🚀 Cuebit - Prompt Versioning and Management for GenAI

Cuebit is an open-source, local-first prompt registry and version control system designed for GenAI development teams — complete with version tracking, version history, lineage tracking, aliases, tagging, and an interactive dashboard.

✨ Features

• 🔐 Prompt version control with full history and lineage tracking
• 🏷️ Alias system (e.g. summarizer-prod → versioned prompt)
• 🧠 Tags and metadata for organizing prompts
• 📁 Project & Task-based prompt grouping
• 📋 Example management for documenting prompt usage patterns
• 📑 Template variables detection and validation
• 🔍 Version comparison with visual diffs
• 📈 Streamlit-based visual dashboard
• ⚙️ REST API powered by FastAPI (/api/v1/...)
• 🔄 Full CLI support for automation
• 🧪 Prompt template preview and rendering
• 👤 Audit trail: created_at, updated_at, updated_by
• 📤 Import/Export functionality for backup and sharing

📦 Installation

# Install from PyPI
pip install cuebit

# Or install from source
git clone https://github.com/iRahulPandey/Cuebit.git
cd cuebit
pip install -e .

🚀 Getting Started

🌱 Initializing Cuebit

# Initialize the Cuebit prompt registry
cuebit init

# Optionally specify a custom data directory
cuebit init --data-dir /path/to/your/data

Cuebit automatically stores its database in a standard user data directory. You can also set the CUEBIT_DB_PATH environment variable to specify a custom database location.

🧪 Registering a Prompt in Python

from cuebit.registry import PromptRegistry

registry = PromptRegistry()
registry.register_prompt(
    task="summarization",
    template="Summarize: {input}",
    meta={"model": "gpt-4", "temperature": 0.7},
    tags=["prod"],
    project="bloggen",
    updated_by="alice",
    examples=[{
        "input": "The quick brown fox jumps over the lazy dog.",
        "output": "A fox jumps over a dog.",
        "description": "Basic example"
    }]
)

🧭 Setting Aliases

registry.add_alias(prompt_id, "summarizer-prod")

🔁 Updating a Prompt (Creates a new version)

registry.update_prompt(
    prompt_id,
    new_template="Summarize concisely: {input}",
    updated_by="bob"
)

📋 Adding Examples

registry.add_example(
    prompt_id,
    input_text="Climate change is a global challenge...",
    output_text="Climate change poses worldwide risks requiring immediate action.",
    description="Climate topic example"
)

Using in a Streamlit App

import streamlit as st
from cuebit.registry import PromptRegistry
from langchain_openai import ChatOpenAI
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain

# Initialize registry - works with zero configuration!
registry = PromptRegistry()

# Get prompt by alias
prompt = registry.get_prompt_by_alias("summarizer-prod")

# Create LangChain PromptTemplate
prompt_template = PromptTemplate(
    input_variables=["input_text"],
    template=prompt.template
)

# Initialize LangChain ChatOpenAI model
llm = ChatOpenAI(
    model=prompt.meta.get("model", "gpt-3.5-turbo"),
    temperature=prompt.meta.get("temperature", 0.7),
    max_tokens=prompt.meta.get("max_tokens", 500)
)

# Create LLM Chain
summarization_chain = LLMChain(
    llm=llm,
    prompt=prompt_template
)

# Generate summary
summary = summarization_chain.run(input_text="Your text to summarize")

🧪 Streamlit Dashboard

cuebit serve --host 127.0.0.1 --port 8000  # launches API + UI

The dashboard provides:

• Project and prompt browsing
• Visual prompt builder with variable detection
• Version history with visual diffs
• Import/Export functionality
• Usage statistics

🔧 Command Line Interface

# Start the server and dashboard
cuebit serve

# List all projects
cuebit list projects

# List prompts in a project
cuebit list prompts --project my-project

# Create a new prompt
cuebit create prompt --task summarization \
    --template "Summarize: {input}" \
    --project my-project --tags "prod,gpt-4"

# Set an alias
cuebit set-alias 123abc summarizer-prod

# Render a prompt with variables
cuebit render --alias summarizer-prod \
    --vars '{"input":"Text to summarize"}'

# Export prompts to JSON
cuebit export --format json --file exports.json

📚 API Reference

• GET /api/v1/projects - List all projects
• GET /api/v1/projects/{project}/prompts - List prompts in a project
• GET /api/v1/prompts - List all prompts (with pagination and filtering)
• POST /api/v1/prompts - Create a new prompt
• GET /api/v1/prompts/{prompt_id} - Get a specific prompt
• PUT /api/v1/prompts/{prompt_id} - Update a prompt (creates new version)
• POST /api/v1/prompts/{prompt_id}/alias - Set an alias for a prompt
• GET /api/v1/prompts/alias/{alias} - Get a prompt by its alias
• POST /api/v1/prompts/render - Render a prompt with variables
• GET /api/v1/prompts/{prompt_id}/history - Get version history
• POST /api/v1/prompts/compare - Compare two prompt versions
• POST /api/v1/prompts/{prompt_id}/rollback - Rollback to a previous version
• DELETE /api/v1/prompts/{prompt_id} - Delete a prompt (soft by default)
• GET /api/v1/export - Export prompts
• POST /api/v1/import - Import prompts

🔍 Advanced Configuration

Environment Variables

• CUEBIT_DB_PATH: Set a custom database location (e.g., sqlite:///path/to/your/prompts.db)

Using With Different Database Backends

Cuebit uses SQLAlchemy, so you can connect to different database backends:

# PostgreSQL example
registry = PromptRegistry("postgresql://user:password@localhost/cuebit")

# MySQL example
registry = PromptRegistry("mysql+pymysql://user:password@localhost/cuebit")

🛠️ Development

To contribute:

# Clone the repository
git clone https://github.com/iRahulPandey/Cuebit.git
cd cuebit

# Install development dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Build package
python -m build

License

MIT