RAG scope limiter for LLM agents to answer only forum-specific questions
Built for Discourse forum admins using AI bots.
“I’m setting up Discourse AI on my forum and have an AI Bot/Agent working with Gemini 2.5 Flash via LiteLLM and Google Vertex AI. The bot is working, but I have…”
The receipts — real demand
“I’m setting up Discourse AI on my forum and have an AI Bot/Agent working with Gemini 2.5 Flash via LiteLLM and Google Vertex AI. The bot is working, but I have a problem with its scope. For example, my forum is a busin…”
Full dossier
Unlock the full dossier — free
Every corroborating quote, the source receipts, and the community echo. One email, no payment.
Why this is a gap
Surfaced from a high-intensity complaint with clear willingness to pay and a specific, reachable audience.
The market
Discourse forum admins running AI bots (Gemini 2.5 Flash + LiteLLM stack) who need LLM responses confined to forum content only. No search volume given; demand appears driven by active Discourse + AI bot adoption, not broad viral interest.
Competition & the opening
LlamaIndex, LangChain, Pinecone, Weaviate, Vectara, and Mendable/Inkeep all offer retrieval scoping, metadata filtering, or corpus-level access control. The gap: none of these are purpose-built *for Discourse admin workflows* — they require custom integration work to wire forum-specific isolation into an existing LLM agent stack. A Discourse-native RAG limiter (plugin or LiteLLM middleware) would
What's hard to build
Integrating tightly with Discourse's post/category/user permission model while also intercepting and validating LLM retrieval calls in real-time is hard. You need robust hooks into both Discourse's data layer and the LiteLLM/Gemini request pipeline, plus robust testing across different forum structures (hierarchies, private categories, group visibility).
Why now
LLM agents on forums need scope guardrails that existing RAG frameworks (LlamaIndex, LangChain) require manual setup to enforce, creating friction for non-engineers deploying Discourse AI.
How you'd monetize
freemium (basic scope rules free, advanced routing/multi-tenant $29/mo)