An AI teammate that answers on behalf of unavailable team members by reasoning over continuously synced personal and shared team wikis.
Architecture
Problem it solves
When a teammate is on leave or unreachable, urgent project context is scattered across Slack, meetings, tickets, emails, and GitHub — forcing others to wait or work with incomplete information.
How it works
Personal and team wikis are auto-synced from meetings, Slack, tickets, emails, and GitHub PRs/discussions into .md files. When someone needs context about an unavailable teammate, the LLM receives both the team wiki and that person's wiki to answer strictly based on stored knowledge.
Approach
Skipped traditional RAG and vector DBs in favor of direct context feeding — modern LLMs with million-token windows can ingest entire wiki directories at once. Knowledge is provided in the initial prompt and persists across the session.