Transforms raw incident reports into structured postmortem analyses — locally, privately, intelligently.

---

overview

PostMortem AI helps engineering teams document, analyze, and learn from incidents by automatically generating structured incident reports and postmortem summaries.

The platform combines a React frontend, FastAPI backend, PostgreSQL database, and locally hosted LLMs via Ollama to provide an end-to-end incident intelligence workflow — with no data leaving your infrastructure.

---

current status

Phase	Status
Phase 1 — Core Incident Intelligence Platform	✅ Complete
Phase 2 — Root Cause Intelligence	🚧 In Progress
Phase 3 — Enterprise Intelligence	🔮 Planned

---

features

✅ phase 1 — core platform v0.1.0

• Incident submission and management
• AI-generated incident reports via local LLM
• PostgreSQL persistence layer
• FastAPI REST API with structured endpoints
• React-based user interface
• Ollama integration for on-device LLM inference (Qwen2.5:3B)
• Structured incident storage and retrieval

🚧 phase 2 — root cause intelligence

• Automated root cause extraction
• Incident severity classification
• Timeline reconstruction
• Action item generation
• Executive summary generation
• Multi-incident trend analysis

🔮 phase 3 — enterprise intelligence

• Incident knowledge base
• Semantic search across incidents
• Retrieval-Augmented Generation (RAG)
• Team dashboards and analytics
• Incident similarity detection
• Predictive incident insights

---

tech stack

layer	technology
Frontend	React
Backend	FastAPI · SQLAlchemy
Database	PostgreSQL
AI / LLM	Ollama · Qwen2.5:3B
Language	Python · JavaScript

---

getting started

prerequisites

• Python 3.10+
• Node.js v18+
• PostgreSQL running locally
• Ollama installed with qwen2.5:3b pulled

ollama pull qwen2.5:3b

run the backend

# Install dependencies
pip install -r requirements.txt

# Start FastAPI server
uvicorn main:app --reload

API available at http://localhost:8000

run the frontend

# Install dependencies
npm install

# Start dev server
npm run dev

App available at http://localhost:5173

---

release

v0.1.0 — Phase 1 Complete

Phase 1 completion release. Core incident intelligence platform is fully functional with local LLM-powered report generation.

git tag -a v0.1.0 -m "Phase 1 Complete"
git push origin v0.1.0

---

license

This project is licensed under the MIT License.

---

Built by Samiksha Patil — Mumbai, India