How rapabru/re-ai-checker works
ReCheck verifies an AI-generated answer by selecting independent providers, querying them concurrently, aggregating their verdicts into a confidence/consensus result, and presenting or saving a report. Users can start this workflow from the CLI or from buttons injected into supported AI chat pages; the browser extension proxies verification requests to the local API, which also exposes a health check. Provider adapters parse structured model responses into shared results. The sampled source supports these main connections; some CLI details and extension rendering logic are documented or only partially sampled.
Main entry points: server.py Local API · recheck.py CLI workflow · __init__.py Verifier selection · content.js Injected UI
Components 18
Entry points
- CLI workflow
recheck.py
Browser extension
- Injected UI
extension/content.js - Status popup
extension/popup.js - Request proxy
extension/background.js
Verification backend
- Local API
server.py
Model providers
- Verifier selection
providers/__init__.py - OpenAI adapter
providers/openai_provider.py - Anthropic adapter
providers/anthropic_provider.py - Gemini adapter
providers/gemini_provider.py - Result parsing
providers/base.py
Results and reports
- Consensus and reports
recheck.py
Other components
- CLI user
- Chat user
- AI chat pages
- OpenAI API
- Anthropic API
- Gemini API
- Saved reports
Connections 31
- CLI user → CLI workflow: provides input
- CLI workflow → Verifier selection: selects verifiers
- Verifier selection → OpenAI adapter: instantiates
- Verifier selection → Anthropic adapter: instantiates
- Verifier selection → Gemini adapter: instantiates
- CLI workflow → OpenAI adapter: runs verification
- CLI workflow → Anthropic adapter: runs verification
- CLI workflow → Gemini adapter: runs verification
- OpenAI adapter → OpenAI API: sends request
- Anthropic adapter → Anthropic API: sends request
- Gemini adapter → Gemini API: sends request
- OpenAI adapter → Result parsing: parses response
- Anthropic adapter → Result parsing: parses response
- Gemini adapter → Result parsing: parses response
- CLI workflow → Consensus and reports: summarizes results
- Consensus and reports → Saved reports: writes reports
- Chat user → AI chat pages: asks question
- AI chat pages → Injected UI: hosts injected UI
- Chat user → Injected UI: starts verification
- Injected UI → Request proxy: sends payload
- Request proxy → Local API: posts verification
- Local API → Verifier selection: selects verifiers
- Local API → OpenAI adapter: runs verification
- Local API → Anthropic adapter: runs verification
- Local API → Gemini adapter: runs verification
- Local API → Consensus and reports: summarizes results
- Local API → Saved reports: writes reports
- Local API → Request proxy: returns results
- Request proxy → Injected UI: returns results
- Injected UI → Chat user: shows report
- Status popup → Local API: checks health