Roo Code Online‑Research & Knowledge‑Pipeline v2

A token‑efficient AI research & knowledge‑management stack for Windows powered by Roo Code custom modes and MCP servers.

---

🔍 What’s new in v2 ?

Area	Before	Now
Online‑research	10‑step, 400‑line prompt	3‑step prompt that instantly spawns the pipeline
Knowledge pipeline	Disabled by default	Streamlined 4‑stage flow auto‑called by online‑research
All modes	“EXACT / STEP N” boiler‑plate	Lean numbered lists (< 700 tokens each)
Non‑reasoning model support	Medium	★ Optimized for fast/light LLMs

---

✨ Key Components

🌐 Online‑Research Mode (v2)

The single entry point.

1. Parse the user query (topic, dates, geo).
2. Build Primary / Broader / Alternative search strings (≤ 150 chars).
3. Immediately delegate to Knowledge‑Source‑Fetcher via new_task.

📊 Knowledge Pipeline 2 → 5

Stage	Mode	Purpose
1	knowledge‑source‑fetcher	Run Brave Search, score relevance, fetch top 3 pages with Firecrawl
2	knowledge‑content‑extractor	Strip boiler‑plate, pull headings, paragraphs, code, data; label High/Med/Low
3	knowledge‑organizer	Auto‑choose document template (General / Comparison / Process) → Markdown file with YAML front‑matter
4	knowledge‑indexer	Update ~/knowledge/index.md, tags.md, and search-index.json

The whole flow is automatic—no user intervention needed after the initial question.

---

🛠️ Quick Start

Requirements

• Windows 10/11, VS Code + Roo Code extension
• Node.js, PowerShell 5.1+
• Brave Search & Firecrawl API keys

Install

git clone https://github.com/yourusername/roo-code-online-research.git
cd roo-code-online-research
:: store API keys
cmdkey /generic:MCP_BRAVE    /user:mcp /pass:YOUR_BRAVE_API_KEY
cmdkey /generic:MCP_FIRECRAWL /user:mcp /pass:YOUR_FIRECRAWL_API_KEY