AI Control Hub Development

We’re seeking a skilled Laravel + React full-stack developer to build a modular, API-based AI Control Hub using a microservice-ready architecture.
The system will initially run as a standalone microservice (Dockerized) but must remain flexible for future integration with our existing Laravel + Angular platform.

Users will log in via our main platform (OAuth/JWT) and automatically access this system based on their subscription tier.
Admins can create unlimited AI Suites and Agents, connect them to external APIs, and control which suites are visible or live for users.
Each AI Suite (Valuation, Compliance, Feasibility, etc.) will function as an independent module and support:

⦁ Web search + internal knowledge data
⦁ File uploads for contextual AI (RAG)
⦁ Multi-model AI calls (OpenAI, Gemini, Mistral, Anthropic)
⦁ PDF/DOCX report generation and download

The goal is a secure, scalable, and modular microservice — where new agents or AI models can be added later without heavy rework.

Note: Use a PingCRM-style Laravel + React (Inertia) setup, not Filament. The goal is an API-first, modular structure that stays flexible for future integration with our main Laravel + Angular platform.



Core Functionality

1. Authentication & Access Control

⦁ Integrate with existing main platform OAuth/JWT.
⦁ Auto-authenticate logged-in users via token (no separate login).
⦁ Role-based permissions using Spatie Laravel-Permission (Admin / User).
⦁ Admin toggle to make Suites “live” or inactive.


2. Admin Dashboard (Laravel + React)

Use a ready-made admin template (Filament, CoreUI, Vuexy, or Materio).

Admin can:

⦁ Create any number of Suites and Agents.
⦁ Assign Agents to Suites (e.g., Market, Valuation, Compliance).
⦁ Define prompt, model, access tier, and external API connections (Crunchbase, Google Patents, etc.).
⦁ Manage visibility — activate/deactivate suites anytime.
⦁ Track usage and token cost with summary charts (Chart.js / ApexCharts).


3. User Interface (React)

Clean, responsive chat interface using open-source React chat templates (Chat UI Kit React or React Chat Widget).

Users can:

⦁ Access available Suites (based on plan/tier).
⦁ Upload documents (PDF, DOCX, XLSX).
⦁ Get AI responses combining web search, uploaded files, and external data sources.
⦁ Download PDF/DOCX reports.
⦁ View recent chats (store history for 1 month).


4. Knowledge Integration & External Data APIs

Connect Agents to public or paid APIs such as:

⦁ Crunchbase (company data)
⦁ Google Patents (intellectual property)
⦁ Brave / Serper / Bing Search (web info)
⦁ AI responses should combine data from:
⦁ Uploaded documents (via vector DB)
⦁ Live web search
⦁ External data APIs (when configured per agent).


5. File Upload & Contextual AI (RAG)

⦁ File uploads: PDF, DOCX, XLSX.
⦁ Text extraction + embeddings stored in FAISS or Chroma.
⦁ Use OpenAI’s text-embedding-3-small for embeddings.
⦁ Isolate user data (no cross-user access).


6. Agent Logic & Workflow Chaining

⦁ Agents can be linked in sequence (e.g., Market → Valuation → Report).
⦁ Simple toggle or drag-list for order management.
⦁ Each Agent can have its own model (OpenAI, Mistral, Gemini, Anthropic, etc.).


7. AI Models

⦁ Default: gpt-3.5-turbo-0125
⦁ Embeddings: text-embedding-3-small
⦁ Structure backend so Admin can add new models later via config (no hardcoding), e.g mistral, gemini.
⦁ All model names must be stored in config or DB for easy updates if APIs change or deprecate.


8. Reports & Exports

⦁ Generate and export chat results as PDF or DOCX using DomPDF or PHPWord.
⦁ Store download history per user.


9. Security & Privacy

⦁ AES-256 encryption for keys and sensitive data.
⦁ HTTPS/TLS enforced; signed internal API requests.
⦁ Row-level data access and signed short-lived file URLs.
⦁ No third-party data logging or analytics.
⦁ GDPR/HIPAA-safe structure (no PHI exposure).



Architecture

Backend: Laravel 10 (API-based)

Frontend: React 18 (template-based)

Database: PostgreSQL

Vector DB: FAISS or Chroma

Auth: OAuth2 / JWT

Containerization: Docker + NGINX

Search APIs: Serper / Brave / Bing

Reports: DomPDF / PHPWord

Charts: Chart.js / ApexCharts



Deliverables

Complete Docker setup (docker-compose.yml, Dockerfiles, .env.example).

Full working source code (GitHub/GitLab repo or zip).

SQL schema + sample data.

Postman or Swagger API docs.

Setup guide with install commands & readme file.

Functional OpenAI integration demo.

Admin and user access for testing.



Development Guidelines

Must be API-first and Docker-ready.

Use templates and open-source UI for faster delivery (no custom design).

Backend and frontend should be framework-independent for easy Angular migration later.

Responsive, mobile-friendly interface.

NDA signing and progress updates every 2 days required.



Integration

User logs into main platform → verifies payment/tier → auto-accesses AI Hub via OAuth/JWT.

Admin can toggle suites between free, paid, or inactive.

Connect agents to external APIs (Crunchbase, Google Patents, etc.) via secure credentials stored in .env.


Final Outcome

A modular AI Control Hub (Laravel + React) that’s secure, scalable, and API-ready — where Admins can create unlimited Suites and Agents, connect them to any AI model or external data API, and control visibility or activation, with seamless integration potential into the main Laravel + Angular platform.