gitglimpse 0.1.4

Extract structured context from your git history — PR descriptions, standups, weekly reports, and LLM-ready JSON.

gitglimpse

Extract structured context from your git history. Standups, PR descriptions, weekly reports, and LLM-ready JSON — from one command.

50KB of raw diffs → 1KB of structured signal. Less noise, fewer tokens, better output.

---

Why gitglimpse?

Modern development (especially with AI) creates more changes across more files than ever.

Git tracks what changed — but not in a way that's easy to understand quickly.

gitglimpse bridges that gap.

It turns raw git history into:

• 🧠 Structured context (JSON) for LLMs and tools
• 📝 Standups & reports for humans
• 🔍 Clean summaries without opening diffs or PRs

Think of it as a context extraction layer for your codebase.

---

✨ Demo

---

Quick Start

pip install gitglimpse
cd your-project
glimpse standup

That’s it. No API keys, no setup required.

---

Core Idea

gitglimpse is not just a “standup generator”.

It’s a Git → Structured Context engine.

glimpse standup --json | your-llm

Instead of dumping raw diffs into an LLM, you give it:

• grouped tasks
• extracted tickets
• filtered noise
• optional diffs
• structured JSON

---

🚀 Commands

glimpse pr

Generate PR summaries from your branch.

glimpse pr
glimpse pr --json
glimpse pr --base develop

---

glimpse standup

Generate a daily summary or structured context.

glimpse standup
glimpse standup --json
glimpse standup --since "3 days ago"
glimpse standup --format markdown

Example:

Yesterday:
  • Add rate limiting middleware (AUTH-42, ~1.5h)
  • Fix pagination bug (BUG-87, ~1h)

Estimated effort: 2.5h

---

glimpse week

Weekly breakdown grouped by day.

glimpse week
glimpse week --json

---

glimpse init

Install Claude Code / Cursor slash commands.

glimpse init

Then in your editor:

/standup
/pr
/week

---

glimpse config

View or edit configuration.

glimpse config show
glimpse config setup

---

🧩 Output Modes

Mode	Description
Template	Fast, deterministic, no LLM
Local LLM	Uses Ollama (privacy-first)
Cloud API	OpenAI / Anthropic / Gemini
JSON	Structured output for pipelines

---

🔥 What Makes It Different

1. Noise Filtering (by default)

Removes:

• merge commits
• lock files
• formatting changes

So you only see meaningful work.

---

2. Task Grouping

Commits → grouped into real tasks:

3 commits → 1 task

---

3. Ticket Extraction

Automatically parses:

feature/AUTH-42-login → AUTH-42

---

4. Works Without LLMs

No AI required.

• Good commits → good summaries
• Bad commits → fallback heuristics
• LLM → optional enhancement

---

5. Built for LLM Workflows

Instead of:

git diff | llm

You do:

glimpse standup --json | llm

Cleaner input → better output.

---

🤖 Claude Code Integration

glimpse init
git add .claude/commands/
git commit -m "add glimpse commands"

Now every dev on your team gets:

/standup
/pr
/week

📌 The repo becomes the distribution channel.

---

GitHub Action

Add automatic PR context to your repository:

- uses: dino-zecevic/gitglimpse@main
  with:
    github-token: ${{ secrets.GITHUB_TOKEN }}

Every pull request gets a structured summary with changes, ticket IDs, and effort estimates. The comment updates on each push.

See action/README.md for full configuration options.

---

🧠 Multi-Project Mode

Run from a parent folder:

cd ~/projects
glimpse standup

gitglimpse will:

• detect repos automatically
• merge timelines
• group by project or task

---

⚙️ Configuration

glimpse config setup

Stored in:

~/.config/gitglimpse/config.toml

Supports:

• local models (Ollama)
• cloud APIs
• context modes (commits, diffs, both)

---

Philosophy

• Privacy-first — works fully offline
• LLM-optional — useful without AI
• Developer-first — not a manager tool
• Composable — JSON output for pipelines
• Honest — no fake precision (effort is approximate)

---

When It’s Actually Useful

• Weekly summaries across repos
• PR descriptions
• Feeding context into coding agents
• Remembering what you did yesterday

---

⚠️ Limitations

• Only sees code changes (not meetings, docs, etc.)
• Effort estimation is heuristic, not accurate
• Depends on git history quality

---

Installation

pip install gitglimpse

Requirements:

• Python 3.11+
• git
• (optional) Ollama or API key

---

Contributing

PRs welcome — especially for:

• better effort estimation
• smarter task grouping
• improved noise filtering

---

License

MIT

---

Built by Dino