distill

Agent command outputs are one of the biggest sources of token waste.

Logs, test results, stack traces… thousands of tokens sent to an LLM just to answer a simple question.

🔥 distill compresses command outputs into only what the LLM actually needs.

Save up to 99% of tokens without losing the signal.

How to use

ollama pull qwen3.5:2b
npm i -g @samuelfaj/distill

Add in your global agent instructions file:

CRITICAL: Always when you invoke a any shell command pipe it through `distill` and tell exactly what you want to know.
`distill` will summarize the response to save tokens, which is critical for us.

Examples:
- `bun test 2>&1 | distill "did the tests pass?"`
- `git diff 2>&1 | distill "what changed?"`
- `terraform plan 2>&1 | distill "is this safe?"`
- `npm audit 2>&1 | distill "extract vulnerabilities as JSON"`

THERE ARE ONLY TWO SCENARIOS WHEN YOU ARE ALLOWED TO NOT USE `distill`:
- Do not use `distill` if exact uncompressed output is required.
- Do not use `distill` if it would break an interactive/TUI workflow.

Usage

logs | distill "summarize errors"
git diff | distill "what changed?"
terraform plan 2>&1 | distill "is this safe?"

Configurations

You can persist defaults locally:

distill config model "qwen3.5:2b"
distill config timeout-ms 90000
distill config thinking false

For pipeline exit mirroring, use pipefail in your shell:

set -o pipefail

Interactive prompts are passed through when distill detects simple prompt patterns like or .