Spaces:
Paused
Paused
Malina
Claude Fable 5.1
Scoped tokens, bounded archives and frames, build deadline, circuit breaker, honest note caps
da9de03 |
Download README.md from NextNeural/LUNA: direct link, hf CLI and curl.
- Browser
- Download file 3.25 kB
-
https://huggingface.co/spaces/NextNeural/LUNA/resolve/main/README.md
- Command line
-
hf download hf://spaces/NextNeural/LUNA/README.md
-
curl -L -o README.md https://huggingface.co/spaces/NextNeural/LUNA/resolve/main/README.md
3.25 kB
| title: LUNA | |
| emoji: 🌙 | |
| colorFrom: yellow | |
| colorTo: purple | |
| sdk: gradio | |
| sdk_version: 6.14.0 | |
| python_version: '3.12' | |
| app_file: app.py | |
| pinned: false | |
| license: apache-2.0 | |
| short_description: Latent understanding & narrative assembly from your files | |
| # LUNA - Latent Understanding & Narrative Assembly | |
| LUNA is the file-synthesis tool of [Next Neural Labs](https://nnlabs.pl). Drop up to | |
| **100 files of any type** (documents, spreadsheets, code, images, audio, video) into a | |
| **task**, tell LUNA *what to build* (e.g. *"Everything about climate change in Italy"*), | |
| and it returns **one structured Note** - Title, Agenda, themed Body, and a Sources | |
| section that credits every file it actually used. | |
| ## How it works | |
| 1. **Categorize** - each upload is sorted by type (text / document / image / audio / | |
| video / spreadsheet / code / …). | |
| 2. **Extract in parallel** - content is pulled from every file concurrently to keep cost | |
| and latency low: documents/sheets/code are parsed to text, images are read by a | |
| vision-language model (description + on-screen text), audio/video are transcribed with | |
| Whisper. | |
| 3. **Map (note-taking)** - a fast model reads each file (long files in parts, in | |
| parallel) and writes notes on everything it contains. Nothing is judged or thrown | |
| away at this stage, so a shaky per-file call can't silently lose a file's content. | |
| 4. **Reduce (synthesis)** - a strong model sees every file's extraction together and is the | |
| final judge of what to use. It fuses the relevant material into the Note (in the language | |
| of your prompt). If your prompt targets one thing (e.g. *"what's on the video"*), it | |
| answers from that and sets the unrelated uploads aside instead of summarizing them "just | |
| in case". A file is either *used* (in the body + Sources) or *not used* (one short line) - | |
| never both. | |
| The front-end lives at **https://app.nnlabs.pl/luna**. Uploaded files are processed and | |
| then purged (after one hour at the latest); only the generated Note is stored, in the | |
| website's database. | |
| ## Running this Space | |
| * **Secrets:** `HF_TOKEN` (inference providers) and `LUNA_SIGNING_SECRET` (shared with the | |
| website's Worker; without it every call is refused). Optional: `TOKEN_REQUIRE_SCOPE=1` | |
| once the Worker emits scoped tokens. | |
| * **Hardware:** CPU (`cpu-upgrade`); sleeps after 15 minutes and takes about a minute to | |
| wake. Whisper `base` is loaded in the background at start. | |
| * **Limits:** 100 files, 25 MB per document, 500 MB per media file and per task; media is | |
| transcribed up to 30 minutes; reading and note-taking stop at 6 minutes and the Note is | |
| written from what was read by then. | |
| ## API | |
| Every endpoint takes a signed, single-use token minted by the website for that | |
| endpoint's scope. There is no anonymous access. | |
| | api_name | inputs | output | | |
| | --- | --- | --- | | |
| | `build` | files, prompt (may end with `⟦FORMAT⟧` directives), token | streams `[status, note]`; the last status is `✅ done` or `⚠️ degraded: reason` | | |
| | `clarify` | prompt, token | one or two sentences restating the request | | |
| Unsupported formats (`.doc`, `.ppt`, `.rtf`, `.xls`, `.ods`) are listed under "Not used" | |
| with a hint to convert them. Archives are read member by member through the same parsers. | |