Self-taught dev ports local voice assistant ARYA from n8n to LangGraph, opens architecture for review

Alternative_Yak_1367 · reddit · 2026-11-05

A self-taught developer migrated ARYA, a local-first voice and chat assistant for Windows, from n8n workflows to a LangGraph Python service and is requesting honest feedback before building further.

Architecture highlights

FastAPI + LangGraph core with streaming replies and speech I/O
Separate MCP tool servers for Gmail, Calendar, Contacts, Spotify, and desktop control
Native tools for weather, web search (Tavily), and a calculator
Approval prompts before risky actions like sending or deleting
A verifier that checks the assistant's action claims against actual tool results
PII tokenization before content reaches models or logs
PostgreSQL + pgvector for conversation checkpoints and memory (hybrid vector + full-text retrieval)

Testing: 135 unit tests, 35 PostgreSQL integration tests, plus small regression evals on gpt-4o-mini (tool routing 18/18, memory routing 46/46, approvals 6/6) — author notes these are small, self-written, and not accuracy claims.

Feedback sought on: whether the LangGraph state/checkpoint design fights the framework, failure modes of the approval flow, whether the verifier pattern is sound, how to structure the MCP servers, and what tests are missing. Known limitations include Windows-only support and small author-written evals.

Original post →

Self-taught dev ports local voice assistant ARYA from n8n to LangGraph, opens architecture for review

More from coding & agent