AI Summary → News Digest with Personas

Scenario

Perigon provides AI-generated summaries for articles, saving you from parsing raw content. This job pulls summarized articles for your industry, creates Mavera personas representing different audience segments, then rewrites the digest in each persona’s voice — so your newsletter can have a section for executives, one for practitioners, and one for students.

Flow: Perigon GET /all (with summaries) → Mavera POST /personas → POST /speak per persona → Multi-voice digest

Code

Python

import os, requests, time

PG_KEY = os.environ["PERIGON_API_KEY"]
PG_BASE = "https://api.goperigon.com/v1"
MV = os.environ["MAVERA_API_KEY"]
MV_BASE = "https://app.mavera.io/api/v1"
MV_H = {"Authorization": f"Bearer {MV}", "Content-Type": "application/json"}

TOPIC = "fintech"

# 1. Fetch articles with summaries
r = requests.get(f"{PG_BASE}/all", params={
    "apiKey": PG_KEY, "q": TOPIC, "sortBy": "date",
    "size": 12, "sourceGroup": "top100",
})
r.raise_for_status()
articles = r.json().get("articles", [])
print(f"Fetched {len(articles)} articles for '{TOPIC}'")

# 2. Build digest from summaries
digest_items = []
for a in articles:
    summary = a.get("summary", a.get("description", ""))
    if not summary:
        continue
    digest_items.append(
        f"**{a.get('title','')}** ({a.get('source',{}).get('name','')})\n{summary[:400]}"
    )

digest_text = "\n\n".join(digest_items[:10])

# 3. Create audience personas
AUDIENCES = [
    {"name": "C-Suite Executive", "desc": "CEO/CFO at a fintech company. Cares about market trends, regulation, fundraising signals, and strategic positioning. Prefers bullet points, numbers, and bottom-line impact."},
    {"name": "Product Manager", "desc": "PM at a fintech startup. Cares about feature trends, user expectations, competitive moves, and technology shifts. Prefers actionable insights with implementation angles."},
    {"name": "Finance Student", "desc": "Graduate student studying financial technology. Cares about industry structure, career opportunities, and understanding how concepts connect. Prefers explanations with context."},
]
persona_ids = []
for aud in AUDIENCES:
    p = requests.post(f"{MV_BASE}/personas", headers=MV_H, json={
        "name": f"Digest: {aud['name']}",
        "description": aud["desc"],
    }).json()
    persona_ids.append({"id": p["id"], "name": aud["name"]})
    time.sleep(0.3)

# 4. Generate persona-voiced digest
for persona in persona_ids:
    speak = requests.post(f"{MV_BASE}/speak", headers=MV_H, json={
        "persona_id": persona["id"],
        "input": [
            {"role": "system", "content": f"You are a {persona['name']}. Rewrite this news digest in YOUR voice — the way you'd summarize it for peers. 5-7 bullet points. Match your communication style."},
            {"role": "user", "content": f"This week's {TOPIC} news:\n\n{digest_text}"},
        ],
        "mode": "interview",
    }).json()

    print(f"\n{'='*60}\nDIGEST FOR: {persona['name']}\n{'='*60}")
    for ex in speak.get("exchanges", speak.get("messages", [])):
        if ex.get("role") == "assistant":
            print(ex["content"][:600])
    time.sleep(0.3)

Example Output

Fetched 12 articles for 'fintech'

DIGEST FOR: C-Suite Executive
============================================================
• Stripe valuation hits $95B — signals public market appetite for payments
  infrastructure. Watch for IPO filing by Q3.
• EU PSD3 draft tightens open banking. Compliance cost: estimate €2-5M for
  mid-size players. Budget now.
• Plaid acquires identity verification startup. Build-vs-buy calculus shifts.
• Neobank customer acquisition costs up 34% YoY. CAC/LTV ratios compressing.

DIGEST FOR: Finance Student
============================================================
• Stripe just hit a $95B valuation — that's the highest for any private fintech.
  To put it in context, that's more than Goldman Sachs was worth 10 years ago.
  An IPO would test whether public markets agree with private valuations.
• The EU is working on PSD3, which updates rules about how banks share data.
  PSD2 created open banking — PSD3 adds more security requirements.

Error Handling

Missing summaries

Not all articles have AI summaries. Fall back to description field. If both are empty, skip the article.

Persona voice drift

If persona output sounds generic, add more specifics to the description — communication style, jargon preferences, typical sentence length.

Digest length

10 articles × 400 chars = ~4,000 chars. Within Mavera’s context limits. For larger digests, split into two Speak sessions.