What is complete, what is active, and what is still planned

Design and implement the database schema and data access patterns. Completion enables persistent storage of all canonical entities.

• ☑ Complete 1.1.1 Design the relational schema covering core entities.
• ☑ Complete 1.1.2 Introduce migration tooling with Alembic.
• ☑ Complete 1.1.3 Implement the repository and unit-of-work layers.

Build the ingestion pipeline that normalizes heterogeneous inputs into canonical TEI. Completion enables content acquisition from diverse sources.

• ☑ Complete 1.2.1 Build the multi-source ingestion service.
• ☑ Complete 1.2.2 Implement source weighting heuristics and conflict resolution.
• ☑ Complete 1.2.3 Capture provenance metadata automatically in TEI headers.

Define and implement the reference-document model for reusable content across series and episodes. Completion enables shared style guides, character profiles, and research briefs.

• ☑ Complete 1.3.1 Define the reusable reference-document model.
• ☑ Complete 1.3.2 Publish approved Entity-Relationship (ER) diagram and glossary.
• ☑ Complete 1.3.3 Document repository and API contract acceptance criteria.
• ☑ Complete 1.3.4 Implement Representational State Transfer (REST) endpoints for reusable reference documents. Requires 1.3.1

Define configuration entities and implement binding resolution for reusable content. Completion enables downstream generators to retrieve structured briefs.

• ☑ Complete 1.4.1 Define series profile and episode template models.
• ☑ Complete 1.4.2 Implement REST endpoints for series profiles and episode templates.
• ☑ Complete 1.4.3 Implement reference-binding resolution. Requires 1.3.1 Completed 2026-03-28. Architectural Decision Record (ADR) 001, the resolver service, brief/API integration, ingestion provenance snapshotting, behavioural coverage, and documentation alignment are all in place.

Scaffold the application services with consistent architecture patterns. Completion enables feature development within enforced boundaries.

• ☑ Complete 1.5.1 Scaffold Falcon 4.2.x HTTP services running on Granian.
  • ☑ Complete Integrate the Granian runtime for Falcon deployment.
• ☑ Complete 1.5.2 Scaffold Celery workers with RabbitMQ integration.
• ☑ Complete 1.5.3 Define hexagonal architecture boundaries and port contracts.
• ☑ Complete 1.5.4 Implement architectural enforcement checks for hexagonal boundaries.

Implement the inference boundary with retry semantics, token budgeting, and content guardrails. Completion enables controlled LLM invocation from orchestration code.

• ☑ Complete 2.1.1 Implement the LLMPort adapter with retry and token budgeting.
• ☑ Complete 2.1.2 Implement guardrail prompts aligned to content templates.

Implement internal quality evaluators that score generated scripts before editorial approval. The initial implementation keeps Pedante, Bromide, Chiltern, Anthem, and Caesura inside LangGraph as structured Large Language Model (LLM) calls that return typed findings and normalized usage metrics per invocation; Chrono is a non-LLM runtime estimator. Later iterations may replace individual evaluators with more cost-effective or service-oriented approaches without changing the orchestration contract.

• ☑ Complete 2.2.1 Implement Pedante for factuality and accuracy checks.
• ☐ Not started 2.2.2 Implement Bromide to detect overused tropes, cliches, and snowclones.
• ☐ Not started 2.2.3 Implement Chiltern to flag pronunciation gaps.
• ☐ Not started 2.2.4 Implement Anthem for brand-guideline evaluation.
• ☐ Not started 2.2.5 Implement Caesura to detect false endings in scripts.
• ☑ Complete 2.2.6 Implement Chrono for runtime estimation.
• ☐ Not started 2.2.7 Persist QA artefacts linked to canonical episodes.

Enrich canonical TEI bodies with structured metadata sourced from templates. Completion enables rich episode content with chapter markers and guest details.

• ☑ Complete 2.3.1 Generate show notes from template expansions.
• ☑ Complete 2.3.2 Generate chapter markers aligned to script segments.
• ☑ Complete 2.3.3 Generate guest bios from reference document bindings.
• ☐ Not started 2.3.4 Generate sponsor reads from template-defined placements.

Implement resumable orchestration with idempotency, queue routing, and cost metering. Completion enables reliable, auditable generation workflows.

• ☑ Complete 2.4.1 Implement structured-output planning and tool-calling execution.
• ☑ Complete 2.4.2 Add LangGraph suspend-and-resume orchestration. Requires 2.1.1
• ☑ Complete 2.4.3 Configure Celery queue routing for workload isolation.
• ☑ Complete 2.4.4 Instrument cost accounting with per-call usage metering.
• ☐ Not started 2.4.5 Extend architecture enforcement to orchestration code.

Implement price snapshots, discovery, metering, and budget reservation for LangGraph nodes. Completion enables current internal metering and preserves the future path to standalone evaluator services with explicit pricing contracts.

• ☐ Not started 2.5.1 Add immutable pricing snapshots for provider and helper pricing inputs.
• ☐ Not started 2.5.2 Add PricingCataloguePort for pricing-source discovery.
• ☐ Not started 2.5.3 Add PricingEngine with concurrency-safe metering.
• ☐ Not started 2.5.4 Extend CostLedgerPort storage for evaluator-node and provider calls.
• ☐ Not started 2.5.5 Extend BudgetPort APIs for reservation workflow.

Define the generation-run domain model and implement API endpoints. Completion enables real-time visibility into generation progress.

• ☑ Complete 2.6.1 Define GenerationRunPort and implement domain model. Requires 2.4.2
• ☐ Not started 2.6.2 Implement repository contracts and Alembic migrations.
• ☐ Not started 2.6.3 Implement REST endpoints for generation runs. Requires 2.6.1

Implement structured script projection for editing without direct TEI manipulation. Completion enables user-friendly script review and patching.

• ☐ Not started 2.7.1 Define ScriptProjectionPort and projection schema.
• ☐ Not started 2.7.2 Implement script projection read endpoint.
• ☐ Not started 2.7.3 Implement script projection patch endpoint. Requires 2.7.2

Define the provider-neutral speech boundary before binding to any one TTS engine. Completion answers whether the audio graph can express the behaviours needed by Inworld, Chatterbox, ElevenLabs v3, and future engines without leaking provider parameter names into the domain model.

• ☐ Not started 3.1.1 Define the TTSPort request and result contracts.
• ☐ Not started 3.1.2 Define the DialogueSpeechPort contract for multi-speaker conversation rendering. Requires 3.1.1
• ☐ Not started 3.1.3 Implement the TTS capability registry and selector. Requires 3.1.1
• ☐ Not started 3.1.4 Implement configurable voice persona support. Requires 3.1.1

Build the pronunciation system as a reusable repository connected to TEI annotations, QA findings, and audio feedback. Completion answers whether mispronunciation fixes can be captured once, audited, and rendered through multiple providers with different pronunciation surfaces.

• ☐ Not started 3.2.1 Define PronunciationRepositoryPort and pronunciation entry models.
• ☐ Not started 3.2.2 Implement pronunciation-pack resolution. Requires 3.2.1
• ☐ Not started 3.2.3 Connect Chiltern findings and audio feedback to pronunciation proposals. Requires 3.2.1
• ☐ Not started 3.2.4 Define provider-neutral performance cues in TEI-backed speech markup. Requires 3.1.1

Implement concrete adapters that prove the speech contracts support different vendor shapes. Completion answers whether the system can mix direct segment TTS, reference-audio voice cloning, and dialogue rendering whilst preserving TEI-backed traceability.

• ☐ Not started 3.3.1 Implement the Inworld TTS adapter. Requires 3.1.3, 3.2.2
• ☐ Not started 3.3.2 Implement the Inworld realtime dialogue adapter. Requires 3.1.2, 3.3.1
• ☐ Not started 3.3.3 Implement the Chatterbox via FAL adapter. Requires 3.1.3, 3.2.2
• ☐ Not started 3.3.4 Implement the ElevenLabs v3 via FAL TTS adapter. Requires 3.1.3, 3.2.2
• ☐ Not started 3.3.5 Implement the ElevenLabs v3 via FAL dialogue adapter. Requires 3.1.2, 3.2.2
• ☐ Not started 3.3.6 Add adapter contract scenarios across capability combinations. Requires 3.3.1, 3.3.2, 3.3.3, 3.3.4, 3.3.5

Build the audio mixing engine with stem management and loudness normalization. Completion enables broadcast-quality audio production.

• ☐ Not started 3.4.1 Integrate background music and sound effect stem management.
• ☐ Not started 3.4.2 Build the mixing engine for narration and stem combination.
• ☐ Not started 3.4.3 Enforce loudness normalization to broadcast standards.

Define the audio-run domain model and implement run lifecycle endpoints. Completion enables audio workflow tracking and iterative refinement.

• ☐ Not started 3.5.1 Define AudioRunPort and implement domain model. Requires 3.1.1, 3.4.2
• ☐ Not started 3.5.2 Implement repository contracts and Alembic migrations.
• ☐ Not started 3.5.3 Implement REST endpoints for audio runs. Requires 3.5.1
• ☐ Not started 3.5.4 Generate shareable previews via PreviewPublisherPort.

Implement standalone voice preview synthesis and export job workflows. Completion enables persona testing and deliverable packaging.

• ☐ Not started 3.6.1 Define VoicePreviewPort and implement synthesis endpoints. Requires 3.1.1, 3.1.4, 3.3.1, 3.3.3, 3.3.4
• ☐ Not started 3.6.2 Define ExportJobPort and implement export endpoints. Requires 3.4.2, 3.4.3
• ☐ Not started 3.6.3 Publish final masters to Content Delivery Network (CDN) endpoints.

Finalize REST API surfaces with consistent contracts and version routing. Completion enables stable API consumption by clients.

• ☑ Complete 4.1.1 Introduce /v1 as the target API prefix.
• ☑ Complete 4.1.2 Finalize REST surfaces for previous phase artefacts.
• ☐ Not started 4.1.3 Integrate request correlation across HTTP, tasks, and outbound provider calls. Requires 1.5.1, 1.5.2

Implement episode lifecycle endpoints and the editorial approval state machine. Completion enables structured approval workflows with audit trails.

• ☐ Not started 4.2.1 Implement editorial approval state machine. Requires 1.1.1
• ☐ Not started 4.2.2 Implement episode REST endpoints. Requires 4.2.1
• ☐ Not started 4.2.3 Integrate notification channels for approval events.

This step answers whether an integration client can drive the narrowest useful source-to-script workflow through JSON/REST: submit show source material, provide presenter context, trigger draft generation without QA, and retrieve the resulting TEI-P5 XML. It validates the /v1 resource contract before the full approval, QA, audio, and export surfaces land.

• ☑ Complete 4.3.1 Implement source and presenter-profile intake for script generation. Requires 1.3.4, 1.4.3, 4.1.1
• ☐ Not started 4.3.2 Implement no-QA generation runs and TEI-P5 retrieval. Requires 2.1.1, 2.4.2, 4.3.1

This step answers whether a single generation run can complete a generation, quality gate, and refinement cycle end-to-end: the run should trigger Chrono on the generated draft, then route short or long drafts back into the same agent loop for targeted redraft. This prepares the implementation for QA-gated generation workflows beyond the initial draft-only slice. Implementations must define single-writer ownership for run state, ordered event publication, and shutdown-safe cancellation before adding new loop transitions. Concurrency model: one graph runner owns mutation for a run, repository writes are guarded by sequence numbers and idempotency keys, and cancellation, shutdown, duplicate-event handling, and resume replay must be deterministic. Property-testing guidance: use Hypothesis to exercise routing, duration-class, and iteration-cap invariants across generated run states.

• ☐ Not started 4.4.1 Add QA-gated mode branching and Graph route plumbing. Requires 2.1.1, 2.2.6, 2.4.2, 4.3.2
• ☐ Not started 4.4.2 Invoke Chrono in-loop and emit duration-based QA findings. Requires 2.2.6, 4.4.1
• ☐ Not started 4.4.3 Trigger targeted redraft from duration outcomes. Requires 4.4.2, 2.4.2
• ☐ Not started 4.4.4 Persist iterative quality context and resume points. Requires 2.6.1, 2.5.5, 4.4.3

Implement real-time event streaming for generation runs. Completion enables live workflow observation and checkpoint intervention.

• ☐ Not started 4.5.1 Define RunEventBusPort and implement WebSocket streaming for generation runs. Requires 2.6.1
• ☐ Not started 4.5.2 Implement WebSocket backpressure and reconnection. Requires 4.5.1

Extend command-line tooling and ship the initial web console. Completion enables operator and editorial self-service.

• ☐ Not started 4.6.1 Extend CLI client for approval workflows.
• ☐ Not started 4.6.2 Ship initial web console for editorial workflows.

Publish OpenAPI and AsyncAPI specifications for all API surfaces. Completion enables client SDK generation and contract validation.

• ☑ Complete 4.7.1 Publish TUI API design document.
• ☐ Not started 4.7.2 Generate OpenAPI specification from endpoint definitions.
• ☐ Not started 4.7.3 Generate AsyncAPI specification from WebSocket schemas.

Implement fine-grained access control, tenant boundaries, and credential lifecycle management. Completion enables secure multi-tenant operation.

• ☐ Not started 5.1.1 Implement fine-grained RBAC across all services.
• ☐ Not started 5.1.2 Implement tenancy isolation for multi-organization support.
• ☐ Not started 5.1.3 Implement automated secrets rotation.

Add security scanning, dependency auditing, and policy-as-code controls. Completion enables proactive vulnerability management.

• ☐ Not started 5.2.1 Add runtime security scanning for container images.
• ☐ Not started 5.2.2 Add dependency auditing with automated alerts.
• ☐ Not started 5.2.3 Implement policy-as-code enforcement in CI/CD.

Implement GitOps-driven recovery drills and backup validation. Completion enables confident incident response.

• ☐ Not started 5.3.1 Roll out GitOps-driven disaster recovery drills.
• ☐ Not started 5.3.2 Implement automated backup verification.
• ☐ Not started 5.3.3 Publish incident runbooks.

Expand monitoring with synthetic probes and customer-facing SLIs; automate compliance evidence collection. Completion enables operational excellence and audit readiness.

• ☐ Not started 5.4.1 Expand observability with synthetic monitoring.
• ☐ Not started 5.4.2 Define and instrument customer-facing SLIs and SLAs.
• ☐ Not started 5.4.3 Certify compliance checkpoints.
• ☐ Not started 5.4.4 Integrate automated compliance evidence collection.

Implement budget controls and cost visibility for operators. Completion enables cost management and financial accountability.

• ☐ Not started 5.5.1 Implement budget enforcement services.
• ☐ Not started 5.5.2 Deliver cost dashboards for operators.

Provision the compute, storage, and messaging infrastructure that underpins the platform. Completion enables service deployments and data persistence.

• ☐ Not started 6.1.1 Bootstrap the Kubernetes control plane on DigitalOcean Kubernetes Service (DOKS).
• ☐ Not started 6.1.2 Deploy CloudNativePG Postgres cluster with high-availability configuration.
• ☐ Not started 6.1.3 Deploy Valkey cache (Redis-compatible) via the Valkey operator.
• ☐ Not started 6.1.4 Deploy RabbitMQ operator with queue definitions for background tasks.
• ☐ Not started 6.1.5 Provision object storage buckets for audio assets and binary artefacts.
• ☐ Not started 6.1.6 Deploy Traefik ingress controller and cert-manager with Let's Encrypt issuers.

Establish declarative deployment pipelines and secure credential handling. Completion enables repeatable, auditable deployments across environments.

• ☐ Not started 6.2.1 Create the GitOps repository via the bootstrap script.
• ☐ Not started 6.2.2 Configure FluxCD with Kustomization overlays for environment promotion.
• ☐ Not started 6.2.3 Configure secret management using SOPS (Secrets OPerationS) and age encryption.

Instrument monitoring, logging, and tracing; publish operational documentation. Completion enables incident response and onboarding.

• ☐ Not started 6.3.1 Instrument clusterwide logging with Loki.
• ☐ Not started 6.3.2 Instrument clusterwide metrics with Prometheus.
• ☐ Not started 6.3.3 Instrument distributed tracing with Tempo.
• ☐ Not started 6.3.4 Define alert routing rules for on-call responders.
• ☐ Not started 6.3.5 Document access controls and networking policies.
• ☐ Not started 6.3.6 Document disaster recovery expectations and restore procedures.
• ☐ Not started 6.3.7 Publish the infrastructure design document.