Bops
A team of AI bots that runs your business ops: inbox, pipeline, invoices, reports. Each one has its own computer, email and phone number, and remembers everything.
翻译准备中
Facil — Scalable AI-Powered Government Digital Services Framework. Document intelligence, workflow automation, compliance monitoring. Currently deployed for Equatorial Guinea.
翻译准备中
正在加载完整 README。 以下为已保存的摘录。
Powering digital transformation for government procedures, workflows, and fiscal services — deployable to any country.
Currently deployed for the Republic of Equatorial Guinea.
Facil is a full-stack, AI-powered framework for digitizing government administrative procedures. It provides a complete platform — backend API, web dashboard, citizen mobile app, and field inspection app — that any government can configure to manage its own fiscal services, workflows, document processing, and citizen interactions.查看完整 README ↗Powering digital transformation for government procedures, workflows, and fiscal services — deployable to any country.
Currently deployed for the Republic of Equatorial Guinea.
Facil is a full-stack, AI-powered framework for digitizing government administrative procedures. It provides a complete platform — backend API, web dashboard, citizen mobile app, and field inspection app — that any government can configure to manage its own fiscal services, workflows, document processing, and citizen interactions.
The framework is country-agnostic by design: workflows, fiscal services, entities, roles, translations, and business rules are all data-driven and configurable through the database and admin UI — not hardcoded. Deploying Facil to a new country means configuring its services, entities, and workflows, not rewriting code.
Current deployment: Republic of Equatorial Guinea — 873 fiscal services, 21 ministries, 20 government entities, 36 business workflows, trilingual (Spanish, French, English).
End-to-end digitization of administrative procedures. 36 predefined workflows (passport, residency, driving license, contracts, inspections...) with configurable steps, dynamic form validation, conditional routing, and multi-entity approval chains. Citizens submit requests through guided wizards; government agents process them via role-based dashboards with SLA-driven priority queues, automatic assignment, and real-time workload balancing across 20 government entities.
A RAG chatbot powered by Gemini 2.5 Flash with role-based specialized agents — each agent has domain-specific tools and access controls respecting strict data confidentiality:
| Agent | Scope | Tools | Confidentiality |
|---|---|---|---|
| Citizen Agent | Public services, fees, procedures | 19 public tools | No access to internal data |
| Authenticated Agent | Personal requests, status tracking | 8 auth tools | Own data only |
| Treasury Agent | Payment validation, revenue analysis | Deep reasoning | Entity-scoped financial data |
| Supervisor Agent | Team performance, anomaly detection | Deep reasoning | Entity-scoped agent metrics |
| Admin Agent | System configuration, audit analysis | Deep reasoning | Full system access with audit trail |
Each agent operates with Chain-of-Thought reasoning (6-step prompt), self-reflection (auto-regeneration if quality score < 5/10), and 15 response formats.
Convert physical documents (contracts, invoices, identity papers, government forms, tax declarations) into structured digital data. The OCR pipeline extracts, validates, and structures data from 39+ document schemas using template-based field mapping with bounding box annotation, AI-powered classification, and cross-document validation. Extract structured text from PDFs while preserving layouts.
Every agent action, payment validation, document review, and permission change is logged in the audit trail (2,800+ entries). The SchemaValidationEngine enforces document integrity with 70+ JSON rules. RiskAnalyzer runs a 12-step assessment pipeline on every document. All API requests pass through rate limiting, JWT validation, RBAC check (335 permissions), and parameterized SQL — ensuring OWASP compliance at every layer.
AI-driven analysis: detect recurring patterns across scanned documents, validate data by cross-referencing multiple sources (OCR extraction vs form input vs database records), and enable enterprise-grade search across the fiscal services catalog (873 services, 21 ministries). Automatic zone resolution with Levenshtein fuzzy matching and commerce type classification for commercial license workflows.
| Metric | Value |
|---|---|
| Fiscal Services Catalog | 873 (869 active) |
| Government Entities | 20 |
| Business Workflows | 36 predefined |
| Database Tables | 145 |
| Database Enums | 50 |
| Roles | 47 (335 granular permissions) |
| OCR Document Schemas | 39 |
| Languages | 3 (ES, FR, EN — extensible) |
| Backend Modules | 30 |
| Frontend Modules | 41+ |
| Communication Templates | 85 (29 email, 31 SMS, 9 push, 16 in-app) |
| Audit Log Entries | 2,800+ |
| Total Commits | 3,800+ |
以上为 README 的开头部分。其余内容请阅读原文。 查看完整 README ↗