CRM Automator: Modular Agentic CRM Toolkit

Author: Trung Le
Team: RealTimeX.ai
Repository: https://github.com/therealtimex/crm-automator

Overview

The CRM Automator is a modular, agentic toolkit designed to transform unstructured data (emails, transcripts, documents) into structured CRM records. Unlike monolithic automation scripts, this toolkit is built as a collection of reusable Tools that can be orchestrated by AI agents or run as standalone pipelines.

Current Status: Production-Ready (v2.0)

Key Features

🧩 Modular Architecture: Decoupled Ingestion (EML/Text), Intelligence (LLM Extraction), and Integration (CRM Sync).
🤖 Agent-Ready: Components like IntelligenceLayer and RealTimeXClient are designed to be imported as tools for frameworks like LangChain or Autogen.
🛡️ Idempotent & Safe: Implements "Search-before-Update" patterns and SQLite-based state tracking to prevent duplicate records.
🧹 Smart Cleaning Pipeline: Automatically converts bloaty HTML into clean Markdown, strips noise, and resolves tracking links (e.g., unwraps Proofpoint/Safelinks and resolves HubSpot redirects) to ensure accurate URL extraction.
🧠 OpenAI-Compatible: Works independently with OpenAI, Anthropic, or local LLMs (via LM Studio/Ollama) using instructor.

Installation

Clone the repository:

git clone https://github.com/therealtimex/crm-automator.git
cd crm-automator

Install dependencies:

# Option A: Standard pip
pip install -r requirements.txt

# Option B: Modern uv (Recommended)
uv sync

Configure environment:

cp .env.example .env
# Edit .env with your CRM and LLM credentials

Usage

1. Process Emails (`.eml`)

Run the automator on a single EML file or an entire directory recursively.

Using uvx (Fastest, zero-install):

uvx eml/eml_automator.py "path/to/email.eml" --env-file ".env"

Using a directory:

uv run python eml/eml_automator.py "path/to/directory" --env-file ".env"

Options:

--force, -f: Force re-processing of an email even if it was already synced.
--verbose, -v: Enable debug logs to see HTTP requests and LLM thought process.

2. Generic Agent Demo

See how the toolkit handles raw text (e.g., meeting transcripts):

python3 eml/agent_demo.py --api-key "your_key"

Architecture

graph TD
    Ingest[Ingestion Layer] --> Intelligence[Intelligence Layer]
    Intelligence <--> Search[Web Search Tool]
    Intelligence --> Client[CRM Client]
    Client --> CRM[(RealTimeX CRM)]

crm_client.py: The "hands" – handles all API calls with retries and 10s timeouts.
intelligence.py: The "brain" – uses Pydantic models to extract structured data from text.
persistence.py: The "memory" – prevents processing the same resource twice.

Built with ❤️ by the RealTimeX.ai Team

Metadata

Release files for crm-automator 1.11.5

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for crm-automator 1.11.5
File	Size	Uploaded
crm_automator-1.11.5.tar.gz	75.0 kB	Jan 13, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for crm-automator 1.11.5
File	Interpreter	ABI	Platform	Reset
crm_automator-1.11.5-py3-none-any.whl 87.1 kB Jan 13, 2026	Python 3	none	any	Details

Total release size: 162.1 kB

Release files / crm_automator-1.11.5.tar.gz

Download URL	crm_automator-1.11.5.tar.gz
Size	75.0 kB
Tags	Source
SHA-256 checksum How to use checksums	`ce18ca82606a4cb67ef46196e69925c1df3347dd49b4ac784eb003d509d7ef05`
BLAKE2b-256 checksum How to use checksums	`f99a13621653d0585cf28d1da09b53651120b22c20a361a8fc4c115884a7cc20`
Upload date	Jan 13, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/6.2.0 CPython/3.10.19`

Release files / crm_automator-1.11.5-py3-none-any.whl

Download URL	crm_automator-1.11.5-py3-none-any.whl
Size	87.1 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`62f7e90322e12bb3d979383fc106d343b0bbc6ed89b838cef6523ffc942c1dff`
BLAKE2b-256 checksum How to use checksums	`380a4b46ae40afe470ce9b6e7043613b011ffbdd68b4f7070596fe34d0aa55c2`
Upload date	Jan 13, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/6.2.0 CPython/3.10.19`