Wix Engineering
Read post

How We Built the Brain Behind Our Self-Healing System: Context Retrieval at Org Scale

Wix Engineering shares how they built a context retrieval system called Wix Octocode Research to power their autonomous bug-fixing orchestrator. The system goes beyond simple GitHub search by using a three-layer architecture: a Planner that unpacks bug reports and queries memory, a Researcher that orchestrates specialized sub-agents (database, logs, code) each with focused toolsets, and a Memory Layer that embeds completed research cycles for future use. Key design decisions include requiring agents to state reasons before tool calls, parallel branch exploration, semantic fallback hints, and uniform tool protocols. They forked Google's ADK to add true concurrency for TypeScript. Ongoing challenges include isolating relevant database records and logs at scale, and capturing organizational knowledge that lives in Slack threads. They're developing Context Driven Development (CDD) to let teams explicitly declare their domain context as a first-class part of the development lifecycle.

    #ai-agents#mcp
Jul 02•10m read time•From wix.engineering
Post cover image
Table of contents
The Problem: Search Doesn't Scale to Org ComplexityWhat We Built: OctoCode, and Then Wix Octocode ResearchWix Octocode Research: Plugging OctoCode Into the Wix Octocode Orchestrator PipelineA Design Principle That Took Us Too Long to LearnWhat's Still HardConclusion: The Centrality of Learning and Measurement at Scale
843 Impressions
Wix Engineering's image
Wix Engineering

36 Followers

•

330 Upvotes

Would you recommend this post?

Copy link
WhatsApp
Facebook
X
New Squad
  • © 2026 Daily Dev Ltd.
  • Guidelines
  • Explore
  • Tags
  • Sources
  • Squads
  • Leaderboard