What problem does it solve?
Pipecat integrates dozens of speech and LLM providers whose models, defaults, and APIs change constantly; this Skill automates the recurring audit that detects new models, deprecations, and unexposed API parameters before users hit breakage.
Core Features & Use Cases
- Per-unit research sweeps: Spawns one researcher subagent per provider/service unit (e.g. cartesia/tts) that probes live APIs, compares catalogues against Pipecat's code, and writes a dated gap report.
- Draft PR proposals: Researchers leave committed branches for clear-cut changes like default model bumps or new Settings fields, which publish.py turns into draft PRs.
- Digest and decision tracking: Aggregates run results into a dated digest issue and maintains per-unit decisions.md files so rejected items are not re-proposed.
- Use Case: Run a weekly sweep across all providers to discover that Cartesia released sonic-4, verify its latency against the current default, and get a ready-to-review branch bumping the default.
Quick Start
Run /provider-watch --only deepgram,groq --limit 2 --concurrency 1 to smoke-test a small research sweep and review the generated reports and branches locally.
