Cloud-Based Jarvis Automation Suite

Cliente Freelancer · Remoto · Remoto · freelance · mid · 3000–5000 GBP

Publicada el 2026-07-16

Descripción de la oferta

AI Agent Platform (Build Brief) About us HOTR is a creator management agency. We manage a roster of 15–19 digital content creators (£100k–137k/month gross) with a team of ~15 (operations, customer support, marketing, a ~10-person messaging team, and finance). We want our own private, cloud-hosted AI assistant for the agency — built on our infrastructure, owned entirely by us. What we need (v1 scope) A web-based chat bot with two specialist AI agents: 1. Acquisition agent Helps run our creator-signing pipeline: Track prospects and their stage (contacted → pitch sent → follow-up → signed / lost) Flag stalled leads (our biggest leak: prospects go cold after the pitch with no follow-up) Draft outreach and follow-up messages in our tone of voice Answer questions about the pipeline ("who's gone cold?", "what's stuck this week?") 2. Finance agent Answers questions on agency finances: Monthly gross / net revenue and trends Creator revenue splits (flat 50/50) and team compensation (hourly + % of net sales) Net-to-agency modelling, churn impact, cash runway Works from figures we upload/enter, plus stored history Both agents must know our agency context — team, roster, economics, goals, constraints — supplied as a knowledge base and retrieved automatically per question (RAG). They must answer using our real numbers, not generic advice. Core requirements Web chat UI (desktop + mobile browser), login-protected, multi-user (5–10 team members) Streaming replies (token-by-token, not a spinner followed by a wall of text) Per-agent system prompts editable by us via an admin screen — no redeploy to tune an agent Knowledge base: upload markdown/docs → chunked + embedded → semantic retrieval (pgvector) Conversation history per agent, persisted and searchable Cloud-hosted, always-on — must survive reboots/restarts without babysitting Secure secrets: all API keys server-side and encrypted; never exposed to the browser Multi-user auth with roles (admin vs team member) Suggested stack (open to your proposal — justify any changes) Frontend: React (Vite) or Next.js Backend: Node / TypeScript Database: Postgres (Supabase) + pgvector for embeddings LLM: Anthropic Claude API (streaming) Embeddings: OpenAI or Gemini Hosting: Railway / Render / Fly.io / VPS Auth: Supabase Auth or Clerk Deliverables Full source code in our GitHub org — all IP assigned to us (non-negotiable) Deployed, working cloud environment running on our accounts and our API keys Admin ability to edit agent prompts and upload/refresh the knowledge base Handover documentation: architecture, how to deploy, how to add a new agent 30 days of post-launch bug support Milestones (propose your own breakdown + price per milestone) Schema, auth, chat skeleton with streaming (one agent working end-to-end) Knowledge base: ingest + embeddings + retrieval, agency context loaded Both agents live with tuned prompts + conversation history Cloud deploy, admin screens, handover Optional extras — quote these SEPARATELY, do not bundle into the base price Revenue-platform data integration — live revenue and creator stats feeding the Finance agent (platform + API details shared with the selected developer) Voice — speak to the bot (speech-to-text + text-to-speech) Additional agents — Marketing, Ops, Sales, Mentor (same framework, new prompts) Notifications — Slack / Telegram / email alerts Questions we'll ask before awarding How would you implement RAG over our documents — chunking, embeddings, retrieval? How do you stream an LLM response to a React UI? How do you store per-tenant API keys securely? Show us an LLM-powered product you've shipped that real users pay for. How would you handle an OAuth integration (e.g. Google/Zoom) including token refresh? To bid, please include One LLM/agent product you've actually shipped (link or demo) Your approach to the knowledge base / RAG Fixed price + timeline for the base scope Separate prices for each optional extra Who is doing the work (you personally, or a team?) Please don't quote a price before reading this brief. Bids without an example of a shipped LLM product will not be considered.

Skills

Fuente original: freelancer