A GPT-Live voice companion for seniors with memory and back end A GitHub repository published by developer mcbuehler provides the backend and assistant configurations for Albert, a Vapi-based voice companion for seniors that uses Hindsight for long-term memory, answers inbound calls, places outbound and scheduled callbacks, archives call records, recordings and transcripts, and serves a staff page for managing customers and reviewing calls. The deployment runs on Python 3.12 with uv and requires Supabase, Hindsight Cloud, Vapi and Brave Search API accounts plus a compatible phone connection, with setup, environment variable, Vapi, deployment, architecture and known-production-issues documentation included; the code is released under the MIT License. The project credits OpenAI and Anthropic for API credits and the ETH Entrepreneur Club's RocketHub accelerator, and its backend started from Vercel's Flask example. Backend and assistant configurations for Albert , a Vapi https://vapi.ai/ voice assistant with long-term memory Hindsight https://ui.hindsight.vectorize.io . Albert answers inbound calls, places outbound and scheduled callbacks, archives call records and available recordings and transcripts, and provides a staff page to manage customers and review calls. Use uv https://docs.astral.sh/uv/ with Python 3.12. A working deployment needs Supabase https://supabase.com/dashboard , Hindsight Cloud https://ui.hindsight.vectorize.io , Vapi https://dashboard.vapi.ai/signup and Brave Search API https://api-dashboard.search.brave.com/ accounts, plus a compatible phone connection. See setup https://github.com/mcbuehler/albert/blob/main/docs/setup.md for account links, credentials, database and Vapi configuration. Review the known production issues https://github.com/mcbuehler/albert/blob/main/docs/known-production-issues.md before connecting real callers. git clone https://github.com/mcbuehler/albert.git cd albert uv sync --locked --python 3.12 cp .env.example .env After completing setup and filling in .env : uv run --locked --python 3.12 --env-file .env gunicorn main:app --bind 127.0.0.1:8000 --timeout 150 Open http://localhost:8000 for the staff page, then follow the smoke test https://github.com/mcbuehler/albert/blob/main/docs/vapi-setup.md smoke-test . - Setup and accounts https://github.com/mcbuehler/albert/blob/main/docs/setup.md · Environment variables https://github.com/mcbuehler/albert/blob/main/docs/configuration.md · Vapi setup https://github.com/mcbuehler/albert/blob/main/docs/vapi-setup.md - Deployment https://github.com/mcbuehler/albert/blob/main/docs/deployment.md · Development and tests https://github.com/mcbuehler/albert/blob/main/docs/development.md no provider accounts needed for tests - Architecture https://github.com/mcbuehler/albert/blob/main/docs/architecture.md · Call archive https://github.com/mcbuehler/albert/blob/main/docs/call-archive-schema.md · Known production issues https://github.com/mcbuehler/albert/blob/main/docs/known-production-issues.md Albert is available under the MIT License https://github.com/mcbuehler/albert/blob/main/LICENSE . Thank you to OpenAI and Anthropic for providing API credits, and to the ETH Entrepreneur Club for its RocketHub accelerator https://www.entrepreneur-club.org/rockethub . Thanks to everyone who used and tested Albert and helped us improve it. The backend started from Vercel's Flask example https://github.com/vercel/vercel/tree/main/examples/flask .