space-ocrlisted

# space ocr space ocr reads a document image on its own servers and returns **structured fields**, each tied to the **exact spot on the page** it was read from (a 4-point box re-anchored to real Vision-API symbols, not an LLM guess). It also gives you **storage** — folders, sheets, and rows — so the documents you process become a queryable workspace **without you building a database or a vector store**. The API is the storage. Everything is driven through one stdlib-only script: [`scripts/space_ocr.py`](scripts/space_ocr.py). There is nothing to install — only `python3`. ## Setup 1. Get an API key at https://space-ocr.com → Developer → API Keys in the top menu (100 free scans, no card). 2. Put it where the script can read it — either: - `export SPACE_OCR_API_KEY=spocr_...`, or - a `.env` file in the project root (see [`.env.example`](.env.example)). 3. That's it. Test with `python3 scripts/space_ocr.py balance`. Optional: `SPACE_OCR_API_BASE` overrides the endpoint (defaults to `https://api.space-ocr.com`). ## Commands Run `python3 scripts/space_ocr.py <command> --help` for flags. All output is JSON on stdout. | Command | What it does | Maps to | |---|---|---| | `balance` | Remaining free / flat-fee / paid scans | `GET /amount` | | `ocr <image> [--auto \| --template ID \| --fields FILE]` | Extract fields from one image (URL, file path, or base64). Prefer `--auto`: the engine looks at the page and picks the fields itself, with the same gating as `/ocr/suggest_fie