AI-powered podcast generation that sounds human, works like software.

Totally not a versificator. Honest.

Episodic is an automated podcast generation platform that transforms your content into professional, scripted audio shows. Using AI orchestration with human oversight, it takes heterogeneous source documents and produces broadcast-quality podcasts with proper tone, compliance checks, and audio mastering.

Think of it as your podcast production team, but with:

• Intelligent content generation using LLMs with multi-layer quality assurance
• Human-in-the-loop workflows for editorial approval and brand compliance
• Professional audio synthesis with TTS, background music, and loudness normalization
• Hexagonal architecture for clean boundaries and testable components
• Cloud-native infrastructure built on Kubernetes with GitOps deployment

Unlike certain dystopian content machines from Orwell's imagination, Episodic emphasizes auditability, compliance, brand guidelines, and human oversight at every stage. It's a tool for creators, not a replacement.

This project is in active early development. We're currently building the platform foundations (Phase 1 of 6).

📋 See docs/roadmap.md for detailed development status and planned features.

• Multi-source document ingestion with conflict resolution
• Canonical TEI (Text Encoding Initiative) content representation using tei-rapporteur, our TEI P5 serialization library for Rust and Python
• Automated provenance tracking and audit trails
• Series profiles and episode templates

• LLM-based script generation with retry and guardrails
• Multi-layer quality assurance (factuality, tone, style)
• Brand guideline compliance checking
• Structured output planning with model tiering for cost control
• LangGraph suspend-and-resume workflows

• Text-to-speech with configurable voice personas
• Background music and sound effect integration
• Professional mixing with ducking and transitions
• Loudness normalization to broadcast standards (-16 LUFS)
• Chapter markers and metadata embedding

• Cost accounting with token usage metering
• Budget enforcement per user/organization
• Comprehensive observability (Prometheus, Loki, Tempo)
• GitOps-driven deployments with FluxCD
• Editorial approval workflows with SLA tracking

• Language: Python 3.14+ (with optional Rust extensions)
• Web Framework: Falcon 4.2.x on Granian (Rust-based ASGI server)
• Task Queue: Celery with RabbitMQ
• Orchestration: LangGraph for agentic workflows
• Database: PostgreSQL (CloudNativePG)
• Cache: Valkey (Redis-compatible in-memory data store)
• Infrastructure: Kubernetes (DigitalOcean DOKS), OpenTofu/Terraform
• Deployment: FluxCD GitOps
• Testing: pytest with BDD (pytest-bdd)

Episodic is built using hexagonal architecture (ports and adapters) to maintain clean boundaries between domain logic and infrastructure concerns. This ensures testability, flexibility, and long-term maintainability.

• Python 3.14+
• uv (fast Python package and project manager)
• Make
• Docker (for local development)

# Install dependencies
uv sync

# Run tests
make test

# Run linters and type checks
make lint
make typecheck

📖 See docs/users-guide.md for the knick-knacks.

User guide coming soon as features are implemented!

Note: The Python package does not expose a stable public API yet. The placeholder hello helper has been removed, so treat the module surface as internal until a supported API is documented.

• Roadmap — Development phases and current status
• System Design — Architecture and component overview
• Infrastructure Design — Kubernetes, GitOps, and observability
• Users Guide — Usage instructions (coming soon)

This project follows strict code quality standards:

• Linting: Ruff with comprehensive rule sets
• Type Checking: Pyright in strict mode
• Testing: pytest with BDD scenarios
• Complexity: Cyclomatic complexity limits enforced
• Architecture: Hexagonal boundary enforcement via lint rules

See AGENTS.md for contributor guidelines and commit gating requirements.

Licensed under the ISC Licence.

Developed by df12 Productions 🌐 https://df12.studio

Remember: With great automation comes great responsibility. Use this power wisely, and keep a human in the loop. 🤖✨