Appearance
Leads and research
Lead page
Facts from the form, salesperson, source link, research brief, call notes, quotes, contract, payment, and a full timeline of everything that happened to the lead.
Lead statuses
stateDiagram-v2
[*] --> new
new --> call_booked
call_booked --> call_done
call_done --> quote_sent
quote_sent --> contract_sent
contract_sent --> signed
signed --> paid
paid --> converted
call_booked --> no_show
quote_sent --> declined
new --> lost
call_done --> lost
new → call booked → call done → quote sent → contract sent → signed → paid → converted, with side states no show, declined, lost (reason required). Vocabulary: LEAD_STATUSES in packages/domain/src/statuses.ts.
Research job
Changed in spec 1.1.0
The scraper guideline is no longer pending: research was ported from Reroot's client research (see Research guide and ADR 0009), and it now runs for clients as well as leads.
- Runs for a lead (triggered on form submit once step 4 exists, and by the owner on demand) and for a client (on conversion and on demand). Same pipeline, same brief.
- Input is the business name and website, plus area and category when known. With no website it searches by name.
- Goal: the owner knows the business (what it sells, who to, its tone, how it takes customers), its social accounts and videos, and its market: the best creators in that niche and area, similar businesses, and competitors, with their social presence.
- Websites are read with Cloudflare Browser Rendering; the model (the Claude app in development, the Claude Platform in production) searches for the rest.
- Tables:
research_job(subject, status, progress, error, provider, model, timestamps) andresearch_brief(versioned JSON). A "Research" panel on the lead page and on the client shows status and results. - Never invent figures. If a source fails, the brief says so under risks. Follower and view counts are kept only when they appear on a page that was read.