The StFinder engine — 87 saints, 240 patronages, a curated synonym map — runs as a persistent service on this machine. Ask it anything; it matches a need to saints and holds a conversation with them.

The one call that matters

curl -s -X POST http://127.0.0.1:8787/api/chat \
  -H 'content-type: application/json' \
  -d '{"text":"THE NEED IN PLAIN WORDS"}'

Reply to the user with the text field as the bot’s answer, and offer the buttons as suggestions the user can pick (render them as a short list of the label values, or as inline buttons if the channel supports them).

Continuing a conversation with one saint

When the user picks a saint — the button data is s:<saintId> — open the conversation:

curl -s -X POST http://127.0.0.1:8787/api/callback \
  -H 'content-type: application/json' \
  -d '{"data":"s:anthony-padua"}'

Then every follow-up message goes to the same endpoint as a normal message, with a session key so the conversation is remembered:

curl -s -X POST "http://127.0.0.1:8787/api/chat?session=USER_ID" \
  -H 'content-type: application/json' \
  -d '{"text":"what do you look after"}'

Use a stable session per user (their Telegram id, or a hash of it) so conversations persist across turns.

The action buttons

Commands the bot understands

If the service is down

`data`	Effect
`q:pray`	the saint’s prayer, to pray together
`q:story`	the saint’s biography
`q:patronage`	what they are patron of
`q:feast`	their feast day
`q:reflect`	a reflection prompt
`b:back`	back to finding saints
`p:<prompt>`	an opening prompt — treat as the need itself

bash /Users/amre/Documents/StFind/scripts/deploy-saint-bot.sh

curl -s http://127.0.0.1:8787/api/health

Rules for the tone

The bot already speaks in the voice of the saints — plain, warm, scriptural, never clinical, never preachy. Relay its text essentially as written. Do not summarise it into corporate mush, do not add “I hope this helps”, and do not editorialize about the match. The match reasons it gives (“for Dentists, Teeth”) are the useful part — keep them.