🎫 Ticket Priority Service

A tiny, production-style micro-service that lets you:

• create, read, update and delete support tickets
• auto-classify each ticket’s priority with an LLM
• browse the data through a Streamlit UI

The code follows a clean “ports & adapters” layout (FastAPI → Service Layer → Repos / LLM adapter) and is fully containerised with Docker Compose.

---

1. Quick-Start & Cross-Platform Setup (Docker Compose)

Prerequisites: Docker 19.03 or newer and Docker Compose v2.

💻 Shell / OS compatibility

The commands you see below are written in standard POSIX-style (bash / zsh). They therefore run unmodified on:

• any Linux distribution
• macOS Terminal / iTerm
• Windows 10/11 when you use WSL 2 or Git-Bash

Running Windows without WSL?

1. Replace the one cp command with its Windows siblings
   PowerShell → Copy-Item env.example .env
   CMD → copy env.example .env
2. All git clone … and docker compose … lines stay exactly the same—Docker Desktop ships the very same v2 CLI plug-in on Windows.

Other than that tweak, every step is identical across operating systems because the heavy lifting happens inside the containers.

# 1 – clone
git clone https://github.com/darkfennertrader/ticket-service-priority.git
cd ticket-service-priority

# 2 – set env-vars (copy env.example if you need to tweak anything)
cp env.example .env

inside the copied .env set the following vars if needed: OPENAI_API_KEY="sk-..." # real key for live priority classification with an OPEN_AI LLM API_HOST_PORT=<your_local_port> # set it to whatever number if port 8000 is already in use on the host machine UI_HOST_PORT=<your_local_port> # set it to whatever number if port 8501 is already in use on the host machine

# 3 – launch the full stack
docker compose up --build

Compose does three things:

1. init-db – one-shot container that creates the tickets table in SQLite
2. api – FastAPI backend on http://localhost:your_local_port
3. frontend – Streamlit UI frontend on http://localhost:your_local_port

Stop everything with CTRL-C and wipe volumes with docker compose down -v.

---

2a. REST API Cheat-sheet

(The interactive Swagger docs live at /docs)

Verb & path	Purpose
POST /tickets	Create a new ticket
GET /tickets	List tickets — optional filters status_filter, priority_filter
GET /tickets/{id}	Retrieve one ticket
PATCH /tickets/{id}	Update title, description or status
DELETE /tickets/{id}	Delete a ticket

Example calls with curl

CREATE:

curl -X POST http://localhost:<YOUR_PORT>/tickets \
     -H "Content-Type: application/json" \
     -d '{ "title": "Prod down", "description": "Login is impossible" }'

LIST ALL TICKETS:

curl "http://localhost:<YOUR_PORT>/tickets"

LIST (only OPEN+HIGH):

curl "http://localhost:<YOUR_PORT>/tickets?status_filter=OPEN&priority_filter=HIGH"

UPDATE STATUS:

curl -X PATCH http://localhost:<YOUR_PORT>/tickets/<UUID> \
     -H "Content-Type: application/json" \
     -d '{ "status": "IN_PROGRESS" }'

DELETE:

curl -X DELETE http://localhost:<YOUR_PORT>/tickets/<UUID>

2b. (OPTIONAL) Point-and-Click UI (🚀 Streamlit)

Don’t feel like typing cURL commands?
Spin up the included Streamlit app instead and drive the API from your browser.

How to launch

docker compose up --build already starts three containers:

1. init-db (one-shot, creates the tickets table)
2. api (FastAPI backend on http://localhost:${API_HOST_PORT:-8000})
3. frontend (Streamlit UI on http://localhost:${UI_HOST_PORT:-8501})

Just open the URL shown above (default http://localhost:<your_local_port>) and you’re in.

What you can do

• Create Ticket – fill a form, hit Create, watch the LLM auto-priority appear.
• Browse Tickets – filter by status/priority, inspect details, update or delete.
• Deleted Tickets – a session-local table thta lists all the deleted tickets.

Everything you do in the UI is a plain HTTP request to the FastAPI service; the browser dev-tools Network tab shows the exact endpoints if you want to peek under the hood.

---

3. LLM Prompt Template

When a ticket is created the service sends exactly the prompt below to the LLM and expects a single word (HIGH, MEDIUM or LOW) in response.
If the call fails (no API key, network error, etc.) the priority silently falls back to TBD.

You are an automated support-ticket triage assistant for our engineering team.

TASK
1. Read the ticket title and description.
2. Decide the priority according to the POLICY below.
3. Reply with ONE WORD ONLY—exactly HIGH, MEDIUM, or LOW—uppercase, with no
   other text, punctuation, or line breaks.

POLICY
HIGH   - Full production outage, data loss, security breach, payment failure,
         or any issue that blocks customers from using a core feature.
MEDIUM - Partial outage, severe performance degradation, significant bug with a
         workaround, or time-sensitive issue that is not mission-critical.
LOW    - Cosmetic defect, minor usability issue, documentation request, feature
         idea, or general question that can be scheduled for later.

---

4. Running the Test-suite

The project ships with:

• unit tests for the AI integration (LLM calls are 100 % mocked)
• black-box integration tests for the full FastAPI router

4.1. Locally (host machine)

python -m venv .venv          # optional but recommended
source .venv/bin/activate
pip install -r requirements.txt
pytest -q

4.2. Inside Docker

You can also execute the very same tests against the API image:

docker compose run --rm \
  -v "$(pwd)/tests":/usr/src/app/tests \
  api pytest -q

Both ways finish in a few seconds and require no real OpenAI key.

---

Happy hacking! 🚀