Intent Analysis System

Takes company details and returns intent evidence — buying/growth signals discovered for that specific company, each with a source URL, a type, and details.

This repository is separate from the sourcing model.

Input / Output

Input:   { name, domain, linkedin? }
Output:  { company, evidence: [ { evidence_url, evidence_type, details } ], analyzed_at }

• schema/company_input.schema.json — input contract (JSON Schema)
• schema/intent_output.schema.json — output contract (JSON Schema)
• intent_analysis/models.py — same contract as Python dataclasses

Evidence types

hiring · pr · funding · tech_stack · news · social_media

How it works (end to end)

ScrapingDog-first — SD does both discovery and extraction. For each evidence type:

1. Discovery — ScrapingDog Google search finds candidate evidence URLs. (Exa /search is an optional fallback: set INTENT_DISCOVERY=exa.)
2. Extraction — ScrapingDog AI mode (ai_query) reads each page and returns a one-sentence details string, or NONE when the page isn't evidence for this company.
3. Verify (optional) — an OpenRouter LLM second pass re-confirms the item is genuine evidence for this company (INTENT_USE_OPENROUTER=1).

Only confirmed items are returned.

company ─▶ SD Google search ─▶ SD AI extract ─▶ (opt) OpenRouter verify ─▶ evidence[]

Usage

pip install -r requirements.txt

export EXA_API_KEY=...
export SCRAPINGDOG_API_KEY=...
export OPENROUTER_API_KEY=...

python cli.py --name "Acme Inc" --domain acme.com
# optional: restrict evidence types
python cli.py --name "Acme Inc" --domain acme.com --types hiring funding news

Or as a library:

from intent_analysis import CompanyInput, analyze_intent

result = analyze_intent(CompanyInput(name="Acme Inc", domain="acme.com"))
print(result.to_dict())

Configuration

Env var	Purpose	Required
SCRAPINGDOG_API_KEY	SD Google search + AI extraction (primary)	yes
EXA_API_KEY	Optional discovery fallback (INTENT_DISCOVERY=exa)	no
OPENROUTER_API_KEY	Optional LLM verify pass (INTENT_USE_OPENROUTER=1)	no
INTENT_LLM_MODEL	Model override (default openai/gpt-4o-mini)	no

See .env.example. Only SCRAPINGDOG_API_KEY is needed for the default path.