// why Share2Brain
Knowledge that dies in the scroll, answers you can verify
Your community already has the knowledge. Make it searchable.
Share2Brain detects every link shared in Discord, fetches and chunks its content, and indexes it in a pgvector knowledge base. Your community builds the knowledge base without changing how they work.
Verifiable answers, not hallucinations.
Every answer includes traceable citations to the original chunk: channel, author, date and link. The agent streams responses so members see the answer form in real time.
You decide who sees what.
Discord roles map to channels inside the agent. Every query respects those permissions: each member only gets answers from channels they can access. RBAC enforced on every vector query.
// features
Everything you need to let your community query its own knowledge
Chat with verifiable citations
The RAG agent streams answers citing indexed fragments. If it can't find an answer, it says so.
Semantic search
Query the entire knowledge base with pgvector. No exact keywords needed.
Discord role-based access
Roles define which channels each member sees. RBAC on every query.
Automatic notifications
Slack and Telegram alerts when new content is indexed.
Read tracking
Unread badges per member and per channel. Know what's left to read.
Knowledge stats
Indexing activity, per-channel coverage and top contributors.
Discord login
OAuth2 with your Discord app. Optional guest access available.
Docker, one command
7 services in one docker-compose up. Backend, bot, workers, web, Postgres, Redis, nginx.
// how it works
An agent that reasons, not a bot that repeats
Share2Brain's core is a Reason-Act-Observe loop built with LangGraph. The agent analyzes the query, decides which tools to use and observes results until it resolves the query — it never guesses.
Step 01 · reason
Reason
The runtime assembles the full context —system prompt, available tools, history and retrieved chunks— and calls the model to decide the next step.
All security and control live in the runtime, not the model. The model reasons; the runtime executes.
// use cases
Real results for real communities
Eliminate recurring questions
Questions that repeat every week answer themselves, citing the exact resource. Your mod team is freed for higher-value work.
Frictionless onboarding
New members find rules, guides and answers without waiting for a moderator. The agent knows everything shared in channels they can access.
Technical support with sources
Questions about your product, API or config answered instantly, citing the original docs or thread. Accuracy without hallucination.
Search your history
Recover conversations, decisions and resources shared in threads and channels using semantic search. No need to recall exact wording.
// installation
One command, 7 services, zero external dependencies
A single docker-compose up starts the backend, the Discord bot, the indexing workers, the web app (search, chat, documents and stats), Postgres with pgvector, Redis and nginx. No SaaS, no subscriptions.
Clone the repository
$ git clone github.com/borjaberrocal87/share2brain
$ cd share2brainConfigure environment variables
$ cp .env.example .env
$ cp Share2Brain.config.yml.example Share2Brain.config.ymlStart all 7 services
$ docker compose up -d
✓ 7 services · http://localhost// documentation
Browsable documentation
Everything you need to deploy and operate your instance, organized by section.
// tech stack
Built on proven foundations
Clean Architecture
Layered separation with inward dependencies.
Domain depends on nothing; outer layers implement its ports.
Event-Driven
Heavy indexing is decoupled from the runtime with Redis Streams and the Outbox pattern.
Your knowledge deserves to be searchable
Open source, self-hosted and verifiable. Deploy your own knowledge agent with one command. Your content never leaves your infrastructure.