apowerb (by thaink2) is an open-source agentic platform and developer framework designed to build, run, and scale AI agents. It bridges the gap between raw LLMs and enterprise execution by combining declarative database-backed agent state, Google ADK execution, and LiteLLM multi-model interoperability.

The framework ships as a Python package providing:

• 🚀 A high-performance FastAPI server serving REST endpoints and SSE streams.
• 💻 A Typer-powered CLI for running, testing, and managing local workflows.
• 🖥️ A companion web UI (apowerb-ui) built with Next.js for visual prompt authoring, monitoring, and debugging.

 ┌────────────────────────────────────────────────────────┐
 │                      Clients / UI                      │
 │    Next.js UI (apowerb-ui)  •  Typer CLI  •  REST/SSE  │
 └──────────────────────────┬─────────────────────────────┘
                            │
 ┌──────────────────────────▼─────────────────────────────┐
 │                FastAPI Orchestrator Layer               │
 │  • Agent Module Materialization   • Webhooks (Pub/Sub) │
 │  • Cron / Event Scheduler         • SSE Output Stream  │
 └──────────────────────────┬─────────────────────────────┘
                            │
 ┌──────────────────────────▼─────────────────────────────┐
 │               Execution Engine (Google ADK)            │
 │  • Base      • Parallel       • Sequential     • Loop  │
 │  • Hierarchical Sub-Agents    • Human-in-the-loop      │
 └─────────────┬───────────────────────────┬──────────────┘
               │                           │
 ┌─────────────▼─────────────┐   ┌─────────▼──────────────┐
 │   LiteLLM Model Router    │   │  Tool Integrations (24+)│
 │ OpenAI, Anthropic, Gemini │   │ M365, Google, SQL, RAG │
 └───────────────────────────┘   └────────────────────────┘

Choose from four built-in execution patterns depending on task complexity:

• base: Single-turn and direct conversational agents with tool execution.
• sequential: Step-by-step pipelines passing context down deterministic stages.
• parallel: Concurrent branch execution aggregating results for low-latency synthesis.
• loop: Evaluator-optimizer cycles that iterate until target conditions or stop tokens are met.
• Hierarchical sub-agents: Delegate specialized tasks to domain-specific worker agents.

Connect to any major model provider (OpenAI, Anthropic, Google Gemini, Mistral, Groq, local Ollama/vLLM) with unified configuration, fallback logic, and zero changes to underlying agent tools.

Incorporate retrieval without external plumbing:

• Automatically parse, chunk, and embed documents, URLs, Amazon S3 objects, and SQL query records.
• Isolated, per-agent knowledge bases searchable via semantic vector indices.

Connect Postgres, MySQL, or Snowflake instances. apowerb introspects table schemas, generates parameterized queries, and runs validated executions against analytical backends.

• Webhooks: Inbound push notifications via Google Cloud Pub/Sub (Gmail) and Microsoft Graph (Outlook) to trigger autonomous email processing agents.
• Scheduling: Native cron-style schedules for recurring background audits, report generation, and polling tasks.
• Real-time SSE: Stream token-by-token completions, RAG ingestion progress, and agent state transitions in real time.

First-class support for enterprise SaaS and developer utilities:

• Workplaces: Google Workspace (Docs, Sheets, Drive, Gmail) and Microsoft 365 (Outlook, Teams, OneDrive).
• Data & Storage: AWS S3, Cloud Object Storage, PostgreSQL, Vector Stores.
• Web & Code: GitHub API, SerpAPI, Tavily, headless browsing, and data visualization generators.

Run PostgreSQL, the apowerb backend engine, and apowerb-ui with a single command:

git clone https://github.com/apowerb/apowerb.git
cd apowerb
cp .env.example .env
docker compose up -d

Visit the dashboard at http://localhost:3000 or inspect the API OpenAPI documentation at http://localhost:8000/docs.

Prerequisites:

• Python 3.12 or 3.13
• PostgreSQL instance running locally or remotely
• uv (recommended) or pip

# Install with uv
uv add apowerb

# Or with pip
pip install apowerb

Configure your environment:

export DATABASE_URL="postgresql+asyncpg://postgres:password@localhost:5432/apowerb"
export OPENAI_API_KEY="your-api-key"

Initialize the database schema and start the service:

# Run migrations
apowerb db upgrade

# Start the server
apowerb server --host 0.0.0.0 --port 8000 --reload

# List all active configured agents
apowerb agents list

# Execute an agent directly from terminal
apowerb run --name "support-agent" --prompt "Summarize pending tickets from yesterday"

# Start webhook receiver daemon
apowerb worker --queues default,webhooks

We welcome pull requests, tool contributions, and bug reports from the developer community!

1. Fork the repo and create your branch from main.
2. Follow our Development Guide.
3. Ensure linting and tests pass:

   uv run ruff check .
   uv run pytest

4. Check out the Release Process to see how PRs are verified and released.

• 📖 Documentation: docs.apowerb.com
• 💬 Discord Community: Join apowerb on Discord
• 🐛 Bug Tracker: GitHub Issues
• 🌐 Company Website: thaink2.com