99 lines
6.6 KiB
Markdown
99 lines
6.6 KiB
Markdown
# OpenForce Benchmark Analysis: CRM Architectures, Custom Object Engines & Visual Workflow Editors
|
|
|
|
## 1. Executive Summary
|
|
This document establishes architectural benchmarks, technical design trade-offs, and proven implementation patterns for **OpenForce**—a Python/Django + Vue 3 open-source CRM targeting Salesforce parity.
|
|
|
|
Key benchmark insights:
|
|
1. **Custom Object Storage**: Salesforce utilizes a multi-tenant universal table architecture (MT_Objects, MT_Fields, MT_Value_Data). Modern PostgreSQL allows us to achieve superior developer ergonomics and query performance via a **Hybrid JSONB + Typed Auxiliary Indexes** architecture (as proven by Twenty CRM and Supabase dynamic schema patterns).
|
|
2. **Visual Workflow Canvas**: `@vue-flow/core` provides native reactivity and customizable node sockets that match the UX paradigms of Salesforce Flow Builder, Zapier, and n8n.
|
|
3. **Execution Runtime**: A lightweight Directed Acyclic Graph (DAG) state machine running atop Celery/Redis guarantees idempotency, branch isolation, and transactional step retries.
|
|
|
|
---
|
|
|
|
## 2. Comparative Analysis of Open-Source CRM Architectures
|
|
|
|
| Dimension | Salesforce (Proprietary) | Twenty CRM (TypeScript/Nest/Postgres) | SuiteCRM 8 (PHP/Symfony) | OpenForce (Proposed: Django/Vue) |
|
|
| :--- | :--- | :--- | :--- | :--- |
|
|
| **Object Engine** | Proprietary Force.com multi-tenant UDM (EAV variant) | Dynamic PostgreSQL tables generated per workspace via TypeORM metadata | Traditional relational EAV (`*_cstm` tables per module) | **Hybrid PostgreSQL JSONB with GIN index + Metamodel Validation** |
|
|
| **Schema Flexibility** | High (Custom fields, relationships, validation rules) | High (Dynamic schema migrations per workspace) | Medium (Alters database schema on field creation) | **High (Instant zero-migration schema updates via JSONB)** |
|
|
| **Workflow Engine** | Salesforce Flow / Process Builder (Apex engine) | Action trigger workflows (Server-side TS) | Logic Hooks + WorkFlow module (Cron/Event triggers) | **DAG Execution Engine with Celery task orchestration** |
|
|
| **Frontend Framework** | Lightning Web Components (LWC) | React + Tailwind + Vite | Angular + Legacy Smarty/Vue | **Vue 3 (Composition API) + Pinia + TailwindCSS** |
|
|
| **API Architecture** | REST / SOAP / GraphQL / Bulk API | GraphQL first + REST | REST API v8 (JSON:API spec) | **Django REST Framework (DRF) + OpenAPI v3 / Dynamic ViewSets** |
|
|
|
|
---
|
|
|
|
## 3. Deep-Dive: Custom Object & Dynamic Field Engine
|
|
|
|
Salesforce's Force.com architecture avoids running `ALTER TABLE` whenever an administrator creates a custom object or field. Instead, it relies on metadata tables describing tenant fields and a massive denormalized storage layer.
|
|
|
|
### 3.1 Evaluated Architecture Patterns
|
|
|
|
#### Pattern A: Strict EAV (Entity-Attribute-Value)
|
|
- **Mechanism**: Table `CustomValues (record_id, field_id, value_string, value_int, ...)`.
|
|
- **Pros**: Pure relational modeling, strict SQL types.
|
|
- **Cons**: Massive JOIN overhead when querying 20+ fields per record; poor sorting/filtering performance across dynamic fields.
|
|
|
|
#### Pattern B: Dynamic DDL (`ALTER TABLE` on the fly)
|
|
- **Mechanism**: Django dynamically creates and executes `ALTER TABLE my_table ADD COLUMN custom_field VARCHAR`.
|
|
- **Pros**: Native DB columns, native indexes, simple SQL.
|
|
- **Cons**: DB lock contention, DDL transaction risks, migration tracking complexity across production instances.
|
|
|
|
#### Pattern C: Hybrid JSONB + Metadata Metamodel (Recommended for OpenForce)
|
|
- **Mechanism**:
|
|
- Standard relational tables for fixed objects (`Account`, `Contact`, `Lead`, `Opportunity`).
|
|
- Metamodel tables: `CustomObjectDefinition` and `CustomFieldDefinition`.
|
|
- Data table: `CustomRecord` with `(id, object_definition_id, data JSONB, created_at, updated_at, created_by)`.
|
|
- Native PostgreSQL GIN indexes on `data`: `CREATE INDEX idx_custom_record_gin ON custom_records USING gin (data jsonb_path_ops);`.
|
|
- Dynamic virtual serialization in DRF with JSONSchema validation at ingestion.
|
|
- **Pros**:
|
|
- Zero-downtime dynamic schema updates.
|
|
- Sub-millisecond JSONB query performance via GIN indexing and expression indexes for frequent filters (e.g. `CREATE INDEX idx_status ON custom_records ((data->>'status'));`).
|
|
- Seamless support for all Salesforce data types (picklist, multiselect, lookup, currency, formula).
|
|
|
|
---
|
|
|
|
## 4. Visual Workflow Automation Engine Benchmark
|
|
|
|
### 4.1 Frontend Canvas Benchmark: Vue Flow vs. Alternatives
|
|
- **Selected: `@vue-flow/core` (v1.30+)**:
|
|
- Direct Vue 3 port inspired by React Flow.
|
|
- Native Vue reactivity: Custom node components can be pure `.vue` Single File Components (SFCs).
|
|
- Out-of-the-box support for interactive controls, minimap, background grid, smooth bezier edges, and multi-socket connection validation.
|
|
- UX Pattern matches Salesforce Flow Builder:
|
|
- **Palette Drawer**: Drag triggers, logic nodes, and action blocks into viewport.
|
|
- **Step Inspector Side-Panel**: Contextual form for editing trigger conditions or email action parameters when a node is selected.
|
|
- **Live Graph Validation**: Visual feedback if branches are orphaned or cycles are detected.
|
|
|
|
### 4.2 Backend Execution Runtime (DAG Engine)
|
|
- **Workflow Representation**: Stored as a JSON DAG:
|
|
```json
|
|
{
|
|
"nodes": [
|
|
{"id": "node_1", "type": "trigger", "data": {"event": "lead.created"}},
|
|
{"id": "node_2", "type": "condition", "data": {"rules": [{"field": "lead_score", "op": "gte", "value": 50}]}},
|
|
{"id": "node_3", "type": "action", "data": {"action": "email.send_template", "template_id": 4}}
|
|
],
|
|
"edges": [
|
|
{"source": "node_1", "target": "node_2"},
|
|
{"source": "node_2", "target": "node_3", "sourceHandle": "true"}
|
|
]
|
|
}
|
|
```
|
|
- **Execution Model**:
|
|
- Synchronous / Celery worker dispatches flow runs on Django model signals (`post_save`).
|
|
- Topologically sorted step resolution with branch isolation.
|
|
- Idempotency keys generated per event trigger to prevent duplicate execution.
|
|
|
|
---
|
|
|
|
## 5. Standard CRM Life-Cycle Recommendations
|
|
|
|
1. **Lead Conversion Flow**:
|
|
- Atomic transaction converting `Lead(status='Qualified')` -> creates `Account` (or merges into existing), creates `Contact`, and optionally creates `Opportunity`.
|
|
- Copies custom fields from Lead to matching target fields on Account/Contact/Opportunity via field mapping metadata.
|
|
2. **Polymorphic Activity Timeline**:
|
|
- Django `GenericForeignKey` to associate `Tasks`, `Notes`, `Emails`, and `StageHistory` with any model (`Lead`, `Opportunity`, or `CustomRecord`).
|
|
3. **Email Outreach Engine**:
|
|
- Queue-backed (Celery) bulk sending with rate-limiting.
|
|
- Template engine using Django templating or Jinja2 syntax (`{{ contact.first_name }}`) with fallback defaults.
|
|
|