Skip to content

## Auto-generated CHANGELOG from task completions and datalad saves ### Problem

Auto-generated CHANGELOG from task completions and datalad saves

Problem

When multiple parallel Claude sessions work on a project, it's hard to know what changed when you come back. Git log is noisy, and there's no curated summary of what each session accomplished.

Proposal

projio should auto-maintain a CHANGELOG.md at the project root, updated on every datalad_save:

  1. On each datalad_save: append an entry to CHANGELOG.md with:
  2. Timestamp
  3. Commit message (from the save)
  4. Session ID (if available from worklog task execution)
  5. Files changed summary (from git diff --stat)

  6. Task note reference (if the save was triggered by a task)

  7. Format: reverse-chronological, grouped by date: 

    ## 2026-04-05

### Population analysis + spectral connectivity (session: 46136288)
- Added behavior/decoder.py, spikes/assemblies.py, spikes/replay.py
- Added spectral/coherence.py, spectral/spike_field.py
- 341 tests passing
- Task: task-arash-20260405-034120-565798

### Docs overhaul (session: ef5a13f0)
- Restructured toctree to Diataxis quadrants
- Added grid cards to homepage

  8. Integration: hook into datalad_save tool — after successful save, auto-append to CHANGELOG.md

Alternative

Could also be a worklog feature — worklog already tracks all task executions and could generate a per-project changelog from queue history. But putting it in projio keeps it closer to the code.

Source context: labpy

Labpy (labpy): sirota lab main python package; initiated as a translation of labbox Matlab code into python

Recent commits: 

b26f692 Tier 1+2 implementation: population analysis, spectral connectivity, spike-field, visualization — docs autodoc + correspondence map update, 341 tests passing
40420e7 batch task 2/2: Tier 2: Spectral connectivity + spike-field analysis
ac9e070 batch task 1/2: Tier 1: Population analysis — Bayesian decoder, assemblies, replay