Viewer-Level Persona Mapping

import os, requests, time
from collections import defaultdict

WS = os.environ["WISTIA_API_TOKEN"]
MV = os.environ["MAVERA_API_KEY"]
WS_BASE = "https://api.wistia.com"
MV_BASE = "https://app.mavera.io/api/v1"
WS_H = {"Authorization": f"Bearer {WS}", "Accept": "application/json"}
MV_H = {"Authorization": f"Bearer {MV}", "Content-Type": "application/json"}

# 1. Fetch all visitors with engagement data (paginated)
all_visitors = []
page = 1
while True:
    resp = requests.get(f"{WS_BASE}/v1/stats/visitors.json", headers=WS_H, params={
        "per_page": 100, "page": page,
    })
    if resp.status_code == 429:
        time.sleep(2)
        continue
    resp.raise_for_status()
    visitors = resp.json()
    if not visitors:
        break
    all_visitors.extend(visitors)
    page += 1
    time.sleep(0.2)

print(f"Total visitors: {len(all_visitors)}")

# 2. Enrich each visitor with per-event details
enriched = []
for visitor in all_visitors[:200]:
    visitor_key = visitor.get("visitor_key", "")
    events = visitor.get("events", [])

    total_watched = 0
    total_available = 0
    video_count = 0
    locations = set()

    for event in events:
        pct = event.get("percent_viewed", 0)
        total_watched += pct
        total_available += 1
        video_count += 1

    avg_percent = total_watched / max(video_count, 1)

    enriched.append({
        "visitor_key": visitor_key,
        "email": visitor.get("visitor_identity", {}).get("email", "anonymous"),
        "name": visitor.get("visitor_identity", {}).get("name", ""),
        "location": visitor.get("visitor_identity", {}).get("location", "unknown"),
        "video_count": video_count,
        "avg_percent_watched": round(avg_percent, 1),
        "total_visits": visitor.get("total_visits", 0),
        "last_active": visitor.get("last_active_at", ""),
    })

# 3. Cluster viewers by behavior
clusters = defaultdict(list)
for v in enriched:
    if v["video_count"] >= 5 and v["avg_percent_watched"] >= 75:
        clusters["binge_watchers"].append(v)
    elif v["video_count"] >= 3 and v["avg_percent_watched"] >= 50:
        clusters["engaged_explorers"].append(v)
    elif v["video_count"] == 1 and v["avg_percent_watched"] < 30:
        clusters["bouncers"].append(v)
    elif v["video_count"] >= 2 and v["avg_percent_watched"] < 50:
        clusters["skimmers"].append(v)
    else:
        clusters["casual_viewers"].append(v)

print(f"\nClusters:")
for name, members in clusters.items():
    print(f"  {name}: {len(members)} viewers")

# 4. Create Mavera personas for each cluster
persona_map = {}
cluster_descriptions = {
    "binge_watchers": "Watches 5+ videos, completes 75%+ of each. Deeply engaged, likely in active evaluation or already a customer seeking mastery. High intent signal.",
    "engaged_explorers": "Watches 3-4 videos at 50-75% completion. Browsing with purpose — comparing options or building understanding. Mid-funnel prospect.",
    "bouncers": "Watched 1 video, left before 30%. Either the content missed their intent, or they arrived by accident. Lowest engagement tier.",
    "skimmers": "Watches 2+ videos but under 50% each. Scanning for specific information — not consuming narrative content. Wants answers, not stories.",
    "casual_viewers": "Moderate engagement that doesn't fit other clusters. May be returning after time away or following a specific recommendation.",
}

for cluster_name, description in cluster_descriptions.items():
    if cluster_name not in clusters:
        continue
    sample = clusters[cluster_name][:5]
    sample_text = "\n".join(
        f"  - {s['email']}: {s['video_count']} videos, {s['avg_percent_watched']}% avg, {s['total_visits']} visits"
        for s in sample
    )

    persona = requests.post(f"{MV_BASE}/personas", headers=MV_H, json={
        "name": cluster_name.replace("_", " ").title(),
        "description": f"{description}\n\nSample viewers:\n{sample_text}",
    }).json()
    persona_map[cluster_name] = persona["id"]
    time.sleep(0.3)

# 5. Analyze via Mave with persona context
cluster_block = "\n\n".join(
    f"CLUSTER: {name.replace('_', ' ').title()} ({len(members)} viewers)\n"
    f"  Avg videos watched: {sum(m['video_count'] for m in members) / len(members):.1f}\n"
    f"  Avg completion: {sum(m['avg_percent_watched'] for m in members) / len(members):.1f}%\n"
    f"  Avg total visits: {sum(m['total_visits'] for m in members) / len(members):.1f}\n"
    f"  Identified (with email): {sum(1 for m in members if m['email'] != 'anonymous')}"
    for name, members in clusters.items() if members
)

analysis = requests.post(f"{MV_BASE}/mave/chat", headers=MV_H, json={
    "message": f"""Map these viewer clusters to marketing personas and suggest outreach strategies.

VIEWER CLUSTERS:
{cluster_block}

Total viewers analyzed: {len(enriched)}

For each cluster:
1. **Persona Profile**: Name, motivation, likely job title/role, what they're looking for
2. **Content Affinity**: What video types/topics would resonate most with this persona?
3. **Outreach Strategy**: Best channel (email, retargeting, sales call) and message angle
4. **Conversion Probability**: Estimate likelihood this cluster converts to customer (high/medium/low)
5. **Content Gaps**: What video content are we missing that this persona would need?

End with: Which cluster represents the highest-value opportunity and what single action would convert them?""",
}).json()

print("\nVIEWER-LEVEL PERSONA MAPPING")
print("=" * 60)
for name, members in clusters.items():
    avg_pct = sum(m["avg_percent_watched"] for m in members) / len(members) if members else 0
    identified = sum(1 for m in members if m["email"] != "anonymous")
    print(f"  {name.replace('_', ' ').title():<22} {len(members):>4} viewers  "
          f"Avg:{avg_pct:>5.1f}%  Identified:{identified:>3}")
print("\n" + analysis.get("content", "")[:2000])

const WS = process.env.WISTIA_API_TOKEN;
const MV = process.env.MAVERA_API_KEY;
const WS_BASE = "https://api.wistia.com";
const MV_BASE = "https://app.mavera.io/api/v1";
const WS_H = { Authorization: `Bearer ${WS}`, Accept: "application/json" };
const MV_H = { Authorization: `Bearer ${MV}`, "Content-Type": "application/json" };

// 1. Fetch visitors (paginated)
const allVisitors = [];
let page = 1;
while (true) {
  const resp = await fetch(
    `${WS_BASE}/v1/stats/visitors.json?per_page=100&page=${page}`, { headers: WS_H }
  );
  if (resp.status === 429) { await new Promise(r => setTimeout(r, 2000)); continue; }
  const visitors = await resp.json();
  if (!visitors.length) break;
  allVisitors.push(...visitors);
  page++;
  await new Promise(r => setTimeout(r, 200));
}

console.log(`Total visitors: ${allVisitors.length}`);

// 2. Enrich
const enriched = allVisitors.slice(0, 200).map(visitor => {
  const events = visitor.events || [];
  const totalWatched = events.reduce((s, e) => s + (e.percent_viewed || 0), 0);
  const videoCount = events.length;
  return {
    visitorKey: visitor.visitor_key || "",
    email: visitor.visitor_identity?.email || "anonymous",
    name: visitor.visitor_identity?.name || "",
    location: visitor.visitor_identity?.location || "unknown",
    videoCount, avgPercentWatched: videoCount > 0 ? Math.round(totalWatched / videoCount * 10) / 10 : 0,
    totalVisits: visitor.total_visits || 0,
    lastActive: visitor.last_active_at || "",
  };
});

// 3. Cluster
const clusters = { binge_watchers: [], engaged_explorers: [], bouncers: [], skimmers: [], casual_viewers: [] };
for (const v of enriched) {
  if (v.videoCount >= 5 && v.avgPercentWatched >= 75) clusters.binge_watchers.push(v);
  else if (v.videoCount >= 3 && v.avgPercentWatched >= 50) clusters.engaged_explorers.push(v);
  else if (v.videoCount === 1 && v.avgPercentWatched < 30) clusters.bouncers.push(v);
  else if (v.videoCount >= 2 && v.avgPercentWatched < 50) clusters.skimmers.push(v);
  else clusters.casual_viewers.push(v);
}

console.log("\nClusters:");
for (const [name, members] of Object.entries(clusters)) {
  if (members.length) console.log(`  ${name}: ${members.length} viewers`);
}

// 4. Create personas
const clusterDescs = {
  binge_watchers: "5+ videos, 75%+ completion. Deep engagement — active evaluation or mastery-seeking.",
  engaged_explorers: "3-4 videos, 50-75%. Browsing with purpose, mid-funnel.",
  bouncers: "1 video, <30%. Mismatched intent or accidental arrival.",
  skimmers: "2+ videos, <50% each. Scanning for answers, not consuming narratives.",
  casual_viewers: "Moderate engagement, doesn't fit other clusters.",
};

const personaMap = {};
for (const [name, desc] of Object.entries(clusterDescs)) {
  if (!clusters[name]?.length) continue;
  const sample = clusters[name].slice(0, 5).map(s =>
    `  - ${s.email}: ${s.videoCount} videos, ${s.avgPercentWatched}% avg, ${s.totalVisits} visits`
  ).join("\n");

  const persona = await fetch(`${MV_BASE}/personas`, {
    method: "POST", headers: MV_H,
    body: JSON.stringify({ name: name.replace(/_/g, " ").replace(/\b\w/g, c => c.toUpperCase()), description: `${desc}\n\nSample:\n${sample}` }),
  }).then(r => r.json());
  personaMap[name] = persona.id;
  await new Promise(r => setTimeout(r, 300));
}

// 5. Analyze
const clusterBlock = Object.entries(clusters)
  .filter(([, m]) => m.length)
  .map(([name, members]) => {
    const avgVids = members.reduce((s, m) => s + m.videoCount, 0) / members.length;
    const avgPct = members.reduce((s, m) => s + m.avgPercentWatched, 0) / members.length;
    const identified = members.filter(m => m.email !== "anonymous").length;
    return `CLUSTER: ${name.replace(/_/g, " ")} (${members.length} viewers)\n  Avg videos: ${avgVids.toFixed(1)} | Avg completion: ${avgPct.toFixed(1)}% | Identified: ${identified}`;
  }).join("\n\n");

const analysis = await fetch(`${MV_BASE}/mave/chat`, {
  method: "POST", headers: MV_H,
  body: JSON.stringify({
    message: `Map viewer clusters to personas with outreach strategies.\n\n${clusterBlock}\n\nTotal: ${enriched.length}\n\nPer cluster:\n1. Persona profile (name, motivation, role)\n2. Content affinity\n3. Outreach strategy (channel + message)\n4. Conversion probability\n5. Content gaps\n\nEnd with: highest-value cluster and single action to convert them`,
  }),
}).then(r => r.json());

console.log("\nVIEWER-LEVEL PERSONA MAPPING");
console.log("=".repeat(60));
for (const [name, members] of Object.entries(clusters)) {
  if (!members.length) continue;
  const avgPct = members.reduce((s, m) => s + m.avgPercentWatched, 0) / members.length;
  const identified = members.filter(m => m.email !== "anonymous").length;
  console.log(`  ${name.replace(/_/g, " ").padEnd(22)} ${String(members.length).padStart(4)} viewers  ` +
    `Avg:${avgPct.toFixed(1).padStart(5)}%  Identified:${String(identified).padStart(3)}`);
}
console.log("\n" + (analysis.content || "").slice(0, 2000));