Anyone can generate a rebuttal; the useful part is the filter in front of it, which routes grief and illness to a gentle answer and conspiracy material to a refusal, so the sharp voice only fires where it belongs. Demo mode is on by default, so you can read what it writes before any account is involved.
A response-generation server for fact-checking in social messages. It classifies the incoming claim, picks a tone, and writes a short reply with a source attached. Sending the message is a separate Instagram server's job; this one writes what gets sent.
- A short fact-check reply with a citation and a deliberate voice, sized for a direct message (generate_sassy_response)
- Tone chosen automatically from the content: sharp for health myths, gentle for grief and illness, plain for professional contexts, and a polite refusal for conspiracy material
- A welcome message written for a new follower (generate_welcome_message)
- Practice claims to work against without connecting a real account, in the demo mode that is on by default (check_instagram_dms)
- A status readout showing whether the messaging side is actually wired up (instagram_integration_status)
- Interaction logging, for daily counts
An Anthropic API key. Demo mode needs nothing else. For live use you also run the separate Instagram messaging server and supply account credentials — and the project advises a dedicated throwaway account, since Instagram restricts accounts it detects as automated. Python, in a virtual environment.