tigs 0.1.2

Talks in Git → Specs - Git-based chat management system for AI-empowered software development

Tigs Python Package

Tigs (Talks in Git → Specs) is a CLI tool for storing and managing LLM chats in Git associated with code commits.

Why Tigs?

The biggest bug in software engineering isn't a crash — it's forgetting why. When someone asks "Why is this function designed this way?", too often the answer is "I think the AI suggested it?"

Tigs solves this by:

• Preserving decision rationale - Never lose that god-tier prompt or design debate
• Creating traceable history - Every "why" has a link you can follow
• Accelerating onboarding - New contributors understand the conversation, not just the code
• Building prompt libraries - Your best AI interactions become reusable team assets

Key Features

• Non-invasive storage - Uses Git notes; never rewrites your commits
• Fast TUI interface - Navigate commits, select chats, and link them effortlessly
• Tool-agnostic - Works with chats from Claude Code, Gemini CLI, Qwen Code and more
• Version-controlled context - Your reasoning becomes greppable, diffable, and reviewable
• Future: Auto-generated specs - AI will read commits + chats to generate precise system specifications

Installation

pip install tigs

Or using uv:

uv pip install tigs

Quick Start

Tigs provides two main interactive TUI (Terminal User Interface) commands:

tigs store - Select and Store Chats

Launch an interactive interface to select commits and messages to associate with code commits:

tigs store

The store interface features:

• Three-pane layout: Commits (left), Messages (center), Logs (right)
• Keyboard navigation:
  • j/k or ↑/↓ - Navigate up/down
  • Space - Toggle selection
  • v - Visual selection mode
  • a - Select all
  • c - Clear selections
  • Tab - Switch between panes
  • s - Store selected items as a chat
  • q - Quit

tigs view - Browse and Read Chats

Explore existing chats associated with your commits:

tigs view

The view interface displays:

• Three-column layout: Commits list, Commit details, Chat content
• Navigation: Browse through commits and view associated chats
• Read-only mode: Safely explore without modifying data

Syncing with Remote Repositories

Share your chats across team members using Git's native push/pull:

# Push chats to remote repository
tigs push

# Fetch chats from remote repository
tigs fetch

The push command validates that all commits with chats are pushed to the remote before pushing the notes, preventing orphaned references.

Low-Level Commands

For automation and scripting, Tigs provides direct CLI commands:

# Add chat to current commit (HEAD)
tigs add-chat -m "Chat content in YAML format"

# Add chat to specific commit
tigs add-chat abc123 -m "Chat content"

# Show chat for current commit
tigs show-chat

# Show chat for specific commit
tigs show-chat abc123

# List all commits that have chats
tigs list-chats

# Remove chat from commit
tigs remove-chat abc123

Interactive Editor

If you don't provide the -m flag, tigs opens your default editor:

# Opens editor for chat content
tigs add-chat

Chat Format

Chats are stored in YAML format following the tigs.chat/v1 schema:

schema: tigs.chat/v1
messages:
- role: user
  content: What does this commit implement?
- role: assistant
  content: This commit adds authentication using JWT tokens.

Git Integration

Tigs stores chats as Git notes in refs/notes/chats, which means:

• Version controlled: Chats are part of Git history
• Distributed: Push/pull chats like any Git data
• Non-invasive: Doesn't modify commits or require rebasing
• Compatible: Works with any Git workflow
• Efficient: Git's fanout structure handles large scale

You can also use standard Git commands:

# View notes directly
git notes --ref=refs/notes/chats show <commit>

# Push notes manually
git push origin refs/notes/chats

Use Cases

• Code Review: Attach review discussions with LLM to specific commits
• Documentation: Add design & implementation notes and decisions
• Learning: Annotate commits with explanations for team members
• AI Assistance: Store AI conversations about code changes
• Debugging: Keep notes about bug investigations tied to commits

Requirements

• Python 3.8+
• Git 2.17+
• Terminal with UTF-8 support
• Unix-like system (Linux, macOS, WSL)