TimeCamp Skill

Two tools: CLI for quick personal actions (timer, entries CRUD) and Data Pipeline for analytics/reports.

Bootstrap (clone if missing)

Before using either tool:

1. Ask user where repos should live (default: ~/utils, but any location is valid).
2. If repos are missing in that chosen location, ask for confirmation to clone.

Example flow and commands:

# Ask first:
# "I don't see TimeCamp repos locally. Clone to ~/utils, or use a different location?"

REPOS_DIR=~/utils  # replace if user picked a different path
mkdir -p "$REPOS_DIR"

if [ ! -d "$REPOS_DIR/timecamp-cli/.git" ]; then
  git clone https://github.com/timecamp-org/timecamp-cli.git "$REPOS_DIR/timecamp-cli"
fi

if [ ! -d "$REPOS_DIR/good-enough-timecamp-data-pipeline/.git" ]; then
  git clone https://github.com/timecamp-org/good-enough-timecamp-data-pipeline.git "$REPOS_DIR/good-enough-timecamp-data-pipeline"
fi

Tool 1: TimeCamp CLI (personal actions)

CLI at ~/utils/timecamp-cli, installed globally via npm link.

Tool 2: Data Pipeline (analytics & reports)

Python pipeline at ~/utils/good-enough-timecamp-data-pipeline. Use this for all analytics, reports, and bulk data fetching.

Run command

cd ~/utils/good-enough-timecamp-data-pipeline && \
uv run --with-requirements requirements.txt dlt_fetch_timecamp.py \
  --from YYYY-MM-DD --to YYYY-MM-DD \
  --datasets DATASETS \
  --format jsonl \
  --output ~/data/timecamp-data-pipeline

Available datasets

Formats: “jsonl`

Output structure

Files land in ~/data/timecamp-data-pipeline/timecamp/*.jsonl.

Examples

cd ~/utils/good-enough-timecamp-data-pipeline && \
uv run --with-requirements requirements.txt dlt_fetch_timecamp.py \
  --from 2026-02-11 --to 2026-02-14 \
  --datasets entries,users,tasks \
  --format jsonl --output ~/data/timecamp-data-pipeline

cd ~/utils/good-enough-timecamp-data-pipeline && \
uv run --with-requirements requirements.txt dlt_fetch_timecamp.py \
  --from 2026-01-01 --to 2026-02-14 \
  --datasets computer_activities,users,application_names \
  --format jsonl --output ~/data/timecamp-data-pipeline

cd ~/utils/good-enough-timecamp-data-pipeline && \
uv run --with-requirements requirements.txt dlt_fetch_timecamp.py \
  --from 2026-01-01 --to 2026-02-14 \
  --datasets computer_activities,users,application_names,entries,tasks \
  --format jsonl --output ~/data/timecamp-data-pipeline

Analytics with DuckDB

Query the persistent data store directly.

DUCKDB=~/.duckdb/cli/latest/duckdb
DATA=~/data/timecamp-data-pipeline/timecamp

# Hours per person
$DUCKDB -c "
SELECT user_name, round(sum(TRY_CAST(duration AS DOUBLE))/3600.0, 1) as hours
FROM read_json_auto('$DATA/entries*.jsonl')
GROUP BY user_name ORDER BY hours DESC
"

# Hours per person per day
$DUCKDB -c "
SELECT user_name, date, round(sum(TRY_CAST(duration AS DOUBLE))/3600.0, 1) as hours
FROM read_json_auto('$DATA/entries*.jsonl')
GROUP BY user_name, date ORDER BY user_name, date
"

# Top applications by time (join activities with app names)
$DUCKDB -c "
SELECT COALESCE(an.full_name, an.application_name, an.app_name, 'Unknown') as app,
       round(sum(ca.time_span)/3600.0, 2) as hours
FROM read_json_auto('$DATA/computer_activities*.jsonl') ca
LEFT JOIN read_json_auto('$DATA/application_names*.jsonl') an
  ON ca.application_id = an.application_id
GROUP BY 1 ORDER BY hours DESC LIMIT 20
"

# People who logged < 30h in a given week
$DUCKDB -c "
SELECT user_name, round(sum(TRY_CAST(duration AS DOUBLE))/3600.0, 1) as hours
FROM read_json_auto('$DATA/entries*.jsonl')
WHERE date BETWEEN '2026-02-03' AND '2026-02-07'
GROUP BY user_name
HAVING sum(TRY_CAST(duration AS DOUBLE))/3600.0 < 30
ORDER BY hours
"

Pattern

1. Check existing data range with DuckDB, if data is missing, fetch it with the pipeline, if it’s already there, use it
2. Query with DuckDB: $DUCKDB -c "SELECT ... FROM read_json_auto('$DATA/entries*.jsonl') ..."

Important Notes

• Duration (entries) is in seconds (3600 = 1h)
• time_span (activities) is also in seconds
• applications_cache.json in pipeline dir caches app name lookups
• For JSONL output, DuckDB glob *.jsonl catches all files for all datasets

Safety

• Confirm before adding, updating, or removing entries
• Show the command before executing modifications
• When stopping a timer, show what was running first