deep call example

gistfile1.txt

SYSTEM_PROMPT = "You are an expert analyst scoring companies on their potential to generate Exa web search API volume.

## What Exa Is
Exa is a neural web search API. Developers use it to programmatically search the internet.

## Core Question
How much web search API volume could a deal with this company generate — now or in the future?
Score ONLY on volume potential, not ease of closing.

## Key Signals
- AI-nativeness × web dependency is the core signal (high on both = good prospect)
- Future potential matters — well-funded AI startups with search-heavy thesis score high even pre-revenue
- Volume sources: product search, training pipelines, background enrichment, platform multiplication

## 7-Point Scale
7 — Crown Jewels: top ~100 globally, could be top-10 Exa account
6 — High Value: strong AI-native, clear search volume path
5 — Strong Prospect: concrete search use case with scale
4 — Moderate: AI meaningful but search is "nice to have" not "need to have"
3 — Low Fit: AI exists but web search is peripheral
2 — Minimal Fit: little AI or no web data connection
1 — No Fit: no AI, no web data need, or Exa competitor

## Output
Ground your score in specific verifiable facts. Don't hallucinate. Missing data = lower confidence, not fabrication."

OUTPUT_SCHEMA = {
    "type": "object",
    "required": ["score", "exa_case", "exa_pitch", "deal_archetype", "confidence", "score_rationale"],
    "properties": {
        "score": {"type": "integer", "description": "1-7 per the framework"},
        "exa_case": {"type": "string", "description": "≤20 word internal summary, volume-centric"},
        "exa_pitch": {"type": "string", "description": "2 sentences for an AE: what product feature uses Exa + volume implication"},
        "deal_archetype": {"type": "string", "description": "Firehose | Training Furnace | Platform Kingmaker | Persistent Intelligence | Inevitable Giant | Volume by Mass | None"},
        "confidence": {"type": "string", "description": "high | medium | low"},
        "score_rationale": {"type": "string", "description": "3-5 sentences: scoring logic, why this tier, volume math"},
    },
}

def build_additional_queries(name, domain):
    return [
        f"{name} ({domain}) product features, technology stack, and AI capabilities",
        f"{name} ({domain}) funding rounds, investors, valuation, and total raised",
        f"{name} ({domain}) revenue, ARR, users, customers, and growth metrics",
        f"{name} ({domain}) employee count, headcount growth, engineering team size",
        f"{name} ({domain}) competitors, market position, and competitive landscape",
        f"{name} ({domain}) latest news, product launches, and partnerships 2025 2026",
        f"{name} ({domain}) use of web search, web data, crawling, or external data APIs",
        f"{name} ({domain}) AI training data, reinforcement learning, data pipelines",
    ]

result = exa.search(
    f"Evaluate {name} ({domain}) as a potential Exa web search API customer. "
    f"Research their product, AI capabilities, scale, funding, growth trajectory, "
    f"and how they use or could use web data. Determine their search volume potential.",
    type="deep",
    system_prompt=SYSTEM_PROMPT,
    output_schema=OUTPUT_SCHEMA,
    additional_queries=build_additional_queries(name, domain),
)