Discovery Detail

Research #5

A single report that combines summary, context signals, web sources, and project-ready ideas.

Research Projects
Research / #5
Research Header

Handle the highest-value next step for this research

done ideas=6 sources=0
Current focus
Highest-value next step
Ideas ready
This research can be turned into a project directly. The highest-value next step is to pick an idea and start building.
Primary action
The primary action to take now
Review Ideas
Mode
How signals were collected
GitHub repo
Output
What’s available so far
summary context ideas
Review Ideas Research Ideas
Summary Context Sources Ideas
Research Query

https://github.com/KeygraphHQ/shannon

done use_web_search=1
Run ID
Research run identifier
#5
Mode
How signals were collected
GitHub repo
Outcome
What this produces
summary + ideas + project seed
Summary

Research summary

Summary

You can replicate the target app (KeygraphHQ/shannon) as a self-hosted knowledge-graph + semantic search service on a single VPS using Docker + Traefik. The MVP focuses on: ingesting notes/markdown, extracting entities/links into a graph, running embeddings for semantic search, and providing a simple web UI + API.

TOC

  1. Deployment notes (VPS, Docker, Traefik, domains, TLS)
  2. Idea 1: Shannon (clone)
  3. Adjacent alternatives (Ideas 2–6)

Deployment notes

  • VPS baseline: 2–4 vCPU, 8–16 GB RAM (embeddings/graph ops benefit from RAM), 80–200 GB SSD.
  • Edge routing: Traefik v3 with Let’s Encrypt (HTTP-01 or DNS-01). Route app.yourdomain.com, api.yourdomain.com, grafana.yourdomain.com.
  • Data: Postgres for core app data; Neo4j (or Postgres + AGE) for graph; Qdrant for vector search; Redis for jobs/queues.
  • Security: SSO/OAuth optional; at minimum password auth + rate limiting + backups + private network between containers.
  • Backups: nightly pg_dump + Qdrant snapshot + Neo4j dump to object storage (S3 compatible).
Context Signals

Actionable signals extracted by the builder

GitHub repo
KeygraphHQ/shannon
Shannon Lite is a fully autonomous AI pentester for web apps and APIs. 96.15% (100/104 exploits) on a hint-free, source-aware variant of the XBOW benchmark.
TypeScript ★ 32061 forks 3191 open 15
Labels (spec clues)
Labels often map directly to feature modules / risks / requirement categories. Tap to copy as brief hints.
Issues (sample)
PRs excluded. Only a small sample is fetched for scope / risks / roadmap hints.
README headings (outline hints)
Ideas

Project directions you can create in one click

Research
Quick Deploy
Clicking Create + Deploy creates a deploy job immediately (you can follow logs).
Shannon (self-hosted clone of KeygraphHQ/shannon)
shannon-clone
L cpu=4 ram=16GB disk=150GB
Teams and individuals want private, searchable knowledge graphs that combine backlinks + semantic search across notes/docs. GitHub-hosted, open tooling signals active interest in self-hosting and extensibility.
Self-hosted Semantic Search for Team Docs (Notion/Confluence alternative search layer)
team-semantic-search
M cpu=2 ram=8GB disk=100GB
Many teams already have docs but lack fast, accurate search across sources. A self-hosted search layer that indexes multiple sources is a common painkiller product.
Personal Knowledge Base with Backlinks (Obsidian-like web app)
web-backlinks-kb
M cpu=2 ram=4GB disk=50GB
Obsidian-style workflows are popular, but many users want a web-based, self-hosted solution with sync, sharing, and easy access from anywhere.
RAG Q&A over Internal Docs (Chat-with-your-docs, self-hosted)
rag-docs-qa
M cpu=4 ram=8GB disk=120GB
RAG assistants are widely adopted to reduce support load and speed up onboarding. Many orgs want on-prem/self-hosted options for privacy and compliance.
Engineering Dependency & Architecture Graph (services, repos, APIs)
engineering-arch-graph
L cpu=2 ram=6GB disk=80GB
Orgs struggle to understand microservice sprawl and dependencies. A graph-based explorer of repos/services/APIs helps incident response, onboarding, and governance.
Lightweight CRM Knowledge Graph (accounts, contacts, interactions)
crm-knowledge-graph
M cpu=2 ram=4GB disk=60GB
Small teams want a simple CRM that highlights relationship context and connections without paying for heavyweight SaaS. Graph modeling improves discovery (who knows whom, deal influence).