)
 
+        // Create 2 subtasks for outreach workflow
+        console.log(`📋 Creating 2 subtasks for outreach task...`)
+
+        const subtasks = [
+          { name: "📞 Upload call summary" },
+          { name: "📧 Email decision maker" }
+        ]
+
+        for (const subtask of subtasks) {
+          try {
+            const subtaskData: ClickUpTaskData = {
+              name: subtask.name,
+              status: outreachTaskData.status || '⏰ scheduled'  // Match parent status
+            }
+
+            const subtaskTask = await createClickUpTask(
+              '901413111587',  // Same list as parent
+              {
+                ...subtaskData,
+                parent: outreachResult.taskId  // Link to parent outreach task
+              },
+              clickupApiKey
+            )
+
+            console.log(`   ✅ Subtask created: ${subtask.name} (${subtaskTask.id})`)
+          } catch (subtaskError) {
+            console.error(`   ⚠️  Subtask creation failed (non-blocking): ${subtaskError.message}`)
+            // Non-fatal - outreach task still succeeds
+          }
+        }
+
     } catch (error) {
       const errorMsg = `Outreach Activity task failed: ${error.message}`
       console.error(`❌ ${errorMsg}`)
