ActionSync

Transform meeting transcriptions into Jira tickets automatically using Claude AI.

Features

• Meeting Mode: Paste a meeting transcript. Claude analyzes it, checks your GitLab repos for technical context, and creates Jira tickets with relevant files, components, acceptance criteria, and links to related issues.
• Ask Mode: Ask questions about your project in natural language. Claude searches across Jira tickets, past meetings, and your codebase to find answers.
• Work Mode: Point Claude at a Jira ticket. It clones the relevant repos, reads the ticket and codebase, and works on the implementation.
• GitLab Integration: Connect GitLab repositories to provide code context for ticket creation and AI work.
• Meeting History & Semantic Search: Store processed meetings with vector embeddings for intelligent search.
• Real-time Updates: WebSocket connection for live processing feedback.
• Per-Project Customization: Custom instructions, GitLab repos, and settings per project.
• Flexible API Support: Works with direct Anthropic/OpenAI APIs or Azure-hosted versions - no Azure subscription required!

Architecture

Frontend (Vanilla JavaScript SPA)
    ↓
FastAPI Backend (Python/Async)
    ↓
PostgreSQL Database (with pgvector)
    ↓
Claude AI (Anthropic or Azure Anthropic API)
OpenAI Embeddings (OpenAI or Azure OpenAI API)
    ↓
Jira REST API + GitLab API

• Backend: FastAPI with async SQLAlchemy (PostgreSQL/SQLite)
• Frontend: Vanilla JavaScript SPA with modern dark theme
• AI: Claude via Anthropic API (direct or Azure) with MCP tools for Jira operations
• Embeddings: OpenAI API (direct or Azure) for semantic search (optional)
• Auth: JWT tokens with bcrypt password hashing

Quick Start

Using Docker Compose (Recommended)

1. Copy the example environment file and configure it:

   cp .env.example .env
   # Edit .env with your values

   Required variables:

   • POSTGRES_PASSWORD - Database password
   • SECRET_KEY - JWT signing secret
   • LLM_PROVIDER - Choose anthropic or azure_anthropic

   For direct Anthropic API (LLM_PROVIDER=anthropic):

   • ANTHROPIC_API_KEY - Your Anthropic API key

   For Azure Anthropic (LLM_PROVIDER=azure_anthropic):

   • AZURE_ANTHROPIC_ENDPOINT - Azure Anthropic API endpoint
   • AZURE_ANTHROPIC_API_KEY - Azure Anthropic API key

2. Start the application:

   docker-compose up -d

3. Access at http://localhost:8080

Manual Setup

1. Install Python dependencies:

   cd backend
   pip install -r requirements.txt

2. Set environment variables (see Environment Variables)

3. Run the server:

   python server.py

Configuration

API Provider Setup

ActionSync supports both direct Anthropic/OpenAI APIs and Azure-hosted versions. See the API Providers Guide for detailed configuration instructions.

Quick start options:

• Direct APIs (no Azure needed): Set LLM_PROVIDER=anthropic and ANTHROPIC_API_KEY
• Azure: Set LLM_PROVIDER=azure_anthropic with Azure endpoint and key

Jira Setup

1. Create an API token at https://id.atlassian.com/manage-profile/security/api-tokens
2. In the app Settings, enter:
   • Jira Base URL (e.g., https://yourcompany.atlassian.net)
   • Your Jira email
   • The API token you created

GitLab Setup (Optional)

1. Create a Personal Access Token in GitLab with read_repository scope
2. In Settings, enter:
   • GitLab URL (e.g., https://gitlab.com)
   • Personal Access Token

Adding Projects

In Settings, add Jira project keys (e.g., PROJ, DEV, SUPPORT) with optional configuration:

• GitLab Projects: Comma-separated repository paths to clone for code context
• Custom Instructions: Additional guidance for Claude when processing meetings
• Embeddings Enabled: Toggle meeting history storage for semantic search

Usage

Meeting Processing Mode

1. Go to Dashboard and select "Meeting" mode
2. Select a project
3. Paste your meeting transcription
4. Click "Process Meeting"
5. Watch Claude analyze and create/update tickets in real-time
6. View results with links to created tickets

Ask Mode

1. Go to Dashboard and select "Ask" mode
2. Select a project
3. Ask a question about your project
4. Claude searches Jira, meeting history, and code to provide answers

Work Mode

1. Go to Dashboard and select "Work" mode
2. Select a project
3. Enter a Jira ticket key (e.g., PROJ-123)
4. Claude clones the configured GitLab repos, reads the ticket, and starts working
5. Watch progress in real-time as Claude analyzes code and implements changes

Meeting History

1. Go to History to view past processed meetings
2. Use search to find meetings by content
3. Click on a meeting to view details and summary

Environment Variables

Core Settings

Variable	Required	Description
SECRET_KEY	Yes	JWT signing secret (change in production)
DATABASE_URL	No	Database connection string (default: SQLite)

LLM Configuration

Variable	Required	Description
LLM_PROVIDER	No	LLM provider: anthropic or azure_anthropic (default: azure_anthropic)

For Direct Anthropic (LLM_PROVIDER=anthropic):

Variable	Required	Description
ANTHROPIC_API_KEY	Yes*	Anthropic API key from api.anthropic.com
ANTHROPIC_MODEL	No	Claude model (default: claude-opus-4-20250514)

For Azure Anthropic (LLM_PROVIDER=azure_anthropic):

Variable	Required	Description
AZURE_ANTHROPIC_ENDPOINT	Yes*	Azure Anthropic API endpoint
AZURE_ANTHROPIC_API_KEY	Yes*	Azure Anthropic API key
AZURE_ANTHROPIC_MODEL	No	Claude model (default: claude-opus-4-5)

Embedding Configuration (Optional - for semantic search)

Variable	Required	Description
EMBEDDING_PROVIDER	No	Embedding provider: openai or azure_openai (default: azure_openai)

For Direct OpenAI (EMBEDDING_PROVIDER=openai):

Variable	Required	Description
OPENAI_API_KEY	Yes*	OpenAI API key from platform.openai.com
OPENAI_EMBEDDING_MODEL	No	Embedding model (default: text-embedding-3-small)

For Azure OpenAI (EMBEDDING_PROVIDER=azure_openai):

Variable	Required	Description
AZURE_OPENAI_ENDPOINT	Yes*	Azure OpenAI endpoint (for embeddings)
AZURE_OPENAI_API_KEY	Yes*	Azure OpenAI API key (for embeddings)
AZURE_OPENAI_EMBEDDING_DEPLOYMENT	No	Embedding deployment name (default: text-embedding-3-small)

API Endpoints

Authentication

Endpoint	Method	Description
/api/auth/register	POST	Create account
/api/auth/login	POST	Login (returns JWT)
/api/auth/me	GET	Get current user

Jira Configuration

Endpoint	Method	Description
/api/jira/config	GET	Get Jira config
/api/jira/config	POST	Create Jira config
/api/jira/config	PUT	Update Jira config

Projects

Endpoint	Method	Description
/api/jira/projects	GET	List projects
/api/jira/projects	POST	Add project
/api/jira/projects/{id}	PUT	Update project settings
/api/jira/projects/{id}	DELETE	Remove project

Processing

Endpoint	Method	Description
/api/meetings/process	POST	Process meeting transcription
/api/jira/ask	POST	Ask a question about the project
/api/work/start	POST	Start working on a Jira ticket
/api/processing/status	GET	Check processing status
/api/processing/abort	POST	Cancel current processing

Meeting History

Endpoint	Method	Description
/api/meetings	GET	List meetings (paginated)
/api/meetings/{id}	GET	Get meeting details
/api/meetings/{id}	DELETE	Delete meeting
/api/meetings/search	POST	Semantic search across meetings

Real-time

Endpoint	Type	Description
/ws?token={jwt}	WebSocket	Real-time processing updates

Deployment

Docker

docker build -t actionsync .
docker run -p 8080:8080 --env-file .env actionsync

Tech Stack

• Python 3.12 with FastAPI
• PostgreSQL 16 with pgvector extension
• Claude AI via Azure Anthropic
• Azure OpenAI for embeddings
• Vanilla JavaScript SPA frontend
• Docker deployment support