Enterprise RAG Practices: Onyx Setup Struggles, Seeking Slack-Integrated Alternatives
Bmjslider · reddit · 2026-08-12
An AI team at a 50-100 person nonprofit shares their pain points and explorations in building an internal enterprise knowledge retrieval (RAG) system.
- Core Needs: Using Google Workspace, Notion, and Slack, the team wants to efficiently index data from the first two and use Slack as the primary interface for Q&A with citations, workflow triggers, and document write-backs.
- Tool Evaluations: They tried the open-source tool Onyx (formerly Danswer), which fits the concept but proved cumbersome to set up and wasn't ready for daily use weeks later. Claude's Tag feature is appealing, but its underlying MCP search for Drive and Notion is slow and unreliable.
- Seeking Advice: The poster asks the community for real-world experiences with platforms like Glean, Dust, and Credal, or whether anyone has abandoned standalone RAG platforms entirely in favor of LLM enterprise connectors.
More from coding & agent
- Peter Steinberger on Building Software with AI Agents and Open Source — andrewchen · 2026-08-12
- Hiring in the AI Era: LeetCode Fails to Vet AI-Fluent Engineers — annetgriffin · 2026-08-12
- CoreWeave Launches TypeScript SDK for Sandboxes to Safely Run AI Agent Code — wandb · 2026-08-12
- RAG Trap: Vector Search Can Find the Answer Yet Fail the LLM — WirelessLife · 2026-08-12
- GitHub Repo Tracks the Evolution and Frameworks of Self-Evolving AI Agents — tom_doerr · 2026-08-12
- Developer Releases Caffeine: Keeps Mac Awake While Claude Code and Codex Run — iannuttall · 2026-08-12