Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/mneedham/devoxx-ai-sports-commentary
https://github.com/mneedham/devoxx-ai-sports-commentary
Last synced: 14 days ago
JSON representation
- Host: GitHub
- URL: https://github.com/mneedham/devoxx-ai-sports-commentary
- Owner: mneedham
- Created: 2024-04-02T12:44:21.000Z (9 months ago)
- Default Branch: main
- Last Pushed: 2024-04-22T10:24:59.000Z (9 months ago)
- Last Synced: 2024-04-22T11:37:09.435Z (9 months ago)
- Language: Python
- Size: 38.1 KB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Devoxx: Game, Set, Match: Transforming Live Sports with AI-Driven Commentary
This is the repository for a talk by Mark Needham and Dunith Danushka at Devoxx UK 2024.
https://www.devoxx.co.uk/talk/?id=11171
>We are both big fans of the live text commentary that the BBC provide for sports like football, tennis, rugby, cricket and more. While there are a lot of novel observations in the commentary, there is also a lot that is effectively summarising what just happened.
>Wouldn't it be cool if the commentator could have a Co-Pilot who can make the process more efficient?
>In this session, we will introduce an AI Co-Pilot for sports commentary based on Redpanda, ClickHouse, Flink, and a Large Language Model. A stream of events will be fed into RedPanda and we'll capture a window of those events on game-by-game and/or time-period buckets using Flink. These events, alongside historical match data, will also be stored in ClickHouse.
>We'll then send the LLM the events that have just happened along with queries on historical data, from which it can come up with suggested text commentary. The commentator can then decide whether they want to use the Co-Pilot's suggestion, edit the suggestion, or just go along with their own version.
This we need
* A stream of events related to a match
* A page that shows commentary of the match
* A page where the live commentary writer can add a new message
** On that page we should show the latest events and have a button to generate an AI message
** A way to bring in stats related to the players/tournament/etc
An example of what the live page should look like:
https://www.bbc.co.uk/sport/live/tennis/66006317/page/4
## Instructions
Start Redpanda
```bash
docker compose up
```
Download ClickHouse
```bash
curl https://clickhouse.com/ | sh
```
On another tab, configure ClickHouse Server
```bash
mkdir clickhouse-server && cd clickhouse-server
./clickhouse server
```
Copy `matches.csv` over
```bash
cp ../data/matches.csv user_files
```
Connect with ClickHouse Client
```bash
./clickhouse client -mn
```
Setup ClickHouse tables
```sql
CREATE TABLE pointsQueue(
match_id String,
id String,
time String,
player1 String,
player2 String,
previous_sets Array(String),
server String,
set String,
game String,
set_score String,
point_score String,
description String,
game_winner String,
publish_time String,
event_type String,
event_round String
)
ENGINE = Kafka(
'localhost:9092',
'points',
'points-consumer-group',
'JSONEachRow'
)
SETTINGS kafka_flush_interval_ms=500;
CREATE TABLE points (
match_id String,
id String,
time String,
player1 String,
player2 String,
previous_sets Array(String),
server String,
set String,
game String,
set_score String,
point_score String,
description String,
game_winner String,
publish_time DateTime32,
event_type String,
event_round String,
)
ENGINE = MergeTree
ORDER BY match_id;
CREATE MATERIALIZED VIEW points_mv TO points AS
SELECT * REPLACE(
parseDateTime32BestEffort(publish_time) AS publish_time
)
FROM pointsQueue;
CREATE TABLE matches
ORDER BY match_id AS
SELECT * REPLACE(
toString(match_id) AS match_id
)
FROM file('matches.csv')
SETTINGS schema_inference_make_columns_nullable=0;
```
Install Python dependencies
```bash
poetry install
```
Configure OpenAI API key
```bash
export OPENAI_API_KEY="sk-xxx"
```
Start commentator app
```bash
poetry run streamlit run apps/commentator.py --server.headless true
```
Navigate to http://localhost:8501