Time Zero v0.1 POC

Semantic Traversal & Origin Verification Engine

[SYS] System initialized. Running in Sandbox/Free-Tier mode.
Ephemeral Trace Graph:
Time Zero
Private FB Group
Format: Image/Screenshot
OCR: Bypassed (Cached Hash)
T-35 Days
→
Amplifier 1
Telegram Channel
Network: Blood Tribe
Format: Text
T-31 Days
→
Bridge
X Influencer
1M+ Followers
Format: Quote Tweet
T-12 Days
→
Mainstream Injection
VP Candidate
National Debate
Format: Live Audio/Transcript
T-0 Days

Actual Source Rating (Time Zero)

Institutional
0.05
Fringe Bias
0.95
Bot Network
0.80

System evaluates THIS source.

Launderer Rating (Mainstream)

Institutional
0.90
Fringe Bias
0.10
Bot Network
0.00

Standard systems evaluate THIS source. (Ignored)

Production Implementation Details

This POC currently mocks the heavy data retrieval. To transition to a live, accurate engine, the following architecture must be deployed.

Layer Free Tier (Current) Production / Paid Integration
Schema Extraction Client-side Regex/Mock OpenAI gpt-4o-mini (Cost: $0.15/1M input tokens)
Vector Similarity Local string matching Sentence-BERT embeddings + Pinecone Serverless
Surface Web Crawl Hardcoded trace graph GDELT Project (Free/S3) + Common Crawl
Dark/Fringe Crawl Hardcoded trace graph Telethon API (Telegram), 4plebs API, Pushshift (Reddit)
Screenshot Wall Bypassed Perceptual Hashing (Local) -> Gemini 1.5 Flash Vision (Triage OCR)

Security Architecture: API Key Protection

Never expose keys to the client. The frontend (this UI) will not make direct calls to OpenAI, Pinecone, or Dataminr.

Implementation: Build a Backend-for-Frontend (BFF) using Node.js or Python (FastAPI). The browser sends `{ "claim": "..." }` to your backend route `POST /api/trace`. The backend server securely reads `.env` variables, orchestrates the LLM schema extraction, vector DB query, and web crawls, structures the Trace Graph JSON, and returns it to the client. Rate limiting (e.g., via Upstash/Redis) is applied at the backend to prevent API abuse.