Skip to content

description: Recovery-grade FIT file processing for Python: parse corrupted .fit files, repair crash files for upload, analyze workouts. Zero dependencies, deterministic, MIT.

Parse anything.
Lose nothing silently.

chiptime is recovery-grade FIT file processing: hand it any workout file — pristine or mangled — and it returns everything genuinely in there, explains every decision it made, and never invents what isn't.

Get started API reference Star on GitHub

Every sports watch and bike computer saves workouts as .fit files — and in the real world those files are often imperfect: devices crash mid-ride, batteries die during the save, firmware writes impossible timestamps, sensors drop out. Most parsers handle the happy path and fail the rest: they crash, silently skip data, or quietly make things up. chiptime is built for the rest.

import chiptime

result = chiptime.parse("inProgressActivity.fit")   # truncated mid-ride crash file
result.ok                                           # True — salvaged
result.recovery.recovered_records                   # what came back
result.activity.sessions[0].rebuilt                 # True: session rebuilt from records
result.to_canonical_json()                          # deterministic, byte-stable JSON

Why chiptime

  • Never loses data silently

    Every drop, repair, and reinterpretation is recorded in a machine-readable provenance log on the output. Silent data loss is treated as the cardinal sin.

  • Deterministic to the byte

    Same input bytes → byte-identical canonical JSON, across runs, machines, and operating systems. Safe to diff, cache, and test against.

  • Repairs, not just reads

    chiptime repair salvages a damaged file and writes a fresh, valid .fit that Garmin Connect and Strava accept — verified against their validators.

  • Sport-aware analytics

    chiptime analyze speaks each sport's language: min/km for runs, watts for rides, min/100m and sets for swims, /500m splits for rowing — computed only from real evidence, never estimated.

  • Built for agents

    Stable machine codes for every error and insight, meaningful exit codes, canonical JSON, and generated reference docs. AI agents are first-class users.

  • Honest by contract

    null means the sensor said nothing; 0 means it said zero. Unknown fields are preserved, not crashed on. If data is truly gone, chiptime says so.

Install

pip install chiptime

Python ≥ 3.11, zero runtime dependencies. pip install "chiptime[pandas]" adds DataFrame export.

Sixty seconds of chiptime

chiptime parse ride.fit                 # human summary
chiptime parse ride.fit --json          # canonical JSON for machines
chiptime repair crashed.fit -o fixed.fit
chiptime validate fixed.fit --platform garmin-connect
chiptime analyze ride.fit --ftp 250     # per-sport report + insights

Continue with Getting started, or jump straight to the Python API reference.