CRM-Lite — Lightweight SQLite CRM for Small Business

Overview

A minimal, file-based CRM built on SQLite for non-technical small business owners. No server, no external dependencies, no auth complexity. Designed for agent-mediated use: the owner interacts via text message with the agent, which reads/writes this database.

When to Use

• Small business (1-10 employees) with no existing CRM
• Agent-mediated access — owners interact via chat/Telegram, not direct DB access
• Lead tracking from website forms, referrals, email inquiries
• Pipeline visibility — estimate sent → accepted/rejected → job scheduled → paid
• Communication log — every email, call, text logged automatically

Don't Use For

• Multi-user concurrent access with role-based permissions
• Complex sales methodologies (MEDDIC, Challenger, etc.)
• Real-time collaboration features
• Integration with enterprise SSO/SCIM

Core Features

Data Model (SQLite: crm.db)

1. customers — Core customer records
   • id (PK), name, email, phone, company, address, city, state, zip
   • lead_source, lead_status (new/qualified/estimate_sent/won/lost)
   • total_value, notes, created_at, updated_at
2. leads — Inbound leads before qualification
   • id, source (website/referral/email/phone), contact_name
   • contact_email, contact_phone, company, property_address
   • sqft, ceiling_height, service_type, urgency
   • status (new/qualified/estimate_ready/scheduled/lost)
   • assigned_to, created_at, updated_at
3. activities — Every touchpoint
   • id, customer_id (FK), lead_id (FK), activity_type (call/email/text/meeting/estimate_sent/estimate_accepted/estimate_rejected/job_scheduled/job_completed/invoice_sent/payment_received)
   • description, outcome, next_action, next_action_date
   • created_at, updated_at
4. estimates — Estimate lifecycle
   • id, customer_id, lead_id, estimate_number, amount
   • status (draft/sent/accepted/rejected/expired)
   • sent_date, accepted_date, rejected_date, expires_date
   • line_items (JSON), notes, created_at, updated_at
5. communications — Message log
   • id, customer_id, lead_id, direction (inbound/outbound)
   • channel (email/sms/telegram/phone), subject, body
   • sent_at, created_at
6. tasks — Action items for owners
   • id, customer_id, lead_id, title, description
   • due_date, priority (low/medium/high/urgent), status (pending/in_progress/done)
   • assigned_to, created_at, updated_at

Quick Start

import sqlite3
from pathlib import Path

DB_PATH = Path("~/.hermes/crm/crm.db").expanduser()
DB_PATH.parent.mkdir(parents=True, exist_ok=True)

def init_db():
    conn = sqlite3.connect(DB_PATH)
    conn.execute("""
        CREATE TABLE IF NOT EXISTS customers (
            id INTEGER PRIMARY KEY AUTOINCREMENT,
            name TEXT NOT NULL, email TEXT, phone TEXT,
            company TEXT, address TEXT, city TEXT, state TEXT, zip TEXT,
            lead_source TEXT, lead_status TEXT DEFAULT 'new',
            total_value REAL DEFAULT 0, notes TEXT,
            created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
            updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
        )
    """)
    # ... other tables
    conn.commit()
    return conn

Agent Integration Patterns

Create Lead from Website Form

User: "New lead from website: John Smith, john@email.com, 555-555-0123, 123 Main St [CITY] [STATE] [ZIP], 2500 sqft, service request"
Agent: Creates lead record, sets status=qualified, creates task for the owner to review

Log Estimate Sent

User: "Sent estimate #EST-0001-0042 to John Smith for $2,850"
Agent: Updates estimate status=sent, creates activity, creates task for follow-up in 3 days

Weekly Pipeline Report (Cron)

Agent: Queries all leads/customers, summarizes by stage, sends to the owner via Telegram

Common Pitfalls

1. Forgetting to create indexes — Add indexes on customers.email, leads.status, activities.customer_id for query speed
2. Not backing up — SQLite file is the entire DB; copy to S3/Drive daily via cron
3. Schema drift — Use migrations (numbered SQL files) rather than ad-hoc ALTER TABLE
4. Concurrent writes — SQLite handles concurrent reads, but writes serialize; keep transactions short

Verification Checklist

• crm.db created with all 6 tables
• Indexes on foreign keys and query columns
• Agent can: create lead, log activity, update estimate status, create task
• Weekly report cron outputs: new leads, estimates sent, conversion rate, pipeline value
• Backup script copies DB to offsite location daily

One-Shot Recipes

"New lead from referral"

conn = init_db()
lead_id = conn.execute("""
    INSERT INTO leads (source, contact_name, contact_email, contact_phone, company, property_address, sqft, service_type, status)
    VALUES ('referral', 'Jane Doe', 'jane@example.com', '555-555-0199', 'Example Realty', '456 Palm Ave [CITY] [STATE] [ZIP]', 3200, 'service_request', 'qualified')
""").lastrowid
conn.commit()

"Get pipeline summary for weekly report"

summary = conn.execute("""
    SELECT
        COUNT(CASE WHEN lead_status='new' THEN 1 END) as new_leads,
        COUNT(CASE WHEN lead_status='estimate_sent' THEN 1 END) as estimates_out,
        COUNT(CASE WHEN lead_status='won' THEN 1 END) as won,
        SUM(CASE WHEN lead_status='won' THEN total_value ELSE 0 END) as revenue
    FROM customers
""").fetchone()