Campaign Knowledge Base

The Knowledge Base is attached to a Campaign and gives the AI the context it needs to write accurate, on-brand comments. Without a KB, the AI works only from your campaign description and advocate profile. With one, it can pull specific product details, pricing, use cases, and brand voice directly from your own content.

What You Can Upload

Images are not indexed. The system extracts text only. If your PDFs or pages contain screenshots or diagrams, the surrounding text is what gets indexed — not the visual content.

How It Works (RAG)

The Knowledge Base uses Retrieval-Augmented Generation (RAG):

1. Documents are split into chunks and stored as vector embeddings.
2. When the AI generates a draft, it retrieves the most relevant chunks for the specific Reddit post it's responding to.
3. Those chunks are injected into the prompt alongside the post context, advocate profile, and campaign settings.

This means the AI only "sees" the relevant parts of your KB for each draft — not the entire document. Keeping content modular (separate pages per topic) improves retrieval quality.

What to Include

Good KB content:

• Product overview — what it does, who it's for, key differentiators
• Feature descriptions — specific capabilities to mention in context
• Use cases — real scenarios where your product solves the user's problem
• Pricing summary — so the AI can answer "how much does it cost?" accurately
• FAQs — your existing public FAQ pages work well
• Tone examples — sample copy that shows the voice you want (even plain text blog posts help)

KB content to avoid:

• Raw code (syntax not useful for comment drafts)
• Image-heavy marketing PDFs where most value is visual
• Internal operational docs irrelevant to product positioning

Web Crawl

From the Campaign KB page, enter a root URL and ReddGrow will crawl up to 100 pages. The crawler follows internal links from the root page only — it won't leave your domain.

Tips for better crawl results:

• Point to your docs site or pricing page as the root if that's where the product detail lives.
• Add ?utm_source=reddgrow to your root URL to filter those sessions in analytics — ReddGrow respects robots.txt.
• If the crawl misses important pages, upload them as separate files instead.

Bulk Upload via CLI

For teams managing multiple campaigns or automating KB updates:

# Install the CLI
npm install -g @reddgrow/cli

# Authenticate
rg auth login

# Upload a file to a campaign's KB
rg kb upload --campaign <campaign-id> --file ./docs/product-overview.md

# Crawl a URL into a campaign's KB
rg kb crawl --campaign <campaign-id> --url https://docs.example.com

You can also pipe content directly for integration with build pipelines:

cat generated-release-notes.md | rg kb upload --campaign <id> --stdin --name "release-notes-v2.1"

Using Claude or Other LLMs to Build Your KB

Several users automate KB population by using Claude with the ReddGrow MCP server:

"Read our product documentation at docs.example.com and upload a
structured product summary to campaign ID <id> using rg kb upload."

This is useful when your docs are extensive and you want a condensed, comment-friendly summary rather than raw documentation.

KB Quality Tips

• Update after major product changes — outdated KB content leads to inaccurate drafts. Set a reminder to refresh after each product release.
• One campaign, one product — don't mix multiple products in the same KB. If you have two products, create two campaigns.
• Size matters less than focus — a tight 5-page KB on one product outperforms a 100-page dump of irrelevant material.