An n8n workflow that acts as a file-retrieval agent for Bale (a Telegram-like Iranian messaging app). Type a natural-language request — e.g. "grab the devops PDFs from this channel" — and the workflow parses the intent with an LLM, fetches matching files from a Bale channel via its bot API, and saves them to disk.

Built as a take-home task for a job interview, focused on demonstrating how to scope an LLM's role narrowly (intent parsing only) while keeping the rest of the pipeline — fetching, filtering, downloading — as plain, deterministic logic.

See inside the folder for:

SETUP.md — how to stand it up (Bale bot/channel creation, n8n import, credentials, known gotchas)
HOW_IT_WORKS.md — architecture and data flow, step by step
NODE_DESCRIPTIONS.md — talking points for walking through the workflow live

`.freebuff`

(add a short description here once you're ready — leaving as a placeholder for now)

Notes

Workflow exports and configs in this repo have credentials and IDs replaced with placeholders (e.g. YOUR_BOT_ID) before committing. Anyone running a workflow locally needs to drop their own values back in.