Keep a long job resumable
Keep a long job on the person's computer (a meeting summary, a transcription) under a key, so starting it twice gives the same job, its output so far is kept in numbered pieces, and any client can resume reading from a cursor or cancel it
Give action: start (key, and seed and version when wanted), append (id, piece), read (id, cursor: the last seq read, 0 for all), finish (id), cancel (id) or list. The jobs are kept in a file in the folder the person granted. Returns the job (id, state, stream_url, expires_at, seed, version, chunks), { seq } for append, { job, pieces, cursor } for read and { jobs } for list.
What it asks for
Nothing: your AI starts it with your words.
Use it from your AI
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to keep a long job on the person's computer (a meeting summary, a transcription) under a key, so starting it twice gives the same job, its output so far is kept in numbered pieces, and any client can resume reading from a cursor or cancel it”