How koushikbhuragoni04-glitch/aura works
Hindsight is an on-call incident triage agent that compares a baseline LLM response with a memory-informed response, while retaining engineer-confirmed resolutions for future alerts. The principal flow is alert submission through Flask, parallel recall/reflection and LLM triage, then evidence-backed results for the UI. A separate seed path loads synthetic incident history. The UI file is named in the README but absent from the supplied tree, so it is not represented as a repository node; relationships below are supported by app.py, agent.py, memory.py, seed.py, and README statements.
Main entry points: app.py Flask API · agent.py Memory-aware triage · memory.py Incident retention · seed.py History loader
Components 16
API and interaction
- Flask API
app.py - Resolution teaching
app.py - Web UI
Triage reasoning
- Triage orchestration
agent.py - Baseline triage
agent.py - Memory-aware triage
agent.py
Incident memory
- Similar recall
memory.py - Pattern reflection
memory.py - Incident retention
memory.py - Incident narrative
memory.py
Incident history
- Demo alert
seed_data.py - History loader
seed.py - Synthetic incidents
seed_data.py
Other components
- On-call engineer
- Groq LLM
- Hindsight bank
Connections 19
- On-call engineer → Flask API: submits alert
app.py - Flask API → Demo alert: serves demo
app.py - Flask API → Triage orchestration: calls triage
app.py - Triage orchestration → Baseline triage: dispatches
agent.py - Triage orchestration → Memory-aware triage: dispatches
agent.py - Baseline triage → Groq LLM: requests completion
agent.py - Memory-aware triage → Similar recall: requests memories
agent.py - Memory-aware triage → Pattern reflection: requests insight
agent.py - Memory-aware triage → Groq LLM: requests completion
agent.py - Similar recall → Hindsight bank: recalls incidents
memory.py - Pattern reflection → Hindsight bank: reflects over
memory.py - Flask API → Resolution teaching: accepts resolution
app.py - Resolution teaching → Incident retention: retains incident
app.py - Incident retention → Incident narrative: formats narrative
memory.py - Incident retention → Hindsight bank: stores incident
memory.py - History loader → Synthetic incidents: loads incidents
seed.py - History loader → Incident retention: retains each
seed.py - Web UI → Flask API: uses API
README.md - On-call engineer → Resolution teaching: teaches agent
app.py