Blogus

package.lock for AI prompts.

Extract prompts from your codebase, version them like dependencies, and keep everything in sync.

Why Blogus?

Approach	Discover	Version	Test	Sync	Lock
Inline strings	-	-	-	-	-
Manual .txt files	-	Git	-	Manual	-
LangChain Hub	-	Yes	-	Manual	-
PromptLayer / Humanloop	-	Yes	Yes	Manual	-
Blogus	Auto	Git	Yes	Auto	Yes

Blogus is different because it:

• Extracts prompts from your existing code (no migration required)
• Locks versions with content hashes (like package.lock)
• Syncs changes back to your source files automatically

Quick Start

# Try it now
uvx blogus scan

# Install
uv add blogus

1. Extract

Find all LLM calls in your codebase:

$ blogus scan

Found 3 LLM API calls:
  src/chat.py:15         OpenAI      unversioned
  src/summarize.py:42    OpenAI      unversioned
  lib/translate.js:28    Anthropic   unversioned

2. Version

Create .prompt files:

$ blogus init

# prompts/summarize.prompt
---
name: summarize
model:
  id: gpt-4o
  temperature: 0.3
variables:
  - name: text
    required: true
---
Summarize in 2-3 sentences: {{text}}

3. Lock

Generate a lock file:

$ blogus lock

# prompts.lock
prompts:
  summarize:
    hash: sha256:a1b2c3d4...
    commit: 4903f76

4. Sync

Update your code when prompts change:

$ blogus fix

# Before
content = "Summarize: " + text

# After
# @blogus:summarize sha256:a1b2c3d4
content = load_prompt("summarize", text=text)

5. Verify (CI)

$ blogus verify || exit 1

Commands

Command	Description
scan	Find LLM API calls in your code
init	Create prompts directory
prompts	List all .prompt files
exec <name>	Run a prompt with variables
analyze	Evaluate prompt effectiveness
test	Generate test cases
lock	Generate prompts.lock
verify	Check lock file (for CI)
check	Find unversioned prompts
fix	Sync code with .prompt files
demo	Launch interactive TUI demo

Web UI

uvx --with blogus[web] blogus-web

Browse prompts, scan projects, and commit changes at http://localhost:8000.

Interactive Demo

Experience Blogus with an interactive terminal demo - perfect for presentations and learning.

# Install with TUI support
uv add blogus[tui]

# Launch the demo
blogus demo

The demo walks through:

• Scan - Live detection of prompts in a sample project
• Analyze - LLM-powered effectiveness scoring
• Compare - Side-by-side multi-model output
• Fix Workflow - Check/Fix/Lock/Verify in action

Options:

blogus demo --speed slow     # Slower animations for talks
blogus demo -p ./my-project  # Demo with your own project

.prompt Format

---
name: code-review
description: Review code for bugs
model:
  id: gpt-4o
  temperature: 0.3
variables:
  - name: code
    required: true
  - name: language
    default: python
---
Review this {{language}} code for bugs:

{{code}}

How It Works

┌──────────┐    ┌──────────┐    ┌──────────┐    ┌──────────┐
│  SCAN    │───▶│  VERSION │───▶│   LOCK   │───▶│   SYNC   │
│          │    │          │    │          │    │          │
│ Find LLM │    │ .prompt  │    │ prompts  │    │ Update   │
│ calls    │    │ files    │    │ .lock    │    │ code     │
└──────────┘    └──────────┘    └──────────┘    └──────────┘

Documentation

• Getting Started
• Core Concepts
• .prompt Format
• CLI Reference
• Web Interface

License

MIT