import os, requests, time
from datetime import datetime, timedelta
DC_TOKEN = os.environ["DISCORD_BOT_TOKEN"]
DC_BASE = "https://discord.com/api/v10"
DC_H = {"Authorization": f"Bot {DC_TOKEN}"}
MV = os.environ["MAVERA_API_KEY"]
MV_BASE = "https://app.mavera.io/api/v1"
MV_H = {"Authorization": f"Bearer {MV}", "Content-Type": "application/json"}
CHANNELS = {
"general": "CHANNEL_ID_1",
"support": "CHANNEL_ID_2",
"feedback": "CHANNEL_ID_3",
}
DAYS_BACK = 7
cutoff = datetime.utcnow() - timedelta(days=DAYS_BACK)
snowflake_cutoff = str(int((cutoff.timestamp() - 1420070400) * 1000) << 22)
# 1. Fetch messages from each channel
all_messages = {}
for name, channel_id in CHANNELS.items():
msgs = []
after = snowflake_cutoff
while True:
r = requests.get(f"{DC_BASE}/channels/{channel_id}/messages", headers=DC_H,
params={"limit": 100, "after": after})
if r.status_code == 429:
retry = r.json().get("retry_after", 1)
time.sleep(retry)
continue
r.raise_for_status()
batch = r.json()
if not batch:
break
msgs.extend(batch)
after = batch[0]["id"]
time.sleep(0.5)
human = [m for m in msgs if not m.get("author",{}).get("bot",False)
and len(m.get("content","")) > 10]
all_messages[name] = human
print(f"#{name}: {len(human)} messages")
# 2. Build per-channel corpus
corpus_parts = []
for name, msgs in all_messages.items():
corpus_parts.append(f"\n## #{name} ({len(msgs)} messages)")
for m in sorted(msgs, key=lambda x: x["id"])[-40:]:
corpus_parts.append(
f"- [{m.get('timestamp','')[:10]}] {m.get('content','')[:300]}"
)
corpus = "\n".join(corpus_parts)
# 3. Mave sentiment analysis
report = requests.post(f"{MV_BASE}/mave/chat", headers=MV_H, json={
"message": f"Community sentiment analyst. Analyze {sum(len(v) for v in all_messages.values())} Discord messages from the past {DAYS_BACK} days.\n\n"
f"{corpus[:8000]}\n\n"
"Produce a WEEKLY COMMUNITY SENTIMENT REPORT:\n\n"
"1. **Overall Sentiment** — Score (1-100) with label (thriving/healthy/neutral/concerning/critical)\n"
"2. **Channel Breakdown** — Sentiment per channel with top topic\n"
"3. **Hot Topics** — Top 5 discussion themes ranked by volume + sentiment\n"
"4. **Praise Highlights** — What the community loves right now (with quotes)\n"
"5. **Pain Points** — Top 3 frustrations (with quotes)\n"
"6. **Emerging Trends** — New topics that weren't discussed last week\n"
"7. **Recommended Actions** — 3 specific things the team should do this week\n\n"
"Include representative quotes for each section."
}).json()
print(f"\n{'='*60}\nWEEKLY COMMUNITY SENTIMENT REPORT\n{cutoff.strftime('%b %d')} – {datetime.utcnow().strftime('%b %d, %Y')}\n{'='*60}")
print(report.get("content", "")[:3000])
