3 ms·
So ... how does it work? Goes without saying, HN is a forum for discussions between tech folks, not an ad platform.
by teiferer 2mo ago
So ... how does it work?
Goes without saying, HN is a forum for discussions between tech folks, not an ad platform.
- cool_coder12 2mo agoI tried posting a long post with how it works but those were flagged (IDK WHY). Here’s an overview. Decispher is a continuous ingestion + retrieval system for engineering context: 1. Ingestion Slack: message + thread classification into constraint / convention / decision / ownership / plan / history GitHub: PR descriptions, review comments, linked issues, commit messages Docs / ADRs: markdown, Confluence, Notion (via connectors) Coding Session: Humans + AI Agents (codex, cursor, grok, VSCE) 2. Processing pipeline Entity + temporal + participant extraction Embedding + fusion into a knowledge graph (decisions are nodes with “why”, “supersedes”, “conflicts-with” edges) Conflict detection when a new context contradicts an older one that is still active 3. Serving layer Humans: search + “why did we decide X?” with citations back to the original Slack/PR message/Dashboard chat Agents: MCP server that injects the relevant decision context into Claude/Cursor/etc. so the agent gets context directly on demand based on task instead of static dump. There is also a open-source tool called Decision-guardian for decision enforcement at PR or CLI.