The language model
The summaries and proposed filings come from a model you choose. The three kinds of provider, exactly what each document sends, and what it costs.
On this page
The summary, the proposed title, category, dates, tags and reminders on each Inbox card come from a language model. Which one is yours to choose, under Settings, Suggestions, with a Test button that makes a real call before saving anything. The key is sealed on the box and never returned to the browser.
Three kinds of provider#
| Provider | Covers | You give it |
|---|---|---|
| Anthropic | Claude, through the official SDK. Requires a provider account and API key. | An API key. |
| OpenAI-compatible | OpenAI, Groq, Mistral, DeepSeek, OpenRouter, and anything else that speaks the same chat endpoint, including Ollama, llama.cpp and LM Studio on your own network. | A base URL, a model name, and a key if the service wants one. |
| None | Heuristics only: the sender becomes the item, a date in the filename becomes the document date. Cards arrive without a summary. | Nothing. |
The commercial services are presets: choosing one fills in the base URL and a model name and asks for a key.
To keep AI processing on your network, connect a local provider such as Ollama. This keeps suggestion requests local; email connections, offsite backups, and other configured services may still use the internet. Summary quality and processing speed depend on the model and hardware. Test it with representative documents and review the results.
What leaves the house, per document#
The filename, the sender's address if it came by email, the page count, the first four thousand characters of its text, and your list of category names and item labels, which includes people's first names. Harbor sends extracted text rather than the original file or page images. For a short document, that text may include its entire contents. The item list can be withheld with SUGGEST_SEND_PEOPLE=false, at the cost of the "who is this about" prefill.
What was sent is reconstructible from the document's text and the recorded model, prompt version and character range; it is not stored twice.
What never leaves#
The OCR worker has no direct internet access. Other services connect as needed: the suggester to your model provider, the mail fetcher to your email hosts, and the backup service to your backup destination. See What runs for the service layout.
What it costs#
Hosted providers charge separately from Harbor. Costs depend on the model, document length, number of requests, and any reprocessing you initiate. Check your provider’s current pricing and usage dashboard before processing a large collection.
A local model avoids hosted API charges but uses your own hardware and electricity. You can also choose None and organize documents without AI summaries.
Changing the model later#
Changing the model or updating a prompt does not automatically regenerate existing suggestions. Reprocessing with a hosted provider may incur charges. To review and reprocess eligible documents, see The harbor command. Documents whose suggestion you already accepted or rejected are skipped.