CCT - AI Trading Intelligence (Cloudflare Workers)

Overview

CCT is a Cloudflare Workers codebase for AI-assisted market intelligence: sentiment analysis, report generation, dashboards, and operational tooling. The core worker lives in src/, Durable Objects are in src/modules/, and the public UI is in public/.

Key characteristics:

• Dual-model sentiment pipeline (primary GPT-OSS 120B, secondary DeepSeek-R1 32B)
• Durable Objects cache as the primary fast path
• D1-backed job storage for reports and runs (with KV fallback)
• Multi-run tracking across scheduled job types
• Polling-based dashboards (SSE disabled)

Project Layout

• src/ - Worker handlers and route modules
• src/routes/ - API v1 route handlers by domain
• src/modules/ - Durable Objects and shared services
• public/ - Dashboard assets and static UI
• docs/ - Guides, architecture notes, and references
• tests/ - Playwright specs and test scripts
• scripts/ - Operational and utility scripts

API Quick Start

Base path for APIs is /api/v1. Route groups are organized by domain:

• Sentiment: /api/v1/sentiment/*
• Reports: /api/v1/reports/*
• Data: /api/v1/data/*
• Jobs: /api/v1/jobs/*
• Sector rotation: /api/v1/sector-rotation/* and /api/v1/sectors/*
• Market intelligence: /api/v1/market-intelligence/*
• Market drivers: /api/v1/market-drivers/*
• Predictive analytics: /api/v1/predictive/*
• Technical analysis: /api/v1/technical/*
• Advanced analytics: /api/v1/analytics/*
• Realtime: /api/v1/realtime/*
• Backtesting: /api/v1/backtesting/*
• Portfolio: /api/v1/portfolio/*
• Risk management: /api/v1/risk/*
• Cache: /api/v1/cache/*
• Production guards: /api/v1/guards/*

For a complete list, see src/routes/ and the API root response at /api/v1.

Authentication

• If X_API_KEY is configured in the environment, non-public /api/v1/* endpoints require X-API-Key.
• Dashboards use public/js/cct-api.js, which looks for the key in sessionStorage.cct_api_key, then localStorage.cct_api_key, then window.CCT_API_KEY.
• Public endpoints include /health, /model-health, /api/v1/data/health, /api/v1/reports/status, /api/v1/jobs/history, and /api/v1/jobs/runs.

Development

npm install
npm run dev

Deployment and testing:

• npm run deploy
• npm run deploy:frontend or npm run deploy:frontend:only
• npm run test:playwright
• npm run test:performance or npm run test:workflows

Wrangler auth uses OAuth for this repo:

unset CLOUDFLARE_API_TOKEN && npx wrangler <command>

Documentation

Start with docs/INDEX.md, then follow the relevant guide:

• docs/guides/USER_GUIDE.md
• docs/guides/DEPLOYMENT_GUIDE.md
• docs/guides/MAINTENANCE_GUIDE.md
• API_DOCUMENTATION.md

Data and Scheduling Notes

• scheduled_job_results.scheduled_date is a write-time tag; the frontend ?date=YYYY-MM-DD is passed through verbatim.
• Multi-run tracking is used across job types with startJobRun() and completeJobRun().
• Prediction job storage is in D1 (job_executions, symbol_predictions, daily_analysis) with KV fallback.

Model Policy

• Primary model: @cf/openai/gpt-oss-120b
• Secondary model: @cf/deepseek-ai/deepseek-r1-distill-qwen-32b
• Deprecated: @cf/openchat/openchat-3.5-0106, @cf/aisingapore/gemma-sea-lion-v4-27b-it

Last Updated: 2026-02-02