CoHost AI Studio
Nothing publishes until it clears the gate.
An automated post-production pipeline for independent podcast studios, with a quality gate that scores eleven dimensions before anything publishes.
- languages
- Python 3.11, TypeScript / React 19, Rust (Tauri)
- pipeline
- 23-step DAG, 4 profile variants
- quality gate
- 11 programmatic scorers + 5 LLM evals
- tests
- 6,029 Python test functions
- history
- 2,462 commits, Feb to May 2026
- repo
- Private
The problem
Independent podcast producers run every episode through five to eight disconnected tools, with a manual handoff at each step and no signal when a bad master or a thin transcript slips through to publish. CoHost collapses that into one instrumented pipeline where nothing ships until quality is verified.
How it is built
-
A DAG pipeline with idempotency and fault isolation
Twenty-three steps run as a topologically sorted DAG on a thread pool. Each step has an idempotency predicate, so partial runs resume cleanly, and per-step cascade behavior isolates a failure (a stitch timeout does not block show-notes generation). The graph is validated for cycles at build time, before the first episode runs.
-
An instrumented quality gate, not a vibe check
Eleven steps are scored against programmatic metrics: LUFS delta for audio mastering, word confidence and speaker resolution for transcripts, artifact completeness for video. A weighted composite plus a per-step hard-fail threshold wires straight into the publish decision; below the bar, the pipeline halts before distribution.
-
LLM evals layered over deterministic scores
Five content-heavy steps add opt-in LLM evaluation via OpenRouter, each returning a criterion, score, and reasoning that merge into the scorecard alongside the programmatic metrics. Fast deterministic checks always run; the slower qualitative ones are opt-in.
By the numbers
Where this honestly stands
Live, pre-launch. Four of five P0 launch criteria are validated; YouTube OAuth and multi-show isolation remain open. Functional and test-validated for single-show local operation.
Want the parts that are not in a public repo? I will walk you through the architecture and the decisions on a call.