GTM Signal Scoring

Open-source go-to-market signal scoring and outbound automation.
Account intelligence, AI-powered reply drafting, lead qualification, and sales routing — powered by Deepline.

Quick Start · Features · Architecture · API Docs · Deepline · License

This repo has been anonymized. Company names, signals, scoring weights, and configuration values have been scrubbed or replaced with representative examples — they are not real or accurate. The architecture, code patterns, and integrations are production-tested, but the specific data you see is illustrative. Customize the Atlas scoring weights, P0 title patterns, ICP definitions, qualification rules, and integration hooks for your own go-to-market motion before deploying.

What is this?

A standalone GTM engineering platform that turns your PostgreSQL database into a ranked account intelligence layer. Built for sales and revenue teams who want signal-based prioritization, automated outbound reply handling, and lead qualification — without relying on expensive all-in-one platforms.

Built with: Next.js, PostgreSQL (Neon), Deepline enrichment, OpenAI, Slack, Lemlist/SmartLead/Instantly/HeyReach, Attio CRM, HubSpot

Features

Signal Scoring (Atlas Algorithm)

Scores every account using tech stack adoption, contact seniority, and P0 penetration signals
30-day score trends with observed vs. derived data clearly labeled
Configurable ICP definitions and qualification rules via YAML
Backtested scoring weights you can override for your GTM motion

Outbound Reply Engine

Multi-provider: Lemlist, SmartLead, Instantly, HeyReach — all routed through Deepline gateway
Receives inbound replies via webhooks (Lemlist) and polling (SmartLead/Instantly/HeyReach via Trigger.dev)
Drafts AI-powered responses using OpenAI GPT-5-mini with full company context
Posts to Slack for human review with Approve / Edit / Reject buttons
Undo Send — approved messages queue for 60s via Upstash QStash before delivery
Round-robin lead routing to sales reps with Attio CRM + HubSpot sync

Lead Qualification Pipeline

AI-powered qualification against configurable ICP definitions
Website scraping + LLM analysis for product-market fit scoring
Automatic routing: qualified leads go to reps, others enter nurture campaigns
Full audit trail of every routing decision

Safety & Security

API key auth, Slack HMAC-SHA256 verification, Lemlist shared secret, QStash signature verification
Sliding-window rate limiting (DB-backed), SSRF protection, PII redaction
LLM prompt injection guardrails, Zod input validation, error sanitization
60-second undo-send queue via QStash with cancellation support

Architecture

┌─────────────┐    ┌──────────────┐    ┌─────────────┐
│  Lemlist     │───▶│  Webhook API │───▶│  LLM Draft  │
│  SmartLead   │    │ /api/outbound│    │  (OpenAI)   │
│  Instantly   │    └──────┬───────┘    └──────┬──────┘
│  HeyReach    │           │                    │
└─────────────┘     ┌──────▼────────────────────▼──────┐
                    │         PostgreSQL (Neon)          │
                    │  inbound.leads | conversations     │
                    │  qualification_results | routing    │
                    │  message_queue | rate_limit_log    │
                    └──────┬────────────────────────────┘
                           │
               ┌───────────▼───────────┐
               │     Slack Approval     │
               │  Approve │ Edit │ Undo │
               └───────────┬───────────┘
                           │
              ┌────────────▼────────────┐
              │  QStash (60s delay)      │
              │  → /api/send-reply       │
              │  → Deepline Gateway      │
              │  → Provider-specific API │
              └────────────┬────────────┘
                           │
              ┌────────────▼────────────┐
              │  Attio + HubSpot CRM    │
              └─────────────────────────┘

Data flow: Deepline enriches company and contact records → Atlas algorithm scores accounts → qualified leads route to sales reps → outbound replies get AI-drafted responses → Slack approval workflow → QStash delayed queue with undo-send → Deepline gateway sends via correct provider → CRM sync.

Quick Start

Prerequisites

Node.js 20+
PostgreSQL (Neon recommended) with dl_resolved and dl_graph schemas
Deepline CLI installed and authenticated (for enrichment)

Install

git clone https://github.com/getaero-io/gtm-signal-scoring
cd gtm-signal-scoring
npm install
cp .env.example .env.local
# Edit .env.local — see Environment Variables below

Run Database Migrations

# Core schema (accounts, scoring)
psql $DATABASE_URL -f scripts/migrate.sql

# Outbound engine (leads, conversations, qualification, routing)
psql $DATABASE_URL -f scripts/migrate-outbound.sql

# Message queue with QStash support
psql $DATABASE_WRITE_URL -f scripts/migrate-qstash-message-id.sql

# Rate limiting
psql $DATABASE_URL -f scripts/migrate-message-queue.sql

Set Up QStash (Undo-Send Queue)

Create a free account at console.upstash.com
Copy your QStash Token, Current Signing Key, and Next Signing Key
Add them to .env.local (see Environment Variables below)
On Vercel: add via Upstash Marketplace integration or manually in Project Settings → Environment Variables

Start Development

npm run dev

Open http://localhost:3000.

Environment Variables

Required

DATABASE_URL=postgresql://user:pass@host:5432/db

Outbound Engine

# Write connection (falls back to DATABASE_URL if not set)
DATABASE_WRITE_URL=postgresql://user:pass@host:5432/db

# OpenAI for LLM reply drafting
OPENAI_API_KEY=sk-...

# Slack integration (https://api.slack.com/apps)
SLACK_BOT_TOKEN=xoxb-...
SLACK_SIGNING_SECRET=...
SLACK_CHANNEL_OUTBOUND=replybot
SLACK_CHANNEL_INBOUND=replybot

# Lemlist (https://app.lemlist.com/settings/integrations)
LEMLIST_API_KEY=...
LEMLIST_WEBHOOK_SECRET=your-shared-secret
LEMLIST_CAMPAIGN_IDS=camp_abc,camp_def

# Deepline gateway — routes to Lemlist/SmartLead/Instantly/HeyReach
DEEPLINE_API_KEY=...
DEEPLINE_CLI_PATH=/usr/local/bin/deepline

QStash (Undo-Send Queue)

Get these from console.upstash.com/qstash:

QSTASH_TOKEN=...
QSTASH_CURRENT_SIGNING_KEY=sig_...
QSTASH_NEXT_SIGNING_KEY=sig_...

Security

# API key for dashboard routes (leave empty in dev to skip auth)
INTERNAL_API_KEY=your-secret-api-key

# Message queue delay in seconds (default: 60)
MESSAGE_SEND_DELAY_SECONDS=60

# Enable test endpoints in production (default: disabled)
ALLOW_TEST_ENDPOINTS=false

AI & Search

# Anthropic (for AI qualification)
ANTHROPIC_API_KEY=sk-ant-...

# Exa web search (fallback enrichment)
EXA_API_KEY=...

Optional Integrations

# HubSpot
ENABLE_HUBSPOT=false
HUBSPOT_ACCESS_TOKEN=
HUBSPOT_PORTAL_ID=

# Notion
ENABLE_NOTION=false
NOTION_API_KEY=
NOTION_DATABASE_ID=

# Email (SMTP)
SMTP_HOST=smtp.example.com
SMTP_PORT=587
SMTP_USER=
SMTP_PASS=
SMTP_FROM="GTM Signal <noreply@example.com>"

Scoring Model

docs/SCORING_MODEL.md is a working example, not a universal template. The weights, title patterns, and thresholds are calibrated for one company's GTM motion and backtested against their pipeline data. Override them with values that reflect your ICP, deal stages, and conversion signals before deploying.

Example weights (customize in lib/scoring/scoring-config.json):

Base: 20 points
Tech adoption: up to +15 per tool (with time decay)
Contact seniority: C-Level +25, VP +20, Director +15, Manager +10
P0 engagement: +10 per qualified contact
Company size: +5 to +20
Max: 100 points

Customization

The scoring model and outbound engine are intentionally opinionated. See docs/CUSTOMIZATION.md to adapt:

Atlas scoring weights (lib/scoring/scoring-config.json)
P0 title/department patterns (lib/scoring/scoring-config.json)
ICP definitions (config/outbound/icp-definitions.yaml)
Qualification rules and filters (config/outbound/qualification-rules.yaml)
Response templates and triggers (config/outbound/response-templates.yaml)
Routing rules and rep assignment (config/outbound/routing-rules.yaml)
Company context for LLM prompts (config/outbound/company-context/)

Integrations

Integration	Type	Purpose	Setup
Deepline CLI	Required	Company/contact enrichment + outbound gateway	deepline.com
PostgreSQL / Neon	Required	Primary datastore	neon.tech
OpenAI	Required (outbound)	LLM reply drafting (GPT-5-mini)	platform.openai.com/api-keys
Slack	Required (outbound)	Approval workflow	api.slack.com/apps
Upstash QStash	Required (outbound)	Undo-send message queue	console.upstash.com/qstash
Lemlist	Outbound provider	Email/LinkedIn campaigns	app.lemlist.com
SmartLead	Outbound provider	Email campaigns (via Deepline)	smartlead.ai
Instantly	Outbound provider	Email campaigns (via Deepline)	instantly.ai
HeyReach	Outbound provider	LinkedIn campaigns (via Deepline)	heyreach.io
Attio	Optional	CRM sync for qualified leads	attio.com
HubSpot	Optional	CRM sync	developers.hubspot.com
Trigger.dev	Optional	Scheduled qualification + polling	trigger.dev
Notion	Optional	Database sync	notion.so
Anthropic	Optional	AI qualification	console.anthropic.com
Exa	Optional	Web search enrichment fallback	exa.ai

Deployment

Vercel (recommended)

npm i -g vercel
vercel link
vercel env pull  # pulls env vars to .env.local

# Add QStash via Upstash marketplace (auto-provisions env vars):
vercel integration add upstash
# Or set manually in Vercel dashboard → Settings → Environment Variables:
#   QSTASH_TOKEN, QSTASH_CURRENT_SIGNING_KEY, QSTASH_NEXT_SIGNING_KEY

vercel deploy --prod

Trigger.dev (scheduled tasks)

npx trigger.dev@latest deploy

Documentation

docs/API.md — API routes, frontend pages, webhook auth, YAML config, scheduled tasks, database schema, security details
docs/SCORING_MODEL.md — Full Atlas algorithm documentation (backtested example)
docs/CUSTOMIZATION.md — How to adapt scoring, ICP, and outbound config
docs/DATABASE_SCHEMA.md — Complete schema reference
docs/INTEGRATIONS.md — Integration setup guides
Deepline — The enrichment engine that powers contact and company data

Keywords

gtm go-to-market signal-scoring outbound lead-generation lead-qualification sales-automation account-intelligence icp-scoring deepline lemlist smartlead instantly heyreach slack attio hubspot crm ai-reply llm qstash next.js postgresql neon

License

MIT — see LICENSE.

Built with Deepline

Name		Name	Last commit message	Last commit date
Latest commit History 115 Commits
api		api
app		app
components		components
config		config
docs		docs
lib		lib
public		public
scripts		scripts
src		src
tests		tests
trigger		trigger
types		types
.gitignore		.gitignore
.vercelignore		.vercelignore
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
README.md		README.md
env.example		env.example
eslint.config.mjs		eslint.config.mjs
next.config.ts		next.config.ts
package-lock.json		package-lock.json
package.json		package.json
postcss.config.mjs		postcss.config.mjs
proxy.ts		proxy.ts
slack-app-manifest-springcash.json		slack-app-manifest-springcash.json
slack-app-manifest.json		slack-app-manifest.json
trigger.config.ts		trigger.config.ts
tsconfig.json		tsconfig.json
vercel.json		vercel.json
vitest.config.ts		vitest.config.ts

Folders and files

Latest commit

History

Repository files navigation

GTM Signal Scoring

What is this?

Features

Signal Scoring (Atlas Algorithm)

Outbound Reply Engine

Lead Qualification Pipeline

Safety & Security

Architecture

Quick Start

Prerequisites

Install

Run Database Migrations

Set Up QStash (Undo-Send Queue)

Start Development

Environment Variables

Required

Outbound Engine

QStash (Undo-Send Queue)

Security

AI & Search

Optional Integrations

Scoring Model

Customization

Integrations

Deployment

Vercel (recommended)

Trigger.dev (scheduled tasks)

Documentation

Keywords

License

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages