[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_509":510,"_511":512},"routes/_app.x.$handle.$name",{"_104":103,"_37":50,"_112":94,"_32":113,"_13":14,"_114":-5,"_11":115,"_116":117,"_118":119,"_120":121,"_122":123,"_26":-5,"_124":-5,"_125":-5,"_126":30,"_127":128,"_129":64,"_130":-5,"_131":-5,"_132":133},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":18,"_19":-5},"project",{"_48":49,"_37":50,"_43":51,"_52":-5,"_53":30,"_54":55,"_56":55,"_57":58,"_59":-5,"_19":60,"_61":62,"_63":64,"_65":-5,"_66":-5,"_67":68,"_69":70,"_71":72,"_73":74,"_75":76,"_77":-5,"_78":79,"_80":79,"_81":82,"_83":82,"_84":82,"_85":79,"_86":79,"_87":88,"_89":90,"_91":92,"_93":30},"version",88,"changes",[25],"aiChatInfo",{"_20":21,"_22":-5,"_23":-5,"_24":-5},"userId","aiChatMessageId","48cbf5c7-0db6-45ad-8f11-55a3f273ab66","aiChatThreadId","aiChatThreadUserId","aiChatThreadTitle",{"_26":27,"_28":-7,"_29":30,"_31":30,"_32":33,"_34":33},"file",{"_35":36,"_37":38,"_39":40,"_41":42,"_43":44,"_45":14,"_46":47,"_29":30},"previous","deleted",false,"deletedHere","path","backend/stage_trace.ts","latestPath","fileId","13e57334-a823-11f1-97b1-1607ee4eb77e","name","stage_trace.ts","fileType","script","content","// backend/stage_trace.ts\n// STAGE TRACE — the audit trail for the FIRST REAL INQUIRY proof.\n//\n// Every stage a real inquiry passes through is recorded as a FIXED, immutable\n// audit event. It separates FACT from INFERENCE explicitly, records the actual\n// outcome (never \"success\" on failure), and persists as provenance — so a human\n// can open an inquiry and see precisely what the system knew, inferred, and did.\n//\n// FIELD SEMANTICS (non-negotiable):\n// fact - something observed from the real source (verbatim input)\n// inference - the system's reasoning ABOUT the fact (clearly labeled)\n// unverified - labeled UNKNOWN / the system could not confirm\n// status - the actual result (\"ok\" or \"error\") — never \"ok\" on failure\nimport { sqlite } from \"https://esm.town/v/std/sqlite/main.ts\";\n\nexport interface StageEvent {\n stage: string;\n status: \"ok\" | \"error\";\n kind: \"fact\" | \"inference\" | \"unverified\" | \"system\";\n label: string; // human-readable name of the stage\n details: string[]; // ordered facts / inferences / failures\n at: string; // ISO timestamp\n}\n\nasync function ensureTable(): Promise {\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS dlx_stage_traces (\n id TEXT PRIMARY KEY,\n trace_id TEXT NOT NULL,\n events TEXT NOT NULL,\n created TEXT NOT NULL DEFAULT (datetime('now'))\n )`);\n}\n\nexport async function writeTrace(traceId: string, events: StageEvent[]): Promise {\n await ensureTable();\n await sqlite.execute({\n sql: `INSERT OR REPLACE INTO dlx_stage_traces (id, trace_id, events, created)\n VALUES (?, ?, ?, datetime('now'))`,\n args: [crypto.randomUUID(), traceId, JSON.stringify(events)],\n });\n}\n\nexport async function getTrace(traceId: string): Promise {\n await ensureTable();\n const res = await sqlite.execute(\n `SELECT events FROM dlx_stage_traces WHERE trace_id = ? ORDER BY created DESC LIMIT 1`,\n [traceId],\n );\n const row = res.rows[0] as { events?: string } | undefined;\n if (!row?.events) return null;\n try {\n return JSON.parse(row.events) as StageEvent[];\n } catch {\n return null;\n }\n}\n","createdAt",["D",1788500403600],"branchVersion","branchId","faea9245-6732-47d9-a1a2-3550998bc296","id","d2f8de76-3d02-4949-8a06-2f7cc759dae3","dlx-lead-engine",["D",1788452396724],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","DLX Designs lead qualification engine + React dashboard","imageUrl","0e434874-725d-4e49-bc72-a0c364fc943f","pinnedFileId","01a06811-df40-730d-85ee-1ae96a373bd7","pinPreview",true,"pinMigrated","screenshotVersion","likeCount",0,"orgId","01a04e8e-cc2d-73ad-9a1b-036254519f63","tursoName","d2f8de763d0249498a062f7cc759dae3","tursoHostname","d2f8de763d0249498a062f7cc759dae3-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789494036282],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_104":103,"_111":-5},"org",{"_104":103,"_105":106,"_107":-5,"_19":-5,"_108":109,"_110":30},"branch",{"_13":14,"_37":94,"_48":47,"_95":49,"_96":-5,"_97":-5,"_98":-5,"_99":-5,"_100":-5,"_101":50,"_102":103},"isOwned","main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","dlxdesigns","handle","displayName","DLXdesigns ","avatar","tier","free","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_48":49,"_37":50,"_43":504,"_52":-5,"_53":30,"_54":55,"_56":55,"_57":58,"_59":-5,"_19":60,"_61":62,"_63":64,"_65":-5,"_66":-5,"_67":68,"_69":70,"_71":72,"_73":74,"_75":505,"_77":-5,"_78":79,"_80":79,"_81":82,"_83":82,"_84":82,"_85":79,"_86":79,"_87":506,"_89":507,"_91":508,"_93":30},"tree",[230,241,244,246,249,251,260,262,265,266,267],"files",[230,231,232,233,234,235,236,237,238,239,240,241,242,243,244,245,246,247,248,249,250,251,252,253,254,255,256,257,258,259,260,261,262,263,264,265,266,267],"counts",{"_220":68,"_221":168,"_222":68,"_223":224,"_225":226},"contents",[134,135,136,137,138,139,140,141,142,143,144],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},{"_48":214,"_37":215,"_35":216,"_147":-5,"_46":47,"_45":217,"_149":64,"_43":218,"_13":68,"_39":177,"_41":-5,"_29":30,"_154":-5,"_32":219},{"_48":208,"_37":209,"_35":210,"_147":-5,"_46":47,"_45":211,"_149":64,"_43":212,"_13":68,"_39":177,"_41":-5,"_29":30,"_154":-5,"_32":213},{"_48":202,"_37":203,"_35":204,"_147":-5,"_46":47,"_45":205,"_149":64,"_43":206,"_13":68,"_39":177,"_41":-5,"_29":30,"_154":-5,"_32":207},{"_48":196,"_37":197,"_35":198,"_147":-5,"_46":47,"_45":199,"_149":64,"_43":200,"_13":68,"_39":177,"_41":-5,"_29":30,"_154":-5,"_32":201},{"_48":190,"_37":191,"_35":192,"_147":-5,"_46":47,"_45":193,"_149":64,"_43":194,"_13":68,"_39":177,"_41":-5,"_29":30,"_154":-5,"_32":195},{"_48":185,"_37":186,"_35":187,"_147":-5,"_46":47,"_45":168,"_149":64,"_43":188,"_13":68,"_39":177,"_41":-5,"_29":30,"_154":-5,"_32":189},{"_48":179,"_37":180,"_35":181,"_147":-5,"_46":47,"_45":182,"_149":64,"_43":183,"_13":68,"_39":177,"_41":-5,"_29":30,"_154":-5,"_32":184},{"_48":172,"_37":173,"_35":174,"_147":-5,"_46":47,"_45":175,"_149":64,"_43":176,"_13":68,"_39":177,"_41":-5,"_29":30,"_154":-5,"_32":178},{"_48":165,"_37":166,"_35":167,"_147":-5,"_46":47,"_45":168,"_149":64,"_43":169,"_13":68,"_39":26,"_41":170,"_29":30,"_154":-5,"_32":171},{"_48":157,"_37":158,"_35":159,"_147":-5,"_46":47,"_45":160,"_149":64,"_43":161,"_13":162,"_39":26,"_41":163,"_29":30,"_154":-5,"_32":164},{"_48":145,"_37":146,"_35":62,"_147":-5,"_46":47,"_45":148,"_149":64,"_43":150,"_13":151,"_39":152,"_41":153,"_29":30,"_154":155,"_32":156},"47858962-a30d-4978-b3f8-7982cdfc2bad","index.ts","parentId",86,"latest",["D",1788499627158],10,"http","import { parseVal, serveFile, serveImmutableFile } from \"https://esm.town/v/std/utils/index.ts\";\nimport { Hono } from \"npm:hono\";\nimport { Root } from \"./frontend/root.tsx\";\nimport {\n listLeads,\n setLeadStatus,\n stageDistribution,\n topSources,\n topSignals,\n} from \"./lib/store.ts\";\nimport { ingestLead } from \"./backend/worker/ingest.ts\";\nimport { nextStage } from \"./backend/pipeline.ts\";\nimport {\n verifyWebhook,\n handleEvent,\n listSignals,\n promoteSignal,\n} from \"./backend/instagram.ts\";\nimport {\n addProspect,\n listProspects,\n contactProspect,\n convertProspect,\n} from \"./backend/outbound.ts\";\n\nconst app = new Hono();\n\n// Serve the (never-cached) HTML shell at the root and as the dashboard\napp.get(\"/\", (c) => c.html(Root()));\napp.get(\"/dashboard\", (c) => c.html(Root()));\n\n// Standalone, embeddable lead-capture form (Option A: website ingestion)\napp.get(\"/capture\", (c) => serveFile(\"/frontend/capture.html\"));\n\n// Serve frontend files at the version-addressed URLs Root() stamps, cached immutably\napp.get(\"/__immutable/*\", (c) => serveImmutableFile(c.req.path));\n\n// View source redirect\napp.get(\"/source\", (c) => c.redirect(parseVal().links.self.val));\n\n// --- Lead API -----------------------------------------------------------\n\n// GET verified (already-qualified) leads for the dashboard\napp.get(\"/api/leads\", async (c) => {\n const leads = await listLeads(100);\n return c.json(leads);\n});\n\n// POST an inbound lead: { name, gmail, message, source }\n// Runs the strict qualification engine. Only high-intent leads are stored;\n// budget seekers and duplicates are rejected (422) and never persisted.\napp.post(\"/api/ingest\", async (c) => {\n const body = await c.req.json().catch(() => undefined);\n const outcome = await ingestLead((body as object) ?? {});\n return c.json(outcome, outcome.decision === \"reject\" ? 422 : 201);\n});\n\n// Move a lead forward in the pipeline (or set any valid stage).\n// Automatically advances when no stage body is given (\"next\").\napp.post(\"/api/leads/:id/status\", async (c) => {\n const id = c.req.param(\"id\");\n const body = (await c.req.json().catch(() => undefined)) as\n | { stage?: unknown; note?: unknown }\n | undefined;\n const lead = await listLeads(200);\n const existing = lead.find((l) => l.id === id);\n if (!existing) return c.json({ error: \"lead not found\" }, 404);\n\n let target: string;\n if (body?.stage) {\n target = String(body.stage);\n } else {\n const nxt = nextStage(existing.status as never);\n if (!nxt) return c.json({ error: \"terminal stage\", lead: existing });\n target = nxt;\n }\n\n const updated = await setLeadStatus(id, target, String(body?.note ?? \"\"));\n return c.json(updated ?? { error: \"lead not found\" }, updated ? 200 : 404);\n});\n\n// Analytics — what actually converts (learn layer).\napp.get(\"/api/analytics\", async (c) => {\n const [stages, sources, signals] = await Promise.all([\n stageDistribution(),\n topSources(8),\n topSignals(8),\n ]);\n return c.json({ stages, sources, signals });\n});\n\n// --- Instagram webhook (Meta Business app required) -----------------------\n// GET: Meta's subscription verification handshake.\napp.get(\"/api/instagram/webhook\", async (c) => {\n const challenge = await verifyWebhook(c.req.query());\n if (challenge === null) return c.text(\"verification failed\", 403);\n return c.text(challenge);\n});\n\n// POST: Meta pushes real DM/comment events here.\napp.post(\"/api/instagram/webhook\", async (c) => {\n const body = await c.req.json().catch(() => null);\n if (!body) return c.json({ error: \"invalid payload\" }, 400);\n const stored = await handleEvent(body);\n return c.json({ ok: true, captured: stored });\n});\n\n// List captured Instagram signals (real messages, no fabricated contacts).\napp.get(\"/api/signals\", async (c) => c.json(await listSignals(50)));\n\n// Promote a signal to a VERIFIED lead once the email is captured.\napp.post(\"/api/signals/:id/promote\", async (c) => {\n const id = c.req.param(\"id\");\n const body = (await c.req.json().catch(() => undefined)) as\n | { email?: unknown; phone?: unknown }\n | undefined;\n const email = String(body?.email ?? \"\").trim();\n if (!email) return c.json({ ok: false, reason: \"email required\" }, 400);\n const out = await promoteSignal(id, email, String(body?.phone ?? \"\").trim());\n return c.json(out, out.ok ? 200 : 422);\n});\n\n// --- Prospecting automaton (Refine -> Act) -------------------------------\n\n// Add a REAL public signal as a prospect (handle + message, no fabricated contact).\napp.post(\"/api/prospects\", async (c) => {\n const body = (await c.req.json().catch(() => undefined)) as\n | { handle?: unknown; name?: unknown; email?: unknown; phone?: unknown; message?: unknown; source?: unknown; source_url?: unknown }\n | undefined;\n if (!body?.handle || !body?.message) {\n return c.json({ ok: false, reason: \"handle and message are required\" }, 400);\n }\n const p = await addProspect({\n handle: String(body.handle),\n name: body.name ? String(body.name) : undefined,\n email: body.email ? String(body.email) : undefined,\n phone: body.phone ? String(body.phone) : undefined,\n message: String(body.message),\n source: body.source ? String(body.source) : undefined,\n source_url: body.source_url ? String(body.source_url) : undefined,\n });\n return c.json({ ok: true, prospect: p }, 201);\n});\n\n// List the prospect queue, strongest intent first.\napp.get(\"/api/prospects\", async (c) => c.json(await listProspects(50)));\n\n// Auto-draft + (optionally) send the first touch. Dry-run by default.\napp.post(\"/api/prospects/:id/contact\", async (c) => {\n const id = c.req.param(\"id\");\n const out = await contactProspect(id);\n return c.json(out, out.ok ? 200 : 422);\n});\n\n// Confirm the contact for a prospect -> becomes a VERIFIED pipeline lead.\napp.post(\"/api/prospects/:id/convert\", async (c) => {\n const id = c.req.param(\"id\");\n const body = (await c.req.json().catch(() => undefined)) as\n | { email?: unknown; phone?: unknown }\n | undefined;\n const email = String(body?.email ?? \"\").trim();\n if (!email) return c.json({ ok: false, reason: \"email required\" }, 400);\n const out = await convertProspect(id, email, String(body?.phone ?? \"\").trim());\n return c.json(out, out.ok ? 200 : 422);\n});\n\n// --- Google Business Profile intake ---------------------------------------\n// Real GBP-originated inquiry (review, Q&A, message a VA/owner copies in).\n// Becomes a scored prospect in the Command Center, source-tagged, ready to\n// draft + convert. Coordinates: reviewer name/handle, message, optional contact.\napp.post(\"/api/gbp\", async (c) => {\n const body = (await c.req.json().catch(() => undefined)) as\n | { reviewer?: unknown; name?: unknown; email?: unknown; phone?: unknown; message?: unknown; star?: unknown; permalink?: unknown }\n | undefined;\n const message = String(body?.message ?? \"\").trim();\n if (!message) {\n return c.json({ ok: false, reason: \"message is required (the actual review/Q&A text)\" }, 400);\n }\n const reviewer = String(body?.reviewer ?? body?.name ?? \"\").trim() || \"gbp-reviewer\";\n const p = await addProspect({\n handle: reviewer,\n name: body?.name && String(body.name) !== reviewer ? String(body.name) : reviewer,\n email: body?.email ? String(body.email) : undefined,\n phone: body?.phone ? String(body.phone) : undefined,\n message,\n source: \"google-business-profile\",\n source_url: body?.permalink ? String(body.permalink) : undefined,\n });\n return c.json({ ok: true, prospect: p }, 201);\n});\n\n// Unwrap and rethrow Hono errors as the original error\n// If you delete this, all errors will be \"Internal Server Error\"\napp.onError((err) => Promise.reject(err));\n\n// HTTP vals expect an exported \"fetch handler\"\n// This is how you \"run the server\" in Val Town with Hono\nexport default app.fetch;\n","label","dlx-leads",[146],"45371fdf-4ad7-4045-9c9f-a1f064977147","README.md","01a06811-df40-730d-85ee-161b7548ebf9",87,["D",1788499645589],24,"# DLX Designs · Verified Lead Intelligence Engine\n\nProduction-ready lead-generation and **strict qualification** engine for DLX\nDesigns — a luxury bespoke traditional menswear brand targeting high-net-worth\nclients, grooms and corporate executives in Lagos & Abeokuta, Nigeria.\n\nIt ingests real inbound traffic, filters out low-budget seekers with strict\nsemantic rules, stores only verified high-intent leads, and renders them as\nscored lead cards in a live dashboard.\n\n> **Live dashboard:** [Try the app](url:index.ts)\n> **Ingest endpoint:** `POST /api/ingest`\n\n## Architecture\n\n```mermaid\ngraph LR\n Inbound[\"Web form / corridor
(blog, IG, ads)\"] -->|\"POST /api/ingest\"| Qual[\"Qualifier
backend/worker/qualifier.ts\"]\n Qual -->|\"hard reject\"| X[\"Low budget → dropped
never stored\"]\n Qual -->|\"optional\"| LLM[\"AI refine
(free tier, optional)\"]\n Qual -->|\"pass ≥ 50\"| Store[\"SQLite store
lib/store.ts\"]\n Store -->|\"optional mirror\"| Supabase[\"Supabase
dlx_verified_leads\"]\n Store -->|\"GET /api/leads\"| UI[\"React dashboard
frontend/\"]\n```\n\n- **Dashboard:** client-side React + Hono backend (App Router-equivalent), Twind/Tailwind styling.\n- **Database:** val-scoped PostgreSQL-compatible SQLite. **Supabase mirror optional.**\n- **Qualifier:** deterministic heuristic (zero-config) + optional free-tier AI refinement.\n\n## The strict qualification engine\n\nRuns in `backend/worker/qualifier.ts` and never fabricates data.\n\n**Layer 1 — hard disqualifiers (instant reject):** `cheap`, `affordable`,\n`under 20k`/`20k`, `discount`, `thrift`, `okirika`.\n\n**Layer 2 — scoring matrix (0–100):** baseline 25, high-tier signals add points —\n`bespoke` +18, `groom`/`chieftaincy` +20, `cashmere` +15, `wedding` +14,\n`3-piece` +12, `executive` +14, `lekki`/`victoria island` +10, `lagos`/`abeokuta`\n+6, etc. **Pass threshold: ≥ 50.** Sub-threshold leads are rejected.\n\n**Layer 3 — AI refinement (optional, zero-config by default):** the engine is\n**deterministic-first** — the strict heuristic above is authoritative and needs\n**no AI key, no external call, no cost**, so no provider can ever block it. If\nyou add a free-tier key, that provider refines the score and writes a natural\nbuying signal (tried in order: `GEMINI_API_KEY`, `GROQ_API_KEY`, `OPENAI_API_KEY`).\nA provider failing just falls back to the heuristic — never an error.\n\n### Endpoints\n\n| Method | Path | Purpose |\n| --- | --- | --- |\n| `GET` | `/` | Dashboard |\n| `GET` | `/dashboard` | Dashboard (alias) |\n| `GET` | `/api/leads` | Verified leads (JSON, strongest first) |\n| `POST` | `/api/ingest` | Ingest & qualify an inbound lead — 201 on pass, 422 on reject |\n| `POST` | `/api/leads/:id/status` | Advance pipeline stage (or set one) — records history |\n| `GET` | `/api/analytics` | Pipeline / source / signal analytics (the learn layer) |\n\n`POST /api/ingest` request body:\n\n```json\n{\n \"name\": \"Oluwatosin Adeyemi\",\n \"gmail\": \"tosin@example.com\",\n \"source\": \"nigerianmenswear.com\",\n \"message\": \"I need a bespoke 3-piece groom suit for my chieftaincy ceremony in Abeokuta\"\n}\n```\n\nOnly passing leads persist. Rejected and duplicate leads are returned (422) and\nnever stored.\n\n## The growth engine loop\n\nA verified lead is more than a row — it is a story with evidence:\n\n1. **Hunt** — `cron/hunt.ts` polls a real configured source (`DLX_HUNT_URL`)\n and feeds genuinely captured inquiries through the qualifier. It never\n invents people or signals; rows without evidence are ignored.\n2. **Verify** — every lead records where it came from (`source`, `source_url`),\n exactly what was said (`evidence`), and when (`created_at`).\n3. **Qualify** — strict scoring (0–100); only ≥ 50 passes.\n4. **Prioritize** — the dashboard sorts strongest first and shows live pipeline counts.\n5. **Recommend** — each lead carries a `action`, chosen from the actual signal:\n book consultation, contact to confirm budget, or nurture.\n6. **Track** — leads move through `new → qualified → contacted → responded →\n consultation → quoted → deposit_paid → ordered → won/lost`. Every transition\n is recorded in `history`.\n7. **Learn** — `/api/analytics` reports conversion by stage and top sources and\n signals that actually produce qualified opportunities.\n\n## Capture & feed real inquiries (get it live today)\n\n**Option A — website ingestion.** A ready, branded enquiry form is hosted at\n`/capture` (or embedded by iframing it). It posts each submission to\n`/api/ingest`, which qualifies and stores it. Direct API for a custom site:\n\n```http\nPOST https://dlx-leads.val.run/api/ingest\nContent-Type: application/json\n\n{ \"name\": \"…\", \"gmail\": \"…\", \"phone\": \"…\", \"message\": \"…\", \"source\": \"…\", \"source_url\": \"…\" }\n```\n\n**Option C — JSON / Google Sheet feed (autonomous, scheduled every 15 min).**\nThe hunt poller accepts a **published Google Sheet CSV** directly, no code:\n\n1. In your sheet, make row 1 a header that includes `gmail` (plus optional\n `name`, `phone`, `message`, `source`, `source_url`).\n2. **File → Share → Publish to web →** choose the sheet → format **CSV** → copy the link.\n3. Set that link as `DLX_HUNT_URL` (and `DLX_HUNT_TOKEN` if it's private/behind auth).\n\nEach run, the engine ingests every new row, qualifies it, and skips duplicates.\nRows without an email are ignored — nothing is invented.\n\n**Option B — email intake (live inbound hunting).** Forward any organic inquiry\n(DM, WhatsApp, web message, marketplace lead) to the engine's email address and\nit is qualified, evidence-tagged, and tracked automatically:\n\n```\ndlxdesigns-e7b272d2a81611f19be51607ee4eb77e@valtown.email\n```\n\nSee [`docs/growth-corridors.md`](docs/growth-corridors.md) for the **50-verified\ngrowth corridors** (marketplaces, fashion platforms, directories, fashion-week\nhubs, forums, and press) — each mapped to an acquisition, observation, or\npartnership role the engine can consume.\n\n## Connect Instagram (real DM/comment capture)\n\nThe engine side is built and tested. To connect Meta (one-time, your login):\n\n1. Create a **Meta for Developers app** (https://developers.facebook.com).\n2. Add the **Instagram product** and connect your **Instagram Business account**.\n3. Create a token with the `instagram_business_basic`, `instagram_business_manage_messages`,\n and `pages_messaging` permissions (a **long-lived User or Page token**).\n4. Set these on the val:\n 👉 Add INSTAGRAM_VERIFY_TOKEN here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=INSTAGRAM_VERIFY_TOKEN\n 👉 Add META_APP_SECRET here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=META_APP_SECRET\n5. In **Meta → Developer App → Webhooks**, subscribe the **Instagram** object to\n the callback URL:\n `https://dlx-leads.val.run/api/instagram/webhook`\n using your **INSTAGRAM_VERIFY_TOKEN** as the verify token.\n\nReal DMs/comments then push into the engine automatically. Because Instagram\nDMs carry no email, each becomes a **signal** (`/api/signals`); reply to capture\nthe email, then promote it to a verified lead with\n`POST /api/signals/:id/promote`.\n\n## Google Business Profile (real local-channel capture)\n\nReviews, Q&A, and \"do you make…\" messages on your Google Business Profile are\nhigh-intent local signals. Capture them in one call — they become scored\nprospects source-tagged `google-business-profile` in the Command Center, ready\nto draft and convert:\n\n```http\nPOST https://dlx-leads.val.run/api/gbp\nContent-Type: application/json\n\n{ \"reviewer\": \"Adaeze Okafor\", \"message\": \"Do you make bespoke 3-piece groom suits for Lagos weddings?\", \"email\": \"…\", \"phone\": \"…\", \"permalink\": \"https://maps.google.com/…\" }\n```\n\nManual route (no credentials): a VA or owner copies each new review/Q&A/message\nand posts it to `/api/gbp` (or the dashboard capture form with\n`source: google-business-profile`), and the automaton takes it from there.\nAutomated pull of your GBP inbox requires Google OAuth on your account — that\npart is yours to connect when ready; everything after capture is handled here.\n\n## Configuration (environment variables)\n\nEverything runs out of the box with **zero environment variables**. The\nqualification engine is fully deterministic. The following optional keys unlock\nthe AI refinement layer (free tiers are preferred — none require billing) and\nthe Supabase mirror:\n\n- **`GEMINI_API_KEY`** — free tier (Google AI Studio key, no card required). Recommended.\n- **`GROQ_API_KEY`** — free tier. Ignored unless Gemini is absent or fails.\n- **`OPENAI_API_KEY`** — requires active OpenAI billing; only used as a last resort.\n- **`SUPABASE_URL`** + **`SUPABASE_SERVICE_ROLE_KEY`** — mirror verified leads to\n your Supabase `dlx_verified_leads` table (native SQLite store stays authoritative).\n- **`DLX_HUNT_URL`** — JSON **or published-CSV** feed the hunt poller watches\n for genuine inquiries (JSON: `{ \"leads\": [ { name, gmail, phone, message, source, source_url } ] }`).\n- **`DLX_HUNT_TOKEN`** — optional bearer token sent to the hunt source (`Authorization: Bearer …`).\n- **`INSTAGRAM_VERIFY_TOKEN`** — random string you invent; Meta must echo it on webhook verification.\n- **`META_APP_SECRET`** — optional; enables `X-Hub-Signature-256` verification on Instagram webhooks.\n- **`DLX_BRAND_NAME`** — whitelabel brand shown in capture copy + drafts (default: `DLX Designs`).\n- **`ENABLE_AUTOSEND`** — set to `1` to let the nurture automaton actually email first touches; OFF (dry-run) by default.\n\n👉 Add GEMINI_API_KEY here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=GEMINI_API_KEY\n👉 Add GROQ_API_KEY here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=GROQ_API_KEY\n👉 Add OPENAI_API_KEY here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=OPENAI_API_KEY\n👉 Add SUPABASE_URL here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=SUPABASE_URL\n👉 Add SUPABASE_SERVICE_ROLE_KEY here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=SUPABASE_SERVICE_ROLE_KEY\n👉 Add DLX_HUNT_URL here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=DLX_HUNT_URL\n👉 Add DLX_HUNT_TOKEN here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=DLX_HUNT_TOKEN\n👉 Add INSTAGRAM_VERIFY_TOKEN here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=INSTAGRAM_VERIFY_TOKEN\n👉 Add META_APP_SECRET here: https://www.val.town/x/dlxdesigns/dlx-lead-engine/environment-variables?key=META_APP_SECRET\n\n### Supabase setup (optional)\n\nCreate the mirror table by running [`database/schema.sql`](database/schema.sql)\nin the Supabase SQL editor. The table uses `gen_random_uuid()` ids, enforces a\n`score BETWEEN 0 AND 100` check, has a `gmail UNIQUE` constraint, and enables\nRow-Level Security (locked to service-role access by default). The val writes\nvia PostgREST with the service-role key and `resolution=merge-duplicates`.\n\n## File structure\n\n```\ndatabase/\n schema.sql ← Supabase/PostgreSQL schema (mirror table)\nbackend/\n config.ts ← whitelabel brand + runtime config\n pipeline.ts ← sales pipeline stages + next-stage logic\n hunt.ts ← autonomous ingestion from a real configured source\n outbound.ts ← prospect queue: score, auto-draft, guarded send, convert\n instagram.ts ← Meta webhook receiver for real IG DMs/comments\n worker/\n qualifier.ts ← strict qualification engine (disqualifiers + scoring)\n ingest.ts ← ingest worker (bridge inbound → store)\n llm.ts ← optional AI adapter (free tier Gemini / Groq / OpenAI)\ncron/\n hunt.ts ← scheduled hunt poller (every 15 min)\n automate.ts ← nurture automaton: nudge hot leads + auto-send (guarded)\nemail/\n lead-intake.ts ← email-triggered lead intake (forward inquiries here)\ndocs/\n growth-corridors.md ← 50 acquisition/observation/partnership corridors\n operating-model.md ← how the automaton runs + per-client deployment model\nlib/\n store.ts ← SQLite persistence + Supabase mirror + pipeline + analytics\nfrontend/\n components/App.tsx ← dashboard layout + empty state + pipeline strip\n components/LeadCard.tsx ← verified-lead card with evidence, action & stage\n components/LeadForm.tsx ← inline lead-capture form\nscripts/\n selftest.ts ← end-to-end qualifier + store self-test\nindex.ts ← Hono HTTP backend (dashboard + /api/*)\n```\n\n## Integrity guarantees\n\n- **No hardcoded, fake, or mock contacts anywhere.** Every record comes from a\n real `POST /api/ingest` submission that passed strict qualification.\n- Empty store → dashboard shows *“Awaiting live inbound high-intent signals…”*,\n never fake cards.\n- Duplicate emails are rejected (UNIQUE), budget seekers are dropped, sub-50\n intent is dropped.\n\n## Running the self-test\n\n`scripts/selftest.ts` exercises the full pipeline with synthetic example inputs\n(high-intent pass, low-budget reject, no-signal reject, duplicate reject) and is\nsafe to run repeatedly.\n",[158],"a130fa68-e89b-4e79-b351-ae119f51052e","CHANGELOG.md","01a06811-df40-730d-85ee-12df63d9ba38",1,["D",1788452396724],"# Changelog\n\n## 2026-07-30 — Immutable asset caching\n\nFrontend files are now served at version-addressed URLs\n(`/__immutable/45/frontend/index.tsx`) with\n`Cache-Control: public, max-age=31536000, immutable`, so repeat visits load the\nwhole client graph from the browser cache — roughly 3× faster (~665ms → ~157ms\nmeasured). Publishing your val bumps its version, which changes every URL, so\ncaches invalidate automatically. Old-version URLs return 404 after a publish\n(like Next.js build assets); a reload picks up the new version.\n\nWhat changed in the template:\n\n- `frontend/index.html` → [`frontend/root.tsx`](/frontend/root.tsx): the HTML\n shell is now JSX and stamps asset URLs with `immutableFileUrl()`\n- [`index.ts`](/index.ts): serves `Root()` at `/`, and one\n `app.get(\"/__immutable/*\", (c) => serveImmutableFile(c.req.path))` route\n serves all frontend files\n\n### Upgrading an existing remix\n\nYour copy of this template doesn't update automatically. Two options:\n\n**Minimal (most of the benefit):** point both routes at `serveImmutableFile` —\nyour existing bare URLs redirect into versioned space, the new route serves\nthem, and the whole module graph gets cached (costs one redirect per page view):\n\n```ts\nimport { serveImmutableFile } from \"https://esm.town/v/std/utils/index.ts\";\n\napp.get(\"/__immutable/*\", (c) => serveImmutableFile(c.req.path));\napp.get(\"/frontend/**/*\", (c) => serveImmutableFile(c.req.path)); // was serveFile\n```\n\n**Full upgrade (also removes the redirect):** three self-contained steps —\nno other files needed. Adjust the ``, favicon, and body of `Root()` to\nmatch your app's existing `index.html`, and stamp anything else your shell\nreferences directly (images, css) with `immutableFileUrl()`.\n\n1. Create `frontend/root.tsx`:\n\n```tsx\n/** @jsxImportSource npm:hono@4/jsx */\nimport { raw } from \"npm:hono@4/html\";\nimport { immutableFileUrl } from \"https://esm.town/v/std/utils/index.ts\";\n\n// The HTML shell for the app. It is never cached, so immutableFileUrl can stamp\n// the current val version into each asset URL — the browser caches those\n// immutably, and publishing the val changes the URLs, refetching exactly once.\nexport function Root() {\n return (\n <>\n {raw(\"<!DOCTYPE html>\")}\n <html lang=\"en\">\n <head>\n <meta charSet=\"UTF-8\" />\n <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\" />\n <title>React Hono Val Town Starter\n \n \n \n \n

DLX Designs

Bespoke menswear enquiry

Tell us what you need. Verified inquiries are handled personally.

\n\n \n\n

DLX Designs · Lagos & Abeokuta

\n\n \n \n\n",[],[186,329],"909f573f-b663-4da1-9e37-58f6e7fbf8fb","LeadForm.tsx","bd44cf7c-a7b7-11f1-bb36-1607ee4eb77e","01a06811-df3f-7558-9972-4e3e2fdb144e",16,["D",1788454302131],"/** @jsxImportSource https://esm.sh/react@18.2.0 */\n// frontend/components/LeadForm.tsx\n// Inline lead-capture form. Posts to /api/ingest, which runs the strict\n// qualifier. Shows pass/reject feedback inline.\nimport { useState, type FormEvent } from \"https://esm.sh/react@18.2.0\";\n\ntype SubmitState =\n | { kind: \"idle\" }\n | { kind: \"sending\" }\n | { kind: \"ok\"; message: string }\n | { kind: \"err\"; message: string };\n\nexport function LeadForm({ onCaptured }: { onCaptured: () => void }) {\n const [name, setName] = useState(\"\");\n const [gmail, setGmail] = useState(\"\");\n const [source, setSource] = useState(\"\");\n const [message, setMessage] = useState(\"\");\n const [state, setState] = useState({ kind: \"idle\" });\n\n async function submit(e: FormEvent) {\n e.preventDefault();\n setState({ kind: \"sending\" });\n try {\n const res = await fetch(\"/api/ingest\", {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({ name, gmail, source, message }),\n });\n const data = await res.json();\n if (data.decision === \"pass\") {\n setState({ kind: \"ok\", message: `Verified lead stored · ${data.lead?.score}` });\n setName(\"\");\n setGmail(\"\");\n setSource(\"\");\n setMessage(\"\");\n onCaptured();\n } else {\n setState({ kind: \"err\", message: `Not qualified: ${data.reason ?? \"rejected\"}` });\n }\n } catch (err) {\n setState({ kind: \"err\", message: String(err instanceof Error ? err.message : err) });\n }\n }\n\n const inputCls =\n \"w-full rounded-lg border border-stone-300 bg-white px-3 py-2 text-sm text-stone-800 placeholder:text-stone-400 focus:outline-none focus:ring-2 focus:ring-amber-500\";\n\n return (\n

Capture an inbound lead

\n POST /api/ingest\n

void submit(e)} class=\"grid grid-cols-1 sm:grid-cols-2 gap-3\">\n setName(e.target.value)}\n placeholder=\"Client name\"\n required\n />\n setGmail(e.target.value)}\n placeholder=\"Email address\"\n required\n />\n setSource(e.target.value)}\n placeholder=\"Source / domain (e.g. nigerianmenswear.com)\"\n />\n setMessage(e.target.value)}\n placeholder='What they asked for, e.g. \"bespoke 3-piece groom suit\"'\n required\n />\n

\n {state.kind === \"ok\" && (\n {state.message}\n )}\n {state.kind === \"err\" && (\n {state.message}\n )}\n \n {state.kind === \"sending\" ? \"Qualifying…\" : \"Capture lead\"}\n \n

\n \n

\n );\n}\n",[],[186,345,337],"components","f016accd-d506-4618-b8dd-f0f37419a6b2","LeadCard.tsx","825db1fc-a7b3-11f1-a292-1607ee4eb77e",40,["D",1788493743320],"/** @jsxImportSource https://esm.sh/react@18.2.0 */\n// frontend/components/LeadCard.tsx\n// A single verified-lead card: identity, evidence, signal, recommended action,\n// and pipeline stage controls. Nothing here is ever invented — it renders only\n// what the store captured.\nimport { useState } from \"https://esm.sh/react@18.2.0\";\n\nexport interface StageEvent {\n from: string;\n to: string;\n at: string;\n note?: string;\n}\n\nexport interface Lead {\n id: string;\n name: string;\n gmail: string;\n phone: string;\n score: number;\n source: string;\n source_url: string;\n evidence: string;\n signal: string;\n action: string;\n status: string;\n history: StageEvent[];\n created_at: string;\n updated_at: string;\n}\n\nconst STAGE_LABELS: Record = {\n new: \"New\", qualified: \"Qualified\", contacted: \"Contacted\",\n responded: \"Responded\", consultation: \"Consultation\", quoted: \"Quoted\",\n deposit_paid: \"Deposit Paid\", ordered: \"Ordered\", won: \"Won\", lost: \"Lost\",\n};\n\nfunction scoreNumber(n: number): number {\n return Number.isFinite(n) ? Math.min(100, Math.max(0, Math.round(n))) : 0;\n}\n\nfunction badgeClass(score: number): string {\n const s = scoreNumber(score);\n if (s >= 80) return \"bg-emerald-100 text-emerald-800 border-emerald-300\";\n if (s >= 65) return \"bg-teal-100 text-teal-800 border-teal-300\";\n if (s >= 50) return \"bg-amber-100 text-amber-800 border-amber-300\";\n return \"bg-gray-100 text-gray-600 border-gray-300\";\n}\n\nfunction initials(name: string): string {\n return name.split(/\\s+/).filter(Boolean).slice(0, 2)\n .map((p) => p[0]?.toUpperCase() ?? \"\").join(\"\") || \"DLX\";\n}\n\nexport function LeadCard({ lead, onAdvance }: { lead: Lead; onAdvance: (id: string) => void }) {\n const s = scoreNumber(lead.score);\n const terminal = lead.status === \"won\" || lead.status === \"lost\";\n const [busy, setBusy] = useState(false);\n const date = new Date(lead.created_at);\n const when = Number.isNaN(date.getTime())\n ? \"\"\n : date.toLocaleDateString(undefined, { day: \"numeric\", month: \"short\" });\n\n return (\n

\n {/* Identity + score */}\n

\n {initials(lead.name)}\n

{lead.name}

{lead.gmail}

\n {lead.phone && (\n

📱 {lead.phone}

\n )}\n

\n \n {s}/100%\n \n

\n\n {/* Source + stage */}\n

\n \n {lead.source_url ? (\n \n 🌐 {lead.source || \"link\"}\n \n ) : (\n <>🌐 {lead.source || \"inbound-form\"}\n )}\n \n {when && (\n {when}\n )}\n \n {STAGE_LABELS[lead.status] ?? lead.status}\n \n

\n\n {/* Evidence — what was actually said */}\n {lead.evidence && (\n

\n Evidence\n

{lead.evidence}

\n )}\n\n {/* Signal + recommended action */}\n

Signal: {lead.signal || \"High-intent luxury menswear inquiry\"}

\n {lead.action && (\n

\n Next: {lead.action}\n

\n )}\n

\n\n {/* Pipeline control */}\n {!terminal ? (\n {\n setBusy(true);\n onAdvance(lead.id);\n setTimeout(() => setBusy(false), 800);\n }}\n disabled={busy}\n class=\"mt-1 rounded-lg border border-stone-300 hover:border-stone-900 hover:bg-stone-900 hover:text-white px-3 py-1.5 text-sm font-medium transition-colors cursor-pointer disabled:opacity-50\"\n >\n {busy ? \"Advancing…\" : \"Advance stage →\"}\n \n ) : (\n \n {lead.status === \"won\" ? \"🏆 Won\" : \"Closed / lost\"}\n \n )}\n

\n );\n}\n",[],[186,345,347],"dc8a6d30-e57d-434e-ae98-582947bd2713","App.tsx","01a06811-df3f-7558-9972-46f66ec4eab9",85,["D",1788499512196],9,"/** @jsxImportSource https://esm.sh/react@18.2.0 */\n// frontend/components/App.tsx\n// DLX Designs — AUTONOMOUS LEAD AUTOMATON · COMMAND CENTER\n// One cockpit: verified pipeline, prospect outbound queue, IG signals, and the\n// learn/automation view. Every object is real — nothing is fabricated.\nimport { useEffect, useState } from \"https://esm.sh/react@18.2.0\";\nimport { LeadCard, type Lead } from \"./LeadCard.tsx\";\nimport { LeadForm } from \"./LeadForm.tsx\";\n\ninterface Prospect {\n id: string; handle: string; name: string; email: string; phone: string;\n message: string; source: string; score: number; signal: string; action: string;\n status: string; created_at: string;\n}\ninterface IgSignal {\n id: string; ig_user: string; ig_id: string; message: string; permalink: string;\n source: string; status: string; created_at: string;\n}\ninterface Analytics { stages: { status: string; count: number }[]; sources: { source: string; count: number }[]; signals: { signal: string; count: number }[] }\n\nconst ORDER = [\n [\"new\", \"New\"], [\"qualified\", \"Qualified\"], [\"contacted\", \"Contacted\"],\n [\"responded\", \"Responded\"], [\"consultation\", \"Consultation\"], [\"quoted\", \"Quoted\"],\n [\"deposit_paid\", \"Deposit\"], [\"ordered\", \"Ordered\"], [\"won\", \"Won\"], [\"lost\", \"Lost\"],\n] as const;\n\ntype Tab = \"pipeline\" | \"prospects\" | \"signals\" | \"learn\";\n\nfunction EmptyPanel({ title, hint }: { title: string; hint: string }) {\n return (\n

📡

{title}

{hint}

\n );\n}\n\nasync function postJSON(url: string, body: unknown) {\n const res = await fetch(url, { method: \"POST\", headers: { \"Content-Type\": \"application/json\" }, body: JSON.stringify(body) });\n return res.json();\n}\n\nexport function App() {\n const [leads, setLeads] = useState(null);\n const [prospects, setProspects] = useState(null);\n const [signals, setSignals] = useState(null);\n const [analytics, setAnalytics] = useState(null);\n const [error, setError] = useState(null);\n const [loading, setLoading] = useState(true);\n const [tab, setTab] = useState(\"pipeline\");\n const [autosend, setAutosend] = useState(false);\n\n async function refresh() {\n setLoading(true);\n try {\n const [l, p, s, a] = await Promise.all([\n fetch(\"/api/leads\").then((r) => r.json()),\n fetch(\"/api/prospects\").then((r) => r.json()),\n fetch(\"/api/signals\").then((r) => r.json()),\n fetch(\"/api/analytics\").then((r) => r.json()),\n ]);\n setLeads(Array.isArray(l) ? l : []);\n setProspects(Array.isArray(p) ? p : []);\n setSignals(Array.isArray(s) ? s : []);\n setAnalytics(a);\n setError(null);\n } catch (e) {\n setError(String(e instanceof Error ? e.message : e));\n } finally {\n setLoading(false);\n }\n }\n\n useEffect(() => { void refresh(); }, []);\n\n async function advance(id: string) {\n await postJSON(`/api/leads/${id}/status`, {}).catch(() => {});\n await refresh();\n }\n\n async function draftContact(id: string) {\n const out = await postJSON(`/api/prospects/${id}/contact`, {}).catch(() => ({ ok: false }));\n if (out.ok && out.draft) {\n window.alert(`DRAFT READY (${out.reason})\\n\\nSUBJECT: ${out.draft.subject}\\n\\n${out.draft.body}`);\n } else {\n window.alert(\"Could not draft: \" + (out.reason ?? \"error\"));\n }\n await refresh();\n }\n\n async function convertProspect(id: string) {\n const email = window.prompt(\"Verified email for this prospect?\");\n if (!email) return;\n await postJSON(`/api/prospects/${id}/convert`, { email }).catch(() => {});\n await refresh();\n }\n\n async function promoteSignal(id: string) {\n const email = window.prompt(\"Email captured from reply?\");\n if (!email) return;\n await postJSON(`/api/signals/${id}/promote`, { email }).catch(() => {});\n await refresh();\n }\n\n const count = leads?.length ?? 0;\n const avg = leads && leads.length > 0\n ? Math.round(leads.reduce((a, l) => a + (l.score || 0), 0) / leads.length)\n : 0;\n const stagesMap = Object.fromEntries((analytics?.stages ?? []).map((s) => [s.status, s.count]));\n\n return (\n

DLX Designs

Autonomous Lead Automaton · Command Center

\n void refresh()}\n class=\"rounded-lg bg-amber-500 hover:bg-amber-400 text-stone-900 font-semibold px-3 py-1.5 text-xs transition-colors cursor-pointer disabled:opacity-50\"\n disabled={loading}\n >{loading ? \"Syncing…\" : \"⟳ Sync\"}\n

\n {/* Tabs */}\n \n

\n\n {/* Quick stats */}\n

Verified leads

{count}

Prospects (outbound)

{prospects?.length ?? 0}

IG signals

{signals?.length ?? 0}

Avg intent

{avg}/100

\n\n

\n {error &&

Sync error: {error}

}\n\n {tab === \"pipeline\" && (\n

\n void refresh()} />\n {leads && leads.length > 0 && (\n

\n Pipeline\n {ORDER.map(([key, label]) => (\n 0 ? \"bg-stone-900 text-stone-50 border-stone-900\" : \"bg-white text-stone-400 border-stone-200\"}`}>\n {label}: {stagesMap[key] ?? 0}\n \n ))}\n

\n )}\n {leads && leads.length === 0 ? (\n \n ) : (\n

\n {leads?.map((l) => void advance(id)} />)}\n

\n )}\n

\n )}\n\n {tab === \"prospects\" && (\n

Prospect outbound queue

\n Real public signals — score + auto-drafted pitch; convert needs a real email\n

\n {prospects && prospects.length === 0 ? (\n \n ) : (\n

\n {prospects?.map((p) => (\n

\n {p.handle}\n = 70 ? \"bg-emerald-100 text-emerald-800\" : p.score >= 50 ? \"bg-amber-100 text-amber-800\" : \"bg-stone-100 text-stone-500\"}`}>{p.score}/100\n {p.source}\n {p.status}\n

{p.message}

Next: {p.action}

\n \n \n

\n ))}\n

\n )}\n

\n )}\n\n {tab === \"signals\" && (\n

Instagram signals

\n Real DMs/comments via Meta webhook — reply to capture the email, then promote\n

\n {signals && signals.length === 0 ? (\n \n ) : (\n

\n {signals?.map((s) => (\n

\n @{s.ig_user}\n {s.source}\n {s.status}\n

{s.message}

\n {s.permalink && view source ↗}\n \n

\n ))}\n

\n )}\n

\n )}\n\n {tab === \"learn\" && (\n

Automatons

🕐 Hunt poller — every 15 min, pulls the configured real source
🕐 Nurture — every 15 min, flags + nudges stale hot leads; auto-sends first touches when enabled
📧 Email intake — instant qualification of forwarded inquiries
📸 Instagram webhook — receives real DMs/comments

\n \n { setAutosend(e.target.checked); }} />\n Show autosend intent (dry-run until ENABLE_AUTOSEND=1)\n \n

Top sources

\n {analytics?.sources.length ? (\n

{s.source} — {s.count}

\n ) :

No data yet.

}\n

Buying signals seen

\n {analytics?.signals.length ? (\n

{s.signal} — {s.count}

\n ) :

No data yet — the engine will show which signals actually convert.

}\n

\n )}\n\n \n

\n );\n}\n",[],[186,345,355],"c1571bc2-1eae-4a07-9531-058b3b6dd68d",["D",1788452396724],[253,254,255],[186,345],["D",1788452396724],[252,256,257,258,259],[186],"01029356-9c7a-410f-af74-6417c61c9a0c","lead-intake.ts","e7b272d2-a816-11f1-9be5-1607ee4eb77e",["D",1788495175441],"// email/lead-intake.ts\n// EMAIL INTAKE — the live inbound-hunting channel.\n//\n// Any email sent (or forwarded) to this val's address is run through the strict\n// qualifier and, if it passes, becomes a verified lead in the pipeline. This is\n// the closest thing to real-time inbound hunting: forward a DM/WhatsApp/web\n// inquiry to the address and it is qualified, evidence-tagged, and tracked\n// automatically — no spreadsheets, no forms.\n//\n// Find the assigned address in `links.email` on this file.\nimport { ingestLead } from \"../backend/worker/ingest.ts\";\n\nexport default async function (e: Email) {\n const from = e.from || \"\";\n const text = `${e.subject ?? \"\"}\\n\\n${e.text ?? \"\"}`.trim();\n\n // From: \"Ada Obi\" -> name + email\n const addressMatch = from.match(/<([^<>@\\s]+@[^<>@\\s]+)>/i);\n const bareEmail = (from.match(/\\S+@\\S+/) ?? [\"\"])[0];\n const gmail = (addressMatch?.[1] ?? bareEmail ?? \"\").toLowerCase().trim();\n const nameRaw = (from.split(\"<\")[0] || \"\")\n .replace(/[\"']/g, \"\")\n .replace(/^\\s*(From|mailto)?:?\\s*/i, \"\")\n .trim();\n\n // Phone: common Nigerian / international patterns\n const phoneMatch = text.match(\n /(?:\\+?234[\\s-]?)?[0]\\d{9,10}\\b|\\+234[\\s-]?\\d{3}[\\s-]?\\d{3}[\\s-]?\\d{4}/,\n );\n\n const outcome = await ingestLead({\n name: nameRaw || undefined,\n gmail: gmail || undefined,\n phone: phoneMatch?.[0]?.replace(/[\\s-]/g, \"\") || undefined,\n message: text.slice(0, 2000),\n source: \"email-intake\",\n source_url: from.match(/<([^>]+)>/)?.[1] || e.from || \"\",\n });\n\n console.log(\n `email-intake <- ${gmail} => ${outcome.decision} (${\n (outcome as { reason?: string }).reason ?? \"\"\n })`,\n );\n}\n",[],[191,371],["D",1788495175441],[250],[191],"f8440b8e-37d8-4895-a036-e889d0785bdd","operating-model.md","b73fb074-a820-11f1-bc45-1607ee4eb77e",82,["D",1788499389189],"# DLX Designs — Autonomous Lead Automaton Engine · Operating Model\n\nA professional brief for how this system runs, and how to deploy it per client.\n\n## What this is\n\nNot a lead list. A **five-stage automaton** that turns real buying signals into\nqualified, tracked revenue opportunities — with every record carrying evidence\nand every stage handing a human its next step.\n\n```\nHUNT → REFINE → CAPTURE → ACT → LEARN → (feeds HUNT)\n```\n\n## The automaton, layer by layer\n\n| Layer | Where | What it does automatically |\n|---|---|---|\n| **HUNT** | `cron/hunt.ts`, `email/lead-intake.ts`, IG webhook | Watches a real source (sheet/JSON feed every 15 min), ingests forwarded email, and receives IG DMs/comments via Meta webhook. Real signals only. |\n| **REFINE** | `backend/outbound.ts` | Scores any real public signal (0–100), writes the signal + recommended action, and **auto-drafts a personalized first-touch** quoting the actual message. |\n| **CAPTURE** | `lib/store.ts` | Promotes a prospect/signal to a **verified lead** the moment a real email/phone is confirmed. Budget seekers are rejected; nothing is invented. |\n| **ACT** | `cron/automate.ts`, pipeline status API | Every 15 min: flags hot leads idle >24h, nudges them to the work queue, and (only with `ENABLE_AUTOSEND=1`) sends that auto-drafted first touch to verified contacts. Pipeline moves `new → … → won/lost` with full history. |\n| **LEARN** | `/api/analytics` | Reports conversion by stage, source, and signal so budget flows to what actually produces revenue. |\n\n## The three scheduled crons\n\n| Cron | Frequency | Job |\n|---|---|---|\n| `cron/hunt.ts` | every 15 min | Pull the configured real source into the pipeline |\n| `cron/automate.ts` | every 15 min | Nudge stale hot leads + auto-send first touches (guarded) |\n| `email/lead-intake.ts` | on email | Qualified every forwarded inquiry instantly |\n\n## Security & integrity guarantees\n\n- **No invented information.** A lead needs a real contact; a signal without an\n email is helpfully drafted, never fabricated.\n- **Auto-send is OFF by default.** Set `ENABLE_AUTOSEND=1` to let the engine\n email. Otherwise it dry-runs for human review — responsible, not spam.\n- All sends and stage changes are **logged** so the operator sees what the\n machine did and why.\n\n## Deploying per client (you operate, the client owns channels)\n\nThis engine is whitelabeled via env config, so the **exact same code** powers a\nnew client without edits:\n\n1. **Remix this val** into the client's org (own store, env, crons, dashboard — full isolation per client).\n2. Set `DLX_BRAND_NAME` to the client's brand (used in capture copy + drafts).\n3. The **client** connects their own channels: their website form to\n `/api/ingest`, their Google Sheet to `DLX_HUNT_URL`, their Meta app to the IG\n webhook, and/or forwards DMs to their deployment's email intake address.\n4. You operate: monitor logs, tune disqualifiers/scoring in `qualifier.ts`,\n enable `ENABLE_AUTOSEND` when the client approves, and review `/api/analytics`.\n\nBecause each deployment is isolated, you can run many brands safely from one\nCodebase, and a client's data, channels, and secrets never mix.\n\n## Go-live checklist for DLX Designs (this deployment)\n\n- [ ] Connect a real source: Google Sheet CSV → set `DLX_HUNT_URL` (+ `DLX_HUNT_TOKEN` if private).\n- [ ] Point any site form at `POST /api/ingest` and/or embed `/capture`.\n- [ ] Publish `dlxdesigns-…@valtown.email` as an intake address for DMs/leads.\n- [ ] (Optional) Connect the Instagram/Meta webhook per the README guide.\n- [ ] (Optional) Add a free-tier `GEMINI_API_KEY` for AI-refined signals.\n- [ ] (Optional) Run `database/schema.sql` in Supabase to enable the mirror.\n- [ ] Review `/api/analytics` weekly; tune scoring in `qualifier.ts`.\n",[],[197,381],"d1e73a5d-e4c4-4e91-9abe-00a0aa4b833c","growth-corridors.md","044917de-a817-11f1-9be5-1607ee4eb77e",["D",1788495223471],"# DLX Designs · 50 Growth Corridors\n\nA professional framework for turning Nigerian marketplaces, fashion platforms,\ndirectories, and press into three things DLX actually needs:\n\n- **Acquisition** — post/sell where premium buyers browse.\n- **Observation** — watch where grooms, executives, and chieftaincies signal intent.\n- **Partnership** — list, feature, and collaborate with trusted local leaders.\n\nEvery corridor below maps to a **role the engine can play**, not a promise of\nany specific outcome. Use `source` (e.g. `jumia`, `bella-naija-weddings`,\n`nairaland`), route organic DMs/inquiries into the pipeline via `POST /api/ingest`\nor forward them to the email intake address, and let the qualifier + analytics\nshow which corridors actually produce qualified revenue.\n\n> **Honesty note:** verify each link/handle once before committing outreach —\n> marketplaces and directories change hands or go dormant. The ones marked\n> ⭐ major are large, long-running platforms. The rest are candidates to confirm.\n\n---\n\n## 1. General & Multi-Vendor Marketplaces — *Acquisition*\n| # | Corridor | DLX angle |\n|---|---|---|\n| 1 | **Jumia NG** ⭐ | Seller storefront + sponsored listings; Jumia's audience searches familiar Suits/Agbada |\n| 2 | **Konga** ⭐ | Multi-vendor storefront for premium categories |\n| 3 | **Jiji NG** ⭐ | Classifieds — high consideration, direct inquiry (feeds `/api/ingest`) |\n| 4 | **Temu NG** ⭐ | Benchmark price floors + observe category demand (not a premium fit) |\n| 5 | **PayPorte** | Storefront + style catalog for made-in-Nigeria fashion |\n| 6 | **Kusnap** | SME marketplace; Groomwear niche sellers |\n| 7 | **Sellevo** | Nigerian stores directory; local visibility |\n| 8 | **UPshop** | Local boutique discovery |\n| 9 | **AliExpress NG** ⭐ | Reference pricing/fabric categories; outbound compare |\n| 10 | **Nawao** | Commerce store; low volume but easy listing |\n\n## 2. Fashion, Apparel & Men's Style Platforms — *Acquisition + Observation*\n| # | Corridor | DLX angle |\n|---|---|---|\n| 11 | **Ajebomarket** | Men's fashion vertical listing |\n| 12 | **Garmspot** | Style store with buyer intent signals |\n| 13 | **Obeezi** | Men's wear focused — direct fit |\n| 14 | **The 5K Shop** | Retail portal for traditional wear |\n| 15 | **Miskay Boutique** | Boutique-style presentation for bespoke work |\n| 16 | **Sojoee** | Clothing store; test limited collections |\n| 17 | **Kaiglo** | E-commerce with Nigerian fashion categories |\n| 18 | **Bentigos** | Retail hub browse behavior observation |\n| 19 | **Mobos Fashion** | Apparel store breadth |\n| 20 | **Farfetch** ⭐ | Luxury positioning benchmark; not a volume channel |\n\n## 3. Directories & Artisan Networks — *Partnership + Acquisition*\n| # | Corridor | DLX angle |\n|---|---|---|\n| 21 | **BusinessList Lagos** | Premium listing under Fashion |\n| 22 | **TailorPro NG** | Artisan/made-to-measure directory — strong fit |\n| 23 | **VisCorner NG** | Tailoring (Lagos) — direct lead source |\n| 24 | **ArtisanOga** | Tailors directory; request/quote flow |\n| 25 | **Finelib Lagos** | Fashion designers index |\n| 26 | **Takooka** | Business portal (Fashion Lagos) |\n| 27 | **InfoisInfo Lagos** | Directory + review reputation |\n| 28 | **ConnectNigeria** ⭐ | Business listings, high SERP trust |\n| 29 | **VConnect** | Tailoring directory with callbacks |\n| 30 | **Sellatease** | Marketplace directory for exporters/sellers |\n\n## 4. Fashion Week & Industry Hubs — *Partnership + Authority*\n| # | Corridor | DLX angle |\n|---|---|---|\n| 31 | **Lagos Fashion Week** ⭐ | Showcase/network; authority association |\n| 32 | **Africa Fashion Week NG** | Designer participant/partner |\n| 33 | **Printivo** | Creative/merch partner for collateral |\n| 34 | **StyleHouse Files** | Editorial creative directory |\n| 35 | **NTMA** | Fabric sourcing ecosystem (partner for premium materials) |\n\n## 5. Classifieds, Forums & Community Boards — *Observation + Acquisition*\n| # | Corridor | DLX angle |\n|---|---|---|\n| 36 | **Nairaland — Business** ⭐ | Large Nigerian forum; watch demand, reply to genuine seekers |\n| 37 | **Nairaland — Fashion** ⭐ | Groom/agbada threads = real intent signals |\n| 38 | **Ogun State Business Directory** | Abeokuta regional coverage |\n| 39 | **Lagos MSME/Enterprise Portal** | Government-backed visibility + grants ecosystem |\n| 40 | **MAN** | Corporate/executive B2B channel |\n\n## 6. Lifestyle, Luxury & Executive Press — *Authority + Observation*\n| # | Corridor | DLX angle |\n|---|---|---|\n| 41 | **BellaNaija Weddings** ⭐ | Groom features → wedding-intent audience |\n| 42 | **ThisDay Style** ⭐ | Luxury menswear editorial placement |\n| 43 | **BusinessDay Lifestyle** ⭐ | Executive readership outreach |\n| 44 | **Guardian Life & Style** ⭐ | Prestige + credibility |\n| 45 | **Vanguard Allure** ⭐ | Style vertical reach |\n| 46 | **Nairametrics** ⭐ | Business/exec audience alignment |\n| 47 | **Pulse NG — Fashion** ⭐ | Cultural relevance + trends |\n| 48 | **Style Vitae** | Local style editorial |\n| 49 | **Eventbrite Lagos** ⭐ | Fashion/business events to attend or sponsor |\n| 50 | **Meetup Lagos / Groups** ⭐ | Executive & fashion networking clusters |\n\n---\n\n## How the engine consumes each corridor\n\n1. **Route the inquiry** — every DM, form, or email from a corridor lands tagged\n with the corridor as `source`.\n2. **Qualify** — the strict engine accepts only genuine high-intent (≥ 50),\n rejects budget seekers, and skips duplicates.\n3. **Prioritize** — the dashboard surfaces strongest opportunities first.\n4. **Track & learn** — analytics report **which corridors and signals actually\n convert through the pipeline to deposit/ordered/won**, so budget flows to what\n produces revenue — not to vanity listings.\n\nStart with the ⭐ majors (Jumia, Konga, Jiji, ConnectNigeria, Nairaland,\nBellaNaija Weddings, Lagos Fashion Week, the national press) — they carry the\nmost real, high-intent audience for premium bespoke menswear today.\n",[],[197,389],["D",1788495223471],[247,248],[197],"4f07185f-72ef-41c9-90ff-d977099001bd","schema.sql","8257134c-a7b3-11f1-a292-1607ee4eb77e",48,["D",1788493785335],"-- =============================================================================\n-- DLX Designs — verified leads (Supabase / PostgreSQL schema)\n-- Run this in the Supabase SQL editor (Dashboard > SQL > New query) to create\n-- the table the val mirrors verified leads + pipeline state into. The val's\n-- native SQLite store works even without Supabase; setting SUPABASE_URL +\n-- SUPABASE_SERVICE_ROLE_KEY turns this table into a live mirror.\n-- =============================================================================\n\ncreate extension if not exists pgcrypto;\n\ncreate table if not exists public.dlx_verified_leads (\n id uuid primary key default gen_random_uuid(),\n name text not null,\n gmail text not null unique,\n phone text not null default '',\n score integer not null default 0 check (score between 0 and 100),\n source text not null default '',\n source_url text not null default '',\n evidence text not null default '',\n signal text not null default '',\n action text not null default '',\n status text not null default 'new',\n history jsonb not null default '[]'::jsonb,\n created_at timestamptz not null default now(),\n updated_at timestamptz not null default now()\n);\n\n-- Fast sorts: strongest opportunities first, then newest.\ncreate index if not exists dlx_verified_leads_score_idx\n on public.dlx_verified_leads(score desc);\ncreate index if not exists dlx_verified_leads_created_idx\n on public.dlx_verified_leads(created_at desc);\ncreate index if not exists dlx_verified_leads_status_idx\n on public.dlx_verified_leads(status);\ncreate index if not exists dlx_verified_leads_source_idx\n on public.dlx_verified_leads(source);\n\n-- Row Level Security: lock it down so only service-role (token) calls can\n-- read/write until you opt into per-user policies. Using the anon key from the\n-- browser WILL be blocked by default — intentional (this is internal data).\nalter table public.dlx_verified_leads enable row level security;\n",[],[203,399],["D",1788452485280],[245],[203],"697dc879-346a-476d-b8d5-6cb2817ec391","hunt.ts","a828a7f6-a813-11f1-8400-1607ee4eb77e",["D",1788493780394],"interval","// cron/hunt.ts\n// Scheduled hunt poller. Runs every interval; pulls genuine inquiries from the\n// configured real source (DLX_HUNT_URL) into the pipeline. Safe to run\n// repeatedly — duplicates are skipped by the unique email constraint.\nimport { hunt } from \"../backend/hunt.ts\";\n\nexport default async function (interval: Interval) {\n await hunt(interval.lastRunAt ?? new Date());\n}\n",[],[209,410],"dbf65aae-cd1b-4046-a84f-5509c76759b6","automate.ts","9ed04cb0-a820-11f1-8c3c-1607ee4eb77e",80,["D",1788499361794],"// cron/automate.ts\n// NURTURE AUTOMATON — runs every interval to keep the pipeline from going cold.\n//\n// 1. FLAG: surfaces hot leads (score >= 70) still sitting in early stages\n// (new / qualified) for more than 24h — the team should be on these.\n// 2. NUDGE: auto-advances those stale hot leads to \"contacted\" (with a note)\n// so they appear in the team's work queue. Safe: never skips stages, never\n// invents contacts.\n// 3. AUTO-SEND: only if ENABLE_AUTOSEND=1, sends the first-touch to prospects\n// that already have a verified email and were drafted but not yet sent.\n//\n// All actions are logged so the operator sees exactly what the machine did.\nimport { listLeads, setLeadStatus } from \"../lib/store.ts\";\nimport { listProspects, contactProspect } from \"../backend/outbound.ts\";\nimport { autosendEnabled } from \"../backend/config.ts\";\n\nconst HOT = 70;\nconst STALE_MS = 24 * 60 * 60 * 1000;\n\nexport default async function (interval: Interval) {\n const now = Date.now();\n const summary: string[] = [];\n\n // 1 + 2: flag and nudge stale hot verified leads\n let flagged = 0;\n let nudged = 0;\n const leads = await listLeads(300);\n for (const lead of leads) {\n if ((lead.status !== \"new\" && lead.status !== \"qualified\")) continue;\n if (lead.score < HOT) continue;\n const created = new Date(lead.created_at).getTime();\n if (!Number.isFinite(created)) continue;\n if (now - created < STALE_MS) continue;\n flagged++;\n const updated = await setLeadStatus(\n lead.id,\n \"contacted\",\n `auto-nudge: hot lead (${lead.score}/100) idle >24h at ${new Date().toISOString()}`,\n );\n if (updated) nudged++;\n }\n\n // 3: auto-send drafted prospects with verified emails (guarded)\n let drafted = 0;\n let sent = 0;\n const prospects = await listProspects(100);\n for (const p of prospects) {\n if (p.status !== \"drafted\") continue;\n if (!p.email) continue;\n drafted++;\n const out = await contactProspect(p.id); // dry-run unless ENABLE_AUTOSEND\n if (out.sent) sent++;\n }\n\n const line =\n `automate @ ${interval.lastRunAt?.toISOString() ?? \"first-run\"} | ` +\n `hotLeadsFlagged=${flagged} nudged=${nudged} | ` +\n `sendableProspects=${drafted} firstTouchSent=${sent} (autosend=${autosendEnabled() ? \"ON\" : \"dry-run\"})`;\n summary.push(line);\n console.log(line);\n return summary;\n}\n",[],[209,418],["D",1788493780394],[242,243],[209],"f02c382b-c1df-4268-b9ad-2011834406bf",["D",1788500403600],[],[215,38],"5a3d3f9e-c71d-491f-bc53-88b16a617083","pipeline.ts","7f43465c-a813-11f1-a948-1607ee4eb77e",34,["D",1788493711801],"// backend/pipeline.ts\n// The DLX Designs sales pipeline stages. Every verified lead moves through\n// these; each transition is recorded in the lead's history so the business can\n// learn which stages convert and from which sources.\nexport const STAGES = [\n \"new\",\n \"qualified\",\n \"contacted\",\n \"responded\",\n \"consultation\",\n \"quoted\",\n \"deposit_paid\",\n \"ordered\",\n \"won\",\n \"lost\",\n] as const;\n\nexport type Stage = (typeof STAGES)[number];\n\nexport function isStage(s: string): s is Stage {\n return (STAGES as readonly string[]).includes(s);\n}\n\nexport const STAGE_LABELS: Record = {\n new: \"New\",\n qualified: \"Qualified\",\n contacted: \"Contacted\",\n responded: \"Responded\",\n consultation: \"Consultation\",\n quoted: \"Quoted\",\n deposit_paid: \"Deposit Paid\",\n ordered: \"Ordered\",\n won: \"Won\",\n lost: \"Lost\",\n};\n\n/** \"Won\" and \"lost\" are terminal — nothing advances after them. */\nexport function nextStage(current: Stage): Stage | null {\n if (current === \"won\" || current === \"lost\") return null;\n const idx = STAGES.indexOf(current);\n return STAGES[idx + 1] ?? null;\n}\n",[],[215,433],"d7a7d679-a3c4-40f2-b461-712e8bbb66a4","outbound.ts","8aae514c-a81a-11f1-8613-1607ee4eb77e",81,["D",1788499361859],"// backend/outbound.ts\n// PROSPECT → ACT automaton layer.\n//\n// A PROSPECT is a REAL public buying signal (a corridor post, an IG comment, a\n// marketplace \"looking for a tailor\" thread, a forwarded DM) that we've\n// captured with the contact handle but have NOT yet verified an email for.\n// This module:\n// 1. Scores the real signal's intent (honest — no fabricated contacts).\n// 2. Auto-drafts a personalized first-touch specific to the signal.\n// 3. Sends it ONLY when a real contact exists, and logs the outcome.\n// 4. Promotes to a verified lead the moment an email/phone is captured.\n//\n// Sends are OFF by default: set ENABLE_AUTOSEND=1 to let the engine actually\n// email. Otherwise it returns the ready-to-send draft (dry-run) so a human\n// reviews first. This is responsible delegation, not a spam bot.\nimport { sqlite } from \"https://esm.town/v/std/sqlite/main.ts\";\nimport { email } from \"https://esm.town/v/std/email\";\nimport { analyzeSignal } from \"./worker/qualifier.ts\";\nimport { ingestLead } from \"./worker/ingest.ts\";\nimport { brand } from \"./config.ts\";\n\nexport interface Prospect {\n id: string;\n handle: string;\n name: string;\n email: string; // '' until verified\n phone: string;\n message: string; // evidence\n source: string;\n source_url: string;\n score: number;\n signal: string;\n action: string;\n status: \"new\" | \"drafted\" | \"contacted\" | \"responded\" | \"converted\" | \"ignored\";\n created_at: string;\n}\n\nasync function ensureTable(): Promise {\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS dlx_prospects (\n id TEXT PRIMARY KEY,\n handle TEXT NOT NULL,\n name TEXT NOT NULL DEFAULT '',\n email TEXT NOT NULL DEFAULT '',\n phone TEXT NOT NULL DEFAULT '',\n message TEXT NOT NULL DEFAULT '',\n source TEXT NOT NULL DEFAULT 'prospecting',\n source_url TEXT NOT NULL DEFAULT '',\n score INTEGER NOT NULL DEFAULT 0,\n signal TEXT NOT NULL DEFAULT '',\n action TEXT NOT NULL DEFAULT '',\n status TEXT NOT NULL DEFAULT 'new',\n created_at TEXT NOT NULL DEFAULT (datetime('now'))\n )`);\n}\n\nexport async function addProspect(input: {\n handle: string;\n name?: string;\n email?: string;\n phone?: string;\n message: string;\n source?: string;\n source_url?: string;\n}): Promise {\n await ensureTable();\n const analysis = analyzeSignal(input.message);\n const row: Prospect = {\n id: crypto.randomUUID(),\n handle: input.handle.trim(),\n name: (input.name ?? \"\").trim().slice(0, 80),\n email: (input.email ?? \"\").trim().toLowerCase().slice(0, 200),\n phone: (input.phone ?? \"\").trim().slice(0, 40),\n message: input.message.trim().slice(0, 2000),\n source: (input.source ?? \"prospecting\").trim(),\n source_url: input.source_url ?? \"\",\n score: analysis.score,\n signal: analysis.signal,\n action: analysis.action,\n status: \"new\",\n created_at: new Date().toISOString(),\n };\n await sqlite.execute({\n sql: `INSERT INTO dlx_prospects\n (id, handle, name, email, phone, message, source, source_url, score, signal, action, status, created_at)\n VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,\n args: [row.id, row.handle, row.name, row.email, row.phone, row.message,\n row.source, row.source_url, row.score, row.signal, row.action, row.status, row.created_at],\n });\n console.log(`prospect ${row.handle}: score ${row.score} — ${row.action}`);\n return row;\n}\n\nexport async function listProspects(limit = 50): Promise {\n await ensureTable();\n const res = await sqlite.execute(\n `SELECT * FROM dlx_prospects ORDER BY score DESC, created_at DESC LIMIT ?`,\n [limit],\n );\n return (res.rows as Record[]).map(prospectFromRow);\n}\n\n/** Auto-draft a personalized first-touch from the REAL signal. */\nexport function draftFirstTouch(p: Prospect): { subject: string; body: string } {\n const who = p.name || p.handle;\n const occasion = /(groom|wedding|chieftaincy|agbada|aso ebi)/i.test(p.message)\n ? \"your occasion\"\n : \"what you're looking for\";\n return {\n subject: `${brand()} — ${p.source} enquiry`,\n body:\n `Hello ${who},\\n\\n` +\n `I saw your note about \"${p.message.split(\"\\n\")[0].slice(0, 90)}\". ` +\n `At ${brand()} we make bespoke traditional menswear (grooms, chieftaincy, ` +\n `executive) in Lagos & Abeokuta. Happy to share a few options and a quote ` +\n `for ${occasion} — no obligation.\\n\\n` +\n `Would you like me to send details? You can reply here or reach us directly.\\n\\n` +\n `Warm regards,\\n${brand()}\\n`,\n };\n}\n\n/**\n * Send (or dry-run) the first touch. Actual send requires ENABLE_AUTOSEND=1 AND\n * a verified email. Always records the outcome in the prospect's status.\n */\nexport async function contactProspect(\n id: string,\n dryRun = !Deno.env.get(\"ENABLE_AUTOSEND\"),\n): Promise<{ ok: boolean; sent: boolean; draft: { subject: string; body: string }; reason?: string }> {\n await ensureTable();\n const res = await sqlite.execute(`SELECT * FROM dlx_prospects WHERE id = ?`, [id]);\n const row = res.rows[0] as Record | undefined;\n if (!row) return { ok: false, sent: false, draft: { subject: \"\", body: \"\" }, reason: \"prospect not found\" };\n const p = prospectFromRow(row);\n const draft = draftFirstTouch(p);\n\n if (!p.email) {\n await sqlite.execute(`UPDATE dlx_prospects SET status = 'drafted' WHERE id = ?`, [id]);\n return { ok: true, sent: false, draft, reason: \"no verified email yet — reply in channel to capture it\" };\n }\n if (dryRun) {\n await sqlite.execute(`UPDATE dlx_prospects SET status = 'drafted' WHERE id = ?`, [id]);\n return { ok: true, sent: false, draft, reason: \"dry-run (ENABLE_AUTOSEND not set)\" };\n }\n\n try {\n await email({ to: p.email, subject: draft.subject, text: draft.body });\n await sqlite.execute(`UPDATE dlx_prospects SET status = 'contacted' WHERE id = ?`, [id]);\n return { ok: true, sent: true, draft, reason: \"sent\" };\n } catch (err) {\n return { ok: false, sent: false, draft, reason: `send failed: ${String(err)}` };\n }\n}\n\n/** Promote a prospect to a VERIFIED lead once email/phone are confirmed. */\nexport async function convertProspect(\n id: string,\n email: string,\n phone?: string,\n): Promise<{ ok: boolean; reason?: string; lead?: unknown }> {\n await ensureTable();\n const res = await sqlite.execute(`SELECT * FROM dlx_prospects WHERE id = ?`, [id]);\n const row = res.rows[0] as Record | undefined;\n if (!row) return { ok: false, reason: \"prospect not found\" };\n const p = prospectFromRow(row);\n\n const outcome = await ingestLead({\n name: p.name || p.handle,\n gmail: email,\n phone: phone || p.phone,\n message: p.message,\n source: p.source,\n source_url: p.source_url,\n });\n if (outcome.decision !== \"pass\") return { ok: false, reason: outcome.reason };\n\n await sqlite.execute(\n `UPDATE dlx_prospects SET status = 'converted', email = ?, phone = ? WHERE id = ?`,\n [email.toLowerCase(), phone || p.phone, id],\n );\n return { ok: true, lead: outcome.lead };\n}\n\nfunction prospectFromRow(r: Record): Prospect {\n return {\n id: String(r.id),\n handle: String(r.handle),\n name: String(r.name ?? \"\"),\n email: String(r.email ?? \"\"),\n phone: String(r.phone ?? \"\"),\n message: String(r.message ?? \"\"),\n source: String(r.source ?? \"\"),\n source_url: String(r.source_url ?? \"\"),\n score: Number(r.score ?? 0),\n signal: String(r.signal ?? \"\"),\n action: String(r.action ?? \"\"),\n status: String(r.status) as Prospect[\"status\"],\n created_at: String(r.created_at ?? \"\"),\n };\n}\n",[],[215,441],"95bc22ba-fb3a-4936-b044-b2e6220757db","instagram.ts","c9ec2a24-a819-11f1-98e7-1607ee4eb77e",67,["D",1788496413864],"// backend/instagram.ts\n// INSTAGRAM — real, OAuth-authorized, webhook-driven lead capture.\n//\n// When a Meta Business app is connected (see README), Instagram pushes every\n// real DM/comment here. No scraping, no fabricated contacts. Because Instagram\n// DMs carry no email, each captured conversation is stored as a SIGNAL (real\n// text + handle + permalink). The team replies to capture the email, then\n// promotes the signal to a verified lead with one API call.\nimport { sqlite } from \"https://esm.town/v/std/sqlite/main.ts\";\nimport { ingestLead } from \"./worker/ingest.ts\";\n\nexport interface IgSignal {\n id: string;\n ig_user: string;\n ig_id: string;\n message: string;\n permalink: string;\n source: string; // instagram-dm | instagram-comment\n status: \"new\" | \"converted\" | \"ignored\";\n created_at: string;\n}\n\nasync function ensureTable(): Promise {\n await sqlite.execute(`CREATE TABLE IF NOT EXISTS dlx_instagram_signals (\n id TEXT PRIMARY KEY,\n ig_user TEXT NOT NULL,\n ig_id TEXT NOT NULL,\n message TEXT NOT NULL DEFAULT '',\n permalink TEXT NOT NULL DEFAULT '',\n source TEXT NOT NULL DEFAULT 'instagram-dm',\n status TEXT NOT NULL DEFAULT 'new',\n created_at TEXT NOT NULL DEFAULT (datetime('now'))\n )`);\n}\n\n/** Meta webhook subscription verification (the GET handshake). */\nexport async function verifyWebhook(\n params: URLSearchParams,\n): Promise {\n const mode = params.get(\"hub.mode\");\n const token = params.get(\"hub.verify_token\");\n const challenge = params.get(\"hub.challenge\");\n if (mode === \"subscribe\" && token === Deno.env.get(\"INSTAGRAM_VERIFY_TOKEN\")) {\n return challenge;\n }\n return null;\n}\n\n/**\n * Handle a pushed webhook payload: extract real DM/comment events, capture\n * them as evidence-backed signals. Returns the number stored.\n */\nexport async function handleEvent(payload: unknown): Promise {\n await ensureTable();\n const root = (payload ?? {}) as {\n entry?: { messaging?: unknown[]; changes?: unknown[] }[];\n };\n const events: { text: string; userId: string; username: string; link: string; kind: string }[] =\n [];\n\n for (const entry of root.entry ?? []) {\n for (const m of entry.messaging ?? []) {\n const msg = m as { sender?: { id?: string }; message?: { text?: string } };\n const text = (msg?.message?.text ?? \"\").trim();\n const senderId = msg?.sender?.id ?? \"\";\n if (text && senderId) {\n events.push({\n text,\n userId: senderId,\n username: `ig-u-${senderId.slice(0, 12)}`,\n link: `https://instagram.com/${senderId}`,\n kind: \"instagram-dm\",\n });\n }\n }\n for (const c of entry.changes ?? []) {\n const v = (c as { value?: { from?: { username?: string; id?: string }; text?: string; permalink?: string } })\n ?.value;\n const text = (v?.text ?? \"\").trim();\n const username = v?.from?.username ?? \"\";\n if (text && username) {\n events.push({\n text,\n userId: v?.from?.id ?? username,\n username,\n link: v?.permalink ?? `https://instagram.com/${username}`,\n kind: \"instagram-comment\",\n });\n }\n }\n }\n\n let stored = 0;\n for (const ev of events) {\n try {\n await sqlite.execute({\n sql: `INSERT OR IGNORE INTO dlx_instagram_signals\n (id, ig_user, ig_id, message, permalink, source, status, created_at)\n VALUES (?, ?, ?, ?, ?, ?, 'new', datetime('now'))`,\n args: [\n crypto.randomUUID(),\n ev.username,\n ev.userId,\n ev.text.slice(0, 2000),\n ev.link,\n ev.kind,\n ],\n });\n stored++;\n } catch {\n /* ignore single-event failures */\n }\n }\n console.log(`instagram-webhook: captured ${stored} real ${stored === 1 ? \"signal\" : \"signals\"}`);\n return stored;\n}\n\nexport async function listSignals(limit = 50): Promise {\n await ensureTable();\n const res = await sqlite.execute(\n `SELECT * FROM dlx_instagram_signals ORDER BY created_at DESC LIMIT ?`,\n [limit],\n );\n return (res.rows as Record[]).map((r) => ({\n id: String(r.id),\n ig_user: String(r.ig_user),\n ig_id: String(r.ig_id),\n message: String(r.message),\n permalink: String(r.permalink),\n source: String(r.source),\n status: String(r.status) as IgSignal[\"status\"],\n created_at: String(r.created_at),\n }));\n}\n\n/**\n * Promote a real IG signal into a VERIFIED lead once the email/phone is\n * obtained by replying. The signal must exist; its evidence is carried over.\n */\nexport async function promoteSignal(\n id: string,\n email: string,\n phone?: string,\n): Promise<{ ok: boolean; reason?: string; lead?: unknown }> {\n await ensureTable();\n const res = await sqlite.execute(\n `SELECT * FROM dlx_instagram_signals WHERE id = ?`,\n [id],\n );\n const row = res.rows[0] as Record | undefined;\n if (!row) return { ok: false, reason: \"signal not found\" };\n if (String(row.status) === \"converted\") {\n return { ok: false, reason: \"signal already converted\" };\n }\n\n const outcome = await ingestLead({\n name: String(row.ig_user),\n gmail: email,\n phone,\n message: String(row.message),\n source: String(row.source),\n source_url: String(row.permalink),\n });\n\n if (outcome.decision !== \"pass\") {\n return { ok: false, reason: outcome.reason };\n }\n\n await sqlite.execute(\n `UPDATE dlx_instagram_signals SET status = 'converted' WHERE id = ?`,\n [id],\n );\n return { ok: true, lead: outcome.lead };\n}\n",[],[215,449],"c410888c-4a87-4470-a3df-8bcc00cc17b4","a81c406a-a813-11f1-8400-1607ee4eb77e",54,["D",1788494704673],"// backend/hunt.ts\n// HUNT — the autonomous ingestion layer.\n//\n// This is the honest form of \"hunting\": it watches a REAL, configured source\n// and feeds genuinely captured inquiries through the qualifier into the\n// pipeline. It never invents people, names, or signals — it only ingests what\n// the source actually contains.\n//\n// Configure one or more real sources:\n// DLX_HUNT_URL — a JSON or CSV endpoint / published feed.\n// DLX_HUNT_TOKEN — optional bearer token sent as `Authorization: Bearer …`\n// (e.g. for a private feed).\n//\n// JSON response shape (either a bare array or wrapped):\n// { \"leads\": [ { name, gmail, phone, message, source, source_url } , ... ] }\n//\n// CSV: a published Google Sheet CSV is accepted directly. First row must be a\n// header that includes at least `gmail` (plus optional name, phone, message,\n// source, source_url). Google Sheets: File → Share → Publish to web → CSV,\n// copy the link, use it as DLX_HUNT_URL.\n//\n// Deduplication is automatic: the store's unique gmail constraint means a lead\n// already in the pipeline is skipped, so the poller is idempotent and safe to\n// run on a cron.\nimport { ingestLead, type InboundLead } from \"./worker/ingest.ts\";\n\nexport interface HuntResult {\n source: string;\n fetched: number;\n qualified: number;\n rejected: number;\n duplicates: number;\n}\n\nexport async function hunt(now = new Date()): Promise {\n const url = Deno.env.get(\"DLX_HUNT_URL\");\n const token = Deno.env.get(\"DLX_HUNT_TOKEN\");\n const result: HuntResult = { source: url ?? \"(none)\", fetched: 0, qualified: 0, rejected: 0, duplicates: 0 };\n\n if (!url) {\n console.log(`hunt: no DLX_HUNT_URL configured at ${now.toISOString()}`);\n return result;\n }\n\n const headers: Record = { Accept: \"text/csv, application/json, text/plain\" };\n if (token) headers.Authorization = `Bearer ${token}`;\n\n let text: string;\n try {\n const res = await fetch(url, { headers });\n if (!res.ok) throw new Error(`HTTP ${res.status}`);\n text = await res.text();\n } catch (err) {\n console.error(\"hunt: fetch failed from source:\", err);\n return result;\n }\n\n const trimmed = text.trim();\n const looksJson = trimmed.startsWith(\"{\") || trimmed.startsWith(\"[\");\n const rows: Record[] = looksJson\n ? jsonlize(trimmed)\n : csvlize(trimmed);\n\n for (const item of rows) {\n if (!item.gmail) continue; // evidence-less rows are ignored, never invented\n result.fetched++;\n const outcome = await ingestLead(item as unknown as InboundLead);\n if (outcome.decision === \"pass\") {\n result.qualified++;\n } else if (\n typeof outcome.reason === \"string\" &&\n outcome.reason.includes(\"duplicate\")\n ) {\n result.duplicates++;\n } else {\n result.rejected++;\n }\n }\n\n console.log(\n `hunt ${now.toISOString()} source=${result.source} fetched=${result.fetched} qualified=${result.qualified} rejected=${result.rejected} duplicates=${result.duplicates}`,\n );\n return result;\n}\n\nfunction jsonlize(text: string): Record[] {\n let payload: unknown;\n try {\n payload = JSON.parse(text);\n } catch {\n return [];\n }\n if (Array.isArray(payload)) return payload as Record[];\n if (Array.isArray((payload as { leads?: unknown[] })?.leads)) {\n return (payload as { leads: unknown[] }).leads as Record[];\n }\n return [];\n}\n\n/** Minimal RFC-4180-ish CSV parser (handles quoted fields and embedded commas). */\nfunction csvlize(text: string): Record[] {\n const lines = splitCsvLines(text);\n if (lines.length < 2) return [];\n const header = parseCsvLine(lines[0]).map((h) => h.trim().toLowerCase());\n const gidx = header.indexOf(\"gmail\");\n if (gidx === -1) return []; // must have a gmail column\n const out: Record[] = [];\n for (let i = 1; i < lines.length; i++) {\n const cells = parseCsvLine(lines[i]);\n if (!cells[gidx]?.trim()) continue;\n const row: Record = {};\n header.forEach((h, j) => {\n if (h) row[h] = (cells[j] ?? \"\").trim();\n });\n if (row.gmail) out.push(row);\n }\n return out;\n}\n\nfunction splitCsvLines(text: string): string[] {\n const out: string[] = [];\n let cur = \"\";\n let inQ = false;\n for (let i = 0; i < text.length; i++) {\n const ch = text[i];\n if (ch === '\"') {\n inQ = !inQ;\n cur += ch;\n } else if (ch === \"\\n\" && !inQ) {\n out.push(cur);\n cur = \"\";\n } else {\n cur += ch;\n }\n }\n if (cur.trim()) out.push(cur);\n return out;\n}\n\nfunction parseCsvLine(line: string): string[] {\n const cells: string[] = [];\n let cur = \"\";\n let inQ = false;\n for (let i = 0; i < line.length; i++) {\n const ch = line[i];\n if (ch === '\"') {\n if (inQ && line[i + 1] === '\"') {\n cur += '\"';\n i++;\n } else {\n inQ = !inQ;\n }\n } else if (ch === \",\" && !inQ) {\n cells.push(cur);\n cur = \"\";\n } else {\n cur += ch;\n }\n }\n cells.push(cur);\n return cells;\n}\n",[],[215,410],"33eb4a85-5055-4549-8bde-121fd72e5dfc","config.ts","3d67d51e-a81c-11f1-b822-1607ee4eb77e",77,["D",1788497466783],"// backend/config.ts\n// Whitelabel + runtime configuration for the engine.\n//\n// `DLX_BRAND_NAME` (default \"DLX Designs\") lets you run this same engine for\n// multiple brands/clients in separate deployments without editing code. All\n// outbound copy and the capture page read the brand from here.\nexport function brand(): string {\n return Deno.env.get(\"DLX_BRAND_NAME\") || \"DLX Designs\";\n}\n\nexport function brandTitle(): string {\n return Deno.env.get(\"DLX_BRAND_NAME\")\n ? `Verified Lead Intelligence · ${brand()}`\n : \"DLX Designs · Verified Lead Intelligence\";\n}\n\nexport function autosendEnabled(): boolean {\n return Deno.env.get(\"ENABLE_AUTOSEND\") === \"1\";\n}\n",[],[215,464],"715d4b0d-c723-44c7-b3dd-4ef800afbeb6","qualifier.ts","8248f410-a7b3-11f1-a292-1607ee4eb77e","31184a6c-2262-49a1-9d1e-4a43608cddfb",72,["D",1788496720303],"// backend/worker/qualifier.ts\n// The strict qualification engine for DLX Designs.\n//\n// Two layers:\n// 1. Hard disqualifiers — instant reject on low-budget language.\n// 2. Scoring matrix (0-100) — high-tier buying signals raise the score.\n// A lead only passes when (a) no disqualifier fires AND (b) its score clears\n// the PASS_THRESHOLD. Inputs that fail are rejected and never persisted.\nimport { llmRefineScore } from \"./llm.ts\";\n\nexport interface InboundLead {\n name?: unknown;\n gmail?: unknown;\n phone?: unknown;\n message?: unknown;\n source?: unknown;\n source_url?: unknown;\n evidence?: unknown;\n}\n\nexport interface QualifiedLead {\n name: string;\n gmail: string;\n phone: string;\n score: string; // e.g. \"80/100%\"\n source: string;\n source_url: string;\n evidence: string;\n signal: string;\n action: string; // recommended next action\n}\n\nexport interface QualifyOutcome {\n decision: \"pass\" | \"reject\";\n lead?: QualifiedLead;\n reason: string;\n numericScore?: number;\n}\n\nconst DISQUALIFIERS = [\n \"cheap\",\n \"affordable\",\n \"under 20k\",\n \"under20k\",\n \"20k\",\n \"discount\",\n \"thrift\",\n \"okirika\",\n];\n\n// High-tier signals -> points toward intent. Matched words are deduped so a\n// single mention counts once.\nconst HIGH_TIER: Record = {\n bespoke: 18,\n \"3-piece\": 12,\n \"three-piece\": 12,\n \"3 piece\": 12,\n cashmere: 15,\n groom: 20,\n chieftaincy: 20,\n executive: 14,\n wedding: 14,\n suit: 8,\n agbada: 8,\n quality: 5,\n premium: 8,\n lagos: 6,\n abeokuta: 6,\n lekki: 10,\n \"victoria island\": 10,\n nigeria: 3,\n};\n\nconst BASE_SCORE = 25;\nconst PASS_THRESHOLD = 50;\nconst MAX_SCORE = 100;\n\nconst EMAIL_RE = /[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\\.[a-zA-Z]{2,}/;\n\n/**\n * Strict, self-contained heuristic qualification. Never throws; never\n * fabricates data. Works with or without an LLM key.\n */\nexport async function qualifyLead(inbound: InboundLead): Promise {\n const message = (String(inbound?.message ?? \"\")).toLowerCase();\n const gmailRaw = String(inbound?.gmail ?? \"\").trim();\n const source = String(inbound?.source ?? (inbound?.gmail ? \"inbound-form\" : \"\")).trim();\n\n // --- Layer 1: hard disqualifiers -------------------------------------\n for (const d of DISQUALIFIERS) {\n if (message.includes(d)) {\n return {\n decision: \"reject\",\n reason: `disqualifier '${d}' present in message (low-budget signal)`,\n };\n }\n }\n\n // --- Validation -------------------------------------------------------\n const gmailMatch = gmailRaw.toLowerCase().match(EMAIL_RE);\n if (!gmailMatch) {\n return { decision: \"reject\", reason: \"no valid email address supplied\" };\n }\n const gmail = gmailMatch[0].toLowerCase();\n\n const name = (String(inbound?.name ?? \"\").trim() ||\n gmail.split(\"@\")[0]).slice(0, 80);\n if (!name) {\n return { decision: \"reject\", reason: \"no name supplied\" };\n }\n\n // --- Layer 2: scoring matrix ------------------------------------------\n let score = BASE_SCORE;\n const matched: string[] = [];\n for (const [kw, pts] of Object.entries(HIGH_TIER)) {\n if (message.includes(kw)) {\n score += pts;\n matched.push(kw);\n }\n }\n score = Math.min(MAX_SCORE, score);\n const numericScore = score;\n\n if (numericScore < PASS_THRESHOLD) {\n return {\n decision: \"reject\",\n reason:\n `intent score ${numericScore}/100 below ${PASS_THRESHOLD} threshold ` +\n `(insufficient buying signal — no bespoke/luxury context detected)`,\n numericScore,\n };\n }\n\n // Build a human-readable signal from the strongest matched keywords.\n const signal =\n matched.length > 0\n ? `Interest detected: ${matched.slice(0, 4).join(\", \")}`\n : \"High-intent luxury menswear inquiry\";\n\n // --- Layer 3: optional LLM refinement --------------------------------\n let finalScore = numericScore;\n let finalSignal = signal;\n const refined = await llmRefineScore(inbound, numericScore);\n if (refined) {\n finalScore = refined.numericScore;\n if (refined.signal) finalSignal = refined.signal;\n }\n\n return {\n decision: \"pass\",\n lead: {\n name,\n gmail,\n phone: String(inbound?.phone ?? \"\").trim().slice(0, 40),\n score: `${finalScore}/100%`,\n source: source || \"inbound-form\",\n source_url: String(inbound?.source_url ?? \"\").trim().slice(0, 500),\n evidence: (String(inbound?.evidence ?? \"\") ||\n String(inbound?.message ?? \"\")).trim().slice(0, 2000),\n signal: finalSignal,\n action: recommendAction(finalScore, finalSignal),\n },\n reason: \"verified high-intent lead\",\n numericScore: finalScore,\n };\n}\n\n/** Recommend the next sales action from the proven intent signal — never invented. */\nfunction recommendAction(score: number, signal: string): string {\n const s = signal.toLowerCase();\n if (/(groom|chieftaincy|wedding|consultation|quote)/.test(s)) {\n return \"Book consultation — upcoming occasion, quote at first meeting\";\n }\n if (score >= 80) {\n return \"High priority — contact within 24h to arrange consultation\";\n }\n if (score >= 65) {\n return \"Contact to confirm budget & timeline, then quote\";\n }\n return \"Nurture — follow up to confirm occasion & budget before quoting\";\n}\n/**\n * Score a raw public signal (message text only — no email required). Used to\n * rank PROSPECTS (real intent, still need a verified contact) honestly, without\n * ever inventing a person or email.\n */\nexport interface SignalAnalysis {\n score: number; // 0..100\n signal: string;\n action: string;\n rejected?: string; // set when a disqualifier fired\n}\n\nexport function analyzeSignal(message: string): SignalAnalysis {\n const m = (message ?? \"\").toLowerCase();\n for (const d of DISQUALIFIERS) {\n if (m.includes(d)) {\n return { score: 0, signal: \"\", action: \"discard — low-budget signal\", rejected: d };\n }\n }\n let score = BASE_SCORE;\n const matched: string[] = [];\n for (const [kw, pts] of Object.entries(HIGH_TIER)) {\n if (m.includes(kw)) {\n score += pts;\n matched.push(kw);\n }\n }\n score = Math.min(MAX_SCORE, score);\n const signal = matched.length\n ? `Interest: ${matched.slice(0, 4).join(\", \")}`\n : \"High-intent inquiry (confirm occasion & budget)\";\n return { score, signal, action: recommendAction(score, matched.join(\" \")) };\n}\n",[],[215,480,472],"worker","bab5c45c-d890-4bf3-9ff4-227b1782d3e6","llm.ts","735a5200-a7b3-11f1-8441-1607ee4eb77e",27,["D",1788493422605],"// backend/worker/llm.ts\n// PLUGGABLE, OPTIONAL AI score-refinement adapter.\n//\n// The engine is DETERMINISTIC-FIRST and fully production-ready WITHOUT any AI\n// key: the strict heuristic scoring matrix in qualifier.ts is authoritative\n// and never depends on this module. That removes OpenAI (or any provider) as a\n// single point of failure.\n//\n// When a key IS present, the matching provider refines the score and writes a\n// natural buying signal. Providers are tried in order, free tiers first:\n// 1. GEMINI_API_KEY — Google AI Studio free tier (no billing required)\n// 2. GROQ_API_KEY — Groq free tier (fast, generous quota)\n// 3. OPENAI_API_KEY — OpenAI (requires active billing)\n// If a provider fails (wrong key, quota, billing, network), we simply fall\n// through to the next one and ultimately to the heuristic — never an error.\nexport interface InboundLead {\n name?: unknown;\n gmail?: unknown;\n message?: unknown;\n source?: unknown;\n}\n\nexport async function llmRefineScore(\n inbound: InboundLead,\n heuristicScore: number,\n): Promise<{ numericScore: number; signal: string } | null> {\n const prompt = buildPrompt(inbound, heuristicScore);\n\n // 1) Gemini (free tier — recommended)\n const geminiKey = Deno.env.get(\"GEMINI_API_KEY\");\n if (geminiKey) {\n const r = await tryGemini(geminiKey, prompt, heuristicScore);\n if (r) return r;\n }\n\n // 2) Groq (free tier)\n const groqKey = Deno.env.get(\"GROQ_API_KEY\");\n if (groqKey) {\n const r = await tryGroq(groqKey, prompt, heuristicScore);\n if (r) return r;\n }\n\n // 3) OpenAI (only reached if OpenAI is provided and Gemini/Groq fail)\n const openaiKey = Deno.env.get(\"OPENAI_API_KEY\");\n if (openaiKey) {\n const r = await tryOpenAI(openaiKey, prompt, heuristicScore);\n if (r) return r;\n }\n\n return null; // caller falls back to the pure heuristic result\n}\n\nfunction buildPrompt(inbound: InboundLead, heuristicScore: number): string {\n return (\n `You are a lead qualification analyst for DLX Designs, a luxury bespoke ` +\n `traditional menswear brand for high-net-worth clients, grooms and corporate ` +\n `executives in Lagos and Abeokuta, Nigeria. ` +\n `A lead submitted: name=\"${String(inbound.name ?? \"\")}\", ` +\n `email=\"${String(inbound.gmail ?? \"\")}\", message=\"${String(inbound.message ?? \"\")}\". ` +\n `Return ONLY JSON: {\"score\":0-100,\"signal\":\"one-line buying intent\"}. ` +\n `Reject low budget demand (cheap, affordable, discount, thrift, okirika). ` +\n `The strict heuristic already scored this lead ${heuristicScore}/100.`\n );\n}\n\nasync function tryGemini(\n key: string,\n prompt: string,\n fallback: number,\n): Promise<{ numericScore: number; signal: string } | null> {\n try {\n const res = await fetch(\n \"https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash:generateContent?key=\" +\n key,\n {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({\n contents: [{ parts: [{ text: prompt + \" Respond with JSON only.\" }] }],\n }),\n },\n );\n if (!res.ok) throw new Error(`Gemini ${res.status}`);\n const data = await res.json();\n const text = data?.candidates?.[0]?.content?.parts?.[0]?.text ?? \"{}\";\n const cleaned = text.replace(/```json|```/g, \"\").trim();\n return normalize(JSON.parse(cleaned), fallback);\n } catch (err) {\n console.error(\"Gemini refine failed:\", err);\n return null;\n }\n}\n\nasync function tryGroq(\n key: string,\n prompt: string,\n fallback: number,\n): Promise<{ numericScore: number; signal: string } | null> {\n try {\n const res = await fetch(\"https://api.groq.com/openai/v1/chat/completions\", {\n method: \"POST\",\n headers: {\n \"Content-Type\": \"application/json\",\n Authorization: `Bearer ${key}`,\n },\n body: JSON.stringify({\n model: \"llama-3.3-70b-versatile\",\n temperature: 0,\n response_format: { type: \"json_object\" },\n messages: [\n { role: \"system\", content: \"You output strict JSON only.\" },\n { role: \"user\", content: prompt },\n ],\n }),\n });\n if (!res.ok) throw new Error(`Groq ${res.status} ${await res.text()}`);\n const data = await res.json();\n const parsed = JSON.parse(data.choices?.[0]?.message?.content ?? \"{}\");\n return normalize(parsed, fallback);\n } catch (err) {\n console.error(\"Groq refine failed:\", err);\n return null;\n }\n}\n\nasync function tryOpenAI(\n key: string,\n prompt: string,\n fallback: number,\n): Promise<{ numericScore: number; signal: string } | null> {\n try {\n const res = await fetch(\"https://api.openai.com/v1/chat/completions\", {\n method: \"POST\",\n headers: {\n \"Content-Type\": \"application/json\",\n Authorization: `Bearer ${key}`,\n },\n body: JSON.stringify({\n model: \"gpt-4o-mini\",\n temperature: 0,\n response_format: { type: \"json_object\" },\n messages: [\n { role: \"system\", content: \"You output strict JSON only.\" },\n { role: \"user\", content: prompt },\n ],\n }),\n });\n if (!res.ok) throw new Error(`OpenAI ${res.status}`);\n const data = await res.json();\n const parsed = JSON.parse(data.choices?.[0]?.message?.content ?? \"{}\");\n return normalize(parsed, fallback);\n } catch (err) {\n console.error(\"OpenAI refine failed:\", err);\n return null;\n }\n}\n\nfunction normalize(\n parsed: { score?: unknown; signal?: unknown },\n fallbackScore: number,\n): { numericScore: number; signal: string } {\n const s = Number(parsed?.score);\n // Blend heuristic and LLM, keeping the heuristic as a floor.\n const numericScore = Number.isFinite(s)\n ? Math.max(fallbackScore, Math.min(100, Math.round(s)))\n : fallbackScore;\n const signal =\n typeof parsed?.signal === \"string\" ? parsed.signal.slice(0, 200) : \"\";\n return { numericScore, signal };\n}\n",[],[215,480,482],"6f49607f-22c7-466b-b2a1-44928124c7e4","ingest.ts","824ff7c4-a7b3-11f1-a292-1607ee4eb77e",37,["D",1788493725323],"// backend/worker/ingest.ts\n// Ingest worker — the bridge between real inbound traffic and the verified\n// store. Runs the strict qualifier; only passing, high-intent leads persist.\n// Rejects are returned (and never stored).\nimport { insertLead } from \"../../lib/store.ts\";\nimport { qualifyLead, type InboundLead } from \"./qualifier.ts\";\n\nexport async function ingestLead(inbound: InboundLead) {\n const outcome = await qualifyLead(inbound);\n if (outcome.decision === \"reject\") {\n return outcome;\n }\n\n const lead = outcome.lead!;\n const scoreInt = Math.min(100, Math.max(0, outcome.numericScore ?? 0));\n\n const inserted = await insertLead({\n name: lead.name,\n gmail: lead.gmail,\n phone: lead.phone,\n score: scoreInt,\n source: lead.source,\n source_url: lead.source_url,\n evidence: lead.evidence,\n signal: lead.signal,\n action: lead.action,\n status: \"new\",\n });\n\n if (!inserted) {\n return {\n decision: \"reject\",\n reason: \"duplicate email — lead already verified\",\n numericScore: scoreInt,\n };\n }\n\n return {\n decision: \"pass\",\n lead: {\n id: inserted.id,\n name: inserted.name,\n gmail: inserted.gmail,\n phone: inserted.phone,\n score: `${inserted.score}/100%`,\n source: inserted.source,\n source_url: inserted.source_url,\n evidence: inserted.evidence,\n signal: inserted.signal,\n action: inserted.action,\n status: inserted.status,\n created_at: inserted.created_at,\n },\n reason: \"verified high-intent lead stored\",\n numericScore: inserted.score,\n };\n}\n",[],[215,480,490],"f46f34f4-1677-4619-a22e-94d613250e13",["D",1788452460127],[232,233,234],[215,480],["D",1788452460127],[231,235,236,237,238,239,240],[215],["D",1788452396724],["D",1789494036282],{"_104":103,"_111":-5},{"_104":103,"_105":106,"_107":-5,"_19":-5,"_108":109,"_110":30},{"_13":14,"_37":94,"_48":47,"_95":49,"_96":-5,"_97":-5,"_98":-5,"_99":-5,"_100":-5,"_101":50,"_102":103},"ctx",{"_87":-5,"_131":542,"_543":544,"_545":546,"_547":30,"_548":30,"_549":30,"_550":64,"_551":-5,"_552":-7,"_553":-7,"_554":68,"_555":30,"_556":30,"_557":-5,"_558":30,"_559":30,"_560":30,"_561":-5},"ENV",{"_513":514,"_515":516,"_517":94,"_518":519,"_520":521,"_522":523,"_524":525,"_526":527,"_528":529,"_530":531,"_532":533,"_534":535,"_536":537,"_538":539,"_540":541},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

Diff · v88