Skip to content

context.md

Latest commit

 

History

History
247 lines (156 loc) · 3.17 KB

context.md

File metadata and controls

247 lines (156 loc) · 3.17 KB
Raw

DevContextIQ - Project Overview

1. Project Summary

DevContextIQ is an AI-powered engineering intelligence platform that:

  • Stores engineering knowledge from repositories and events
  • Answers project questions using memory
  • Analyzes pull request merge risk
  • Analyzes incidents and outages
  • Shows workspace integration health
  • Uses FastAPI, Supabase, embeddings, and frontend UI

It can be described as:

ChatGPT + GitHub Memory + DevOps Copilot

2. Core Architecture

Frontend

Currently running locally on:

http://localhost:5173

Likely built with:

  • React
  • Vite
  • TypeScript

Frontend responsibilities:

  • UI pages
  • User inputs
  • API calls
  • Status checks
  • Result rendering

Backend

Backend is deployed and healthy.

Version shown:

  • v2.0.0

API Base:

http://127.0.0.1:8001/api/v1

Likely backend stack:

  • FastAPI
  • REST APIs
  • OpenAI / Hugging Face model calls
  • Supabase database access

Database

Using Supabase.

Main tables:

  • nodes
  • edges
  • node_embeddings

3. Data Model Explanation

nodes Table

Stores important entities.

Examples:

  • Repo
  • Author
  • Decision
  • Service
  • Incident

Example:

Type Label
repo Dinol-ino/DevContext
author Dinol-ino
decision approved

edges Table

Stores relationships between nodes.

Examples:

  • belongs_to_repo
  • owned_by_author
  • related_to_incident

Example:

From To Relation
decision repo belongs_to_repo
decision author owned_by_author

node_embeddings Table

Stores vector embeddings of engineering memory chunks.

Used for:

  • Semantic search
  • Retrieval-Augmented Generation (RAG)
  • Context retrieval

4. Main Features

4.1 Chat Assistant

Allows users to ask questions about engineering memory.

Examples:

  • Explain auth architecture
  • What changed recently?
  • Who approved PR?
  • Show backend decisions

How It Works

  1. Frontend sends question
  2. API call to /api/v1/ask
  3. Backend searches embeddings
  4. Fetches relevant nodes
  5. Builds context
  6. Sends to LLM
  7. Returns final answer

This is a RAG system.

4.2 Governance Check

PR merge risk analyzer.

Inputs:

  • PR URL
  • Diff text

Outputs:

  • Safe to merge or not
  • Risk level
  • Conflicts
  • Review recommendation

Endpoint:

POST /api/v1/governance/check

Use Cases

  • Reviewers
  • Engineering managers
  • CI gatekeeping

4.3 Incident Analysis

Analyzes production alerts and failures.

Inputs:

  • Alert title
  • Service name
  • Error snippet

Outputs:

  • Likely cause
  • Severity
  • Suggested actions
  • Warnings

Endpoint:

POST /api/v1/incident

Example:

Input:

Too many DB connections

Possible Output:

  • Connection leak
  • Pool saturation
  • Restart workers
  • Inspect ORM sessions

4.4 Workspace Settings

Shows integration health.

Current examples:

  • Repo selected
  • Backend healthy
  • Auth config
  • API base

Future additions:

  • GitHub Connect
  • Billing
  • Slack Bot
  • Organization Memory

5. API Endpoints

Health Check

GET /api/v1/health

Example Response:

{
  "status": "healthy",
  "version": "2.0.0"
}