Run Tools in Parallel
Use parallel tools for short, independent checks that belong in the same conversation.
#Scenario: map the tracker before changing a scraper
Before editing a provider scraper, you need to understand three separate parts of the project:
- where scrapers are registered
- the shape stored in
data/deprecations.json - how calendar events are generated
Ask Flux to inspect them together:
Inspect these parts of the repository in parallel:
1. Read @scraper/__init__.py and list the registered providers.
2. Read @scraper/base.py and @data/deprecations.json and summarize the entry shape.
3. Read @generators/ics_generator.py and explain which entries become calendar events.
Do not edit files. Combine the results into one short data-flow summary.Flux can run independent reads and searches concurrently, then return the evidence to the same conversation.
#Keep dependencies in order
Do not parallelize steps when one changes what the next step reads. State the order explicitly:
First update the provider registry. After that edit succeeds, inspect the final registry and generated outputs.Keep overlapping edits and state-changing commands sequential. Use parallel subagents when each investigation is large enough to need its own context.
#Expected result
You get one concise map of the scraper registry, shared data shape, and calendar output without waiting for each independent read in sequence.