A next-generation AI bot designed to drastically reduce Mean Time To Resolution (MTTR) for TechOps issues. It analyzes incoming Jira tickets, retrieves similar historical cases using RAG (Retrieval-Augmented Generation), and posts potential solutions or diagnostic steps directly as comments.
Visual Preview
This project is currently in the **Design & Prototyping** phase.
TechOps teams spend countless hours answering the same repetitive questions or searching for "that one ticket from 6 months ago."
This bot aims to be the **first responder** that never sleeps.
1. **Ingestion Pipeline**:
* Fetch all historical closed Jira tickets.
* Clean and chunk text (Description, Comments, Resolution).
* Generate embeddings using **OpenAI text-embedding-3-small**.
* Upsert into **AWS OpenSearch** vector index.
2. **Real-Time Inference**:
* **Trigger**: Jira Webhook (`issue_created`).
* **Search**: Query OpenSearch for Top-3 most similar resolved tickets.
* **Reasoning**: Pass new ticket + similar tickets to **GPT-4o**.
* **Action**: Bot posts a comment on Jira: *"Based on similar issues (KEY-123, KEY-456), here is a suggested fix..."*
3. **Accuracy & Feedback**:
* Human-in-the-loop initial validation.
* "Was this helpful?" buttons in Slack notifications to fine-tune the prompt.
* **Backend**: Python / FastAPI
* **AI**: LangChain, OpenAI API
* **Database**: AWS OpenSearch (Vector), PostgreSQL (Metadata)
* **Integration**: Atlassian Jira REST API