How bakulya014-dot/tumar-ai works
Tumar.AI is a multilingual cybersecurity assistant. In the main user workflow, a person submits a message, URL, email, or image through the Streamlit app; the Trust Engine routes it to an analyzer, which produces a shared TrustReport that the UI renders and may retain in session history. Other product branches include breach checks, dashboard/history, optional sample contributions, and a REST API with customer-key management. The README documents additional password-health and API capabilities; their implementation is not fully represented in the sampled excerpts, so those links are documented rather than source-verified.
Main entry points: trust_engine.py Trust Engine · views Analysis pages · main.py REST API · scam_analysis.py Scam analysis
Components 26
User experience
- Streamlit app
app.py - Analysis pages
views - Dashboard
views/home.py - Translations
i18n/__init__.py - UI components
components.py
Analysis platform
- Trust Engine
services/trust_engine.py - Scam analysis
services/scam_analysis.py - Deterministic scanner
services/heuristics.py - URL scanner
services/url_scanner.py - Breach checker
services/breach_checker.py - Image analysis
services/vision_analysis.py - Password health
services - TrustReport schema
services/trust_report.py
AI integrations
- Provider registry
services/providers/__init__.py - Gemini client
services/gemini_client.py - Compatible providers
services/providers/openai_compatible.py
REST and contributions
- Contribution consent
contribution_ui.py - REST API
api/main.py - Customer keys
api/keys.py - Key store
api/keystore.py - Detection corpus
api/corpus_store.py
Reports and history
- Report renderer
report_view.py - Session report history
report_store.py
Other components
- User
- AI providers
- Have I Been Pwned
Connections 35
- User → Streamlit app: submits checks
- Streamlit app → Analysis pages: routes to
- Analysis pages → Trust Engine: analyzes subjects
views/home.py - Analysis pages → Breach checker: checks emails
views/breach_checker.py - Analysis pages → Image analysis: analyzes images
views/image_scanner.py - Trust Engine → Scam analysis: dispatches to
services/trust_engine.py - Trust Engine → URL scanner: dispatches to
services/trust_engine.py - Trust Engine → Breach checker: dispatches to
services/trust_engine.py - Trust Engine → Image analysis: dispatches to
services/trust_engine.py - Trust Engine → Password health: supports declared kind
services/trust_engine.py - Scam analysis → Deterministic scanner: scans message
services/scam_analysis.py - Scam analysis → Provider registry: requests provider
services/scam_analysis.py - URL scanner → Gemini client: requests analysis
services/url_scanner.py - Image analysis → Provider registry: requests vision
services/vision_analysis.py - Provider registry → Compatible providers: builds provider
services/providers/__init__.py - Provider registry → Gemini client: builds Gemini
- Gemini client → AI providers: sends prompts
services/gemini_client.py - Compatible providers → AI providers: sends requests
services/providers/openai_compatible.py - Scam analysis → TrustReport schema: creates report
services/trust_report.py - Trust Engine → TrustReport schema: returns report
services/trust_engine.py - Analysis pages → Report renderer: renders report
views/image_scanner.py - Report renderer → UI components: composes components
report_view.py - Report renderer → Translations: translates labels
report_view.py - Analysis pages → Session report history: stores reports
views/home.py - Dashboard → Session report history: summarizes history
views/home.py - Analysis pages → Contribution consent: offers contribution
views/home.py - Contribution consent → Detection corpus: stores samples
contribution_ui.py - Contribution consent → REST API: optionally submits
contribution_ui.py - User → REST API: sends API request
- REST API → Trust Engine: analyzes request
api/main.py - REST API → Detection corpus: handles contributions
api/main.py - Customer keys → Key store: manages keys
api/keys.py - REST API → Key store: checks and meters
api/main.py - Breach checker → Have I Been Pwned: checks range
- Password health → Have I Been Pwned: checks hash prefix