wrenautomation

Keep the credentials an automation signs in with: sealed at rest, shared through SSM, every use audited, expiries watched.

The web in, documents out. Polite HTTP fetching with robots and a fetch ladder up to a browser; YouTube, any other video and podcasts read from a URL through yt-dlp and @wrenautomation/falcon; what a channel or search holds. The sources, runner and ledger

Ask in, plan out. The planner of Wren's research engine: typed queries with who proposed them and why, proposals merged into one plan under a plan file per ask kind, read-or-skip on hits, and the edge to take after a round. Pure functions and prompts; the

One door to every model: a Model takes text, image, video, audio and PDF parts, answers with text, tokens and USD, and parses an answer to a zod shape. Adapters for Anthropic, Gemini, Cohere, any OpenAI-compatible server and a shell command (claude -p); e

Research from the open web, driven by an agent one step per command: plan, search, triage, open, extract behind a quote gate, check, review, every record in a ledger. The agent decides; the engine runs the fixed steps. The ask and plan files, the document

The web and files in, text out. Polite HTTP fetching with robots and a fetch ladder up to a browser; HTML and PDF to text, a page to title, text and links; a video or podcast from a URL or a file (captions, ffmpeg and a vision model, audio to markdown, wh