<img width="2318" height="1448" alt="image" src="https://github.com/user-attachments/assets/bca71d69-4972-420c-b14e-c8cf74a17de9" /> ## Summary Everything a knowledge base could not read, and the pieces it cut them into saying what they are. **Reading a PDF.** A PDF records glyphs at coordinates; whether they are a heading, a table cell or a running footer is inferred from how they look. There are now two readers. The geometric one infers structure from type size and position and needs nothing but the file. The other asks deepdoc's models, ported into `raven/knowledge/parser/deepdoc`: a layout model that says what each region is, a detector and recognizer that read a page carrying no text at all, and a table-structure model that rebuilds a grid from a picture of one. Detecting boxes around text the file already spells out exactly costs an inference pass per page to learn less, and recognising it costs accuracy. The layout graph answers with a shape its own postprocessing raises on, and its classes are DocLayout-YOLO's order rather than the list Both were established by running the model, and the label order is pinned by a test. **Tables go in as tables.** `<table>` is the only shape that keeps what a grid states -- which column a cell stood under, which cells are merged, which row is the header -- and it is the shape a model reads a table in, which is what these chunks are for. PDF, Word and PowerPoint all state their merges exactly, and they now survive as `colspan` and `rowspan`. The flat form this replaces had to repeat a header standing over two columns across both of them, because there was nowhere to record a span; it had no callers left afterwards and is gone. Two deviations from upstream, both about what the string is for: every cell's text is escaped, where upstream interpolates it raw and a cell holding a `<` produces markup that is no longer a table, and the span attributes are written only where there is a span, where upstream emits `<td >` for every ordinary cell. **A picture of where each chunk came from.** A chunk is a few hundred words with a page number on it, which is enough to cite and not enough to answer whether the parser read the page correctly. So every PDF chunk that knows where it sits carries one image: the regions it covers, cropped out of the rendered pages and stacked. here they are already structured in the parser's element spans, so the markers would only put text nobody wrote into the thing being embedded. The panel draws a chunk's tables as tables, parsed and rebuilt as elements rather than injected: a person can rewrite any piece from that same panel, so what comes back from the store is user input however it got there. **The page that reads them.** Upstream retired the knowledge page while this branch was out; it is rebuilt against the new architecture, and the half that never existed is the one a reader needs -- what a document actually became once it was cut and embedded. Documents sit in one-level folders and arrive by four routes: a file, a folder, a URL, a written note. Adding and indexing are two calls on the contract and nothing drains that queue on its own, so a page that stopped after the first left every upload saying Queued for good. Opening a file takes the whole page: the document on the left, the pieces the index holds of it on the right. The left half is the file drawn as its own pages rather than framed by the browser's PDF viewer, which reads where to open once and answers to nothing afterwards -- and which nothing can be drawn on. Pages are rendered once and kept, and the region a piece was cut from is washed over the page it came from, placed in fractions of the page so the picture's own scale never enters the arithmetic. Clicking a piece takes the file there; double-clicking opens it for rewriting. The pieces can be searched, filtered by state, turned on and off, deleted, rewritten and added to, one at a time or in a batch. Every one of those was already on the contract with no reader for it. A base's settings are back and fuller: how many pieces a search returns, how documents are cut, and how much prose comes with a table or a figure. The embedding model is a picker while the base holds no pieces and a stated fact once it does -- moving it then is the collection made again at the new width and every document re-cut, which is not a thing a panel does by being saved. **An agent can read them.** The retrieval engine was there and nothing could reach it: no agent had a tool that opened a library. A reader now points a conversation at bases from the composer, the way they pin a working directory, and the selection rides in the same place -- the session's metadata, bound around the turn. The tool takes a query and never a base. The scope is the reader's choice, so a model cannot reach material nobody offered it, and it is withheld where nothing was picked rather than offered and refused. A hit says which document and page it came from, because a hit a reader cannot place is a hit they cannot check -- and the passages are pressable in the transcript, walking into the base, the document, the folder it is filed under and then the piece. A citation a reader cannot follow is a claim they have to take on trust, and the whole point of searching their own material is that they can go and look. **The word segmenter and the resources.** `raven.core.Tokenizer` is Infinity's `rag_tokenizer`, ported: a forward and backward maximum match over a dictionary trie, with stemming and lemmatisation on the latin path. XGBoost is not carried over. None of the data lives in git -- the dictionary, the NLTK corpora and the ONNX graphs run to about a hundred megabytes, so `make fetch-resources` on an install and an optional Docker layer behind `RAVEN_RESOURCES` both point at one script. A checkout that could not reach the network still works, and every feature that reads this data says so when asked. ## Type - [ ] Fix - [x] Feature - [ ] Docs - [ ] CI / tooling - [ ] Refactor - [ ] Other ## Verification Run on macOS against Python 3.12, with the deepdoc weights fetched (the model tests skip without them): ``` uv run pytest 27450 passed, 111 skipped, 3 failed npm run test --prefix ui-web 3276 passed npm run test --prefix ui-tui 2106 passed npm run type-check --prefix ui-web clean (tsc) npm run lint --prefix ui-web clean (4 pre-existing hook warnings) make lint-python clean (ruff check + format) make lint-types clean (ty) make lint-imports 10 contracts kept, 0 broken make check-source-language clean make check-large-files clean make build-ui built, both boot snapshots match node ui-web/scripts/check-css.mjs clean node ui-web/scripts/check-class-namespace.mjs clean ``` The three Python failures are not this branch's and were each confirmed on the base alone: one asks a host without LibreOffice to refuse a render and this host has LibreOffice installed, and two simulation cases fail on an unmodified checkout of the base commit. The RPC contract was regenerated for both consumers and both match it (`npm run gen --prefix ui-web`, `npm run gen:rpc --prefix ui-tui`, then `gen:check`). The page's boot snapshots and region goldens were refreshed where the composer bar gained a button, and each records that button and nothing else. - [x] Relevant tests pass locally - [x] Relevant lint / type checks pass locally - [x] User-facing docs or screenshots are updated when needed ## Risk A PDF's tables, and a Word or PowerPoint file's, are now indexed as HTML rather than as header-prefixed lines. Chunk text changes shape for those documents, so a base indexed before this and searched after it holds both forms until its documents are reindexed; retrieval keeps working either way, because both forms are text. The same applies to how much prose a table or a figure carries with it: the setting now reaches as far as it says, where before it stopped at the nearest paragraph, so a chunk cut before this keeps the narrower context until its document is rebuilt. Rolling back is reverting the branch and reindexing. The deepdoc models are optional. An install that never ran `make fetch-resources` reads PDFs geometrically, exactly as before, and says so when the models are asked for by name. Nothing here reaches the network at request time. Chunk crops and rendered pages are written at index time and on first view, both under the state directory rather than in the repository, and both are deleted with their document. A page is immutable for the life of its document id, so the route serving them caches; the document route does not. All three routes take ids and no path, and the two image routes answer under the strictest sandbox in the tree. What an agent may search is the reader's choice and nothing else. The tool takes no base argument, so a model cannot name one, and it is absent from the schema entirely for a conversation that named none -- a lane nobody pointed at a base is unchanged by this branch. The bases a session carries are stored beside its working directory and travel no further. - [x] Security impact considered - [x] Backward compatibility considered - [x] Rollback path is clear for risky changes ## Related Issues N/A --------- Co-authored-by: Claude (claude-opus-5) <noreply@anthropic.com>
163 lines
5.8 KiB
Python
163 lines
5.8 KiB
Python
"""Reading a page that carries no text: the detector and the recognizer.
|
|
|
|
Skipped whole when the weights are not installed, which is a checkout where
|
|
`make fetch-resources` has not run.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import asyncio
|
|
|
|
import pytest
|
|
|
|
from raven.knowledge.parser._pdfdoc import open_document
|
|
from raven.knowledge.parser.deepdoc import _onnx, _pipeline
|
|
from raven.knowledge.parser.pdf_parser import PdfParser
|
|
|
|
pytestmark = pytest.mark.skipif(
|
|
not _onnx.available("det", "rec", "layout"), reason="the deepdoc recognition models are not installed"
|
|
)
|
|
|
|
_LINES = [
|
|
"Quarterly report",
|
|
"Revenue grew across every region we measured this quarter.",
|
|
"The northern market remains the largest by some distance.",
|
|
]
|
|
|
|
|
|
def _typed() -> bytes:
|
|
"""A page with a text layer, the ordinary case."""
|
|
import pymupdf
|
|
|
|
document = pymupdf.open()
|
|
page = document.new_page()
|
|
page.insert_text(pymupdf.Point(60, 80), _LINES[0], fontsize=20)
|
|
for index, line in enumerate(_LINES[1:]):
|
|
page.insert_text(pymupdf.Point(60, 140 + index * 30), line, fontsize=12)
|
|
raw = document.tobytes()
|
|
document.close()
|
|
return raw
|
|
|
|
|
|
def _scanned() -> bytes:
|
|
"""The same page as a picture of itself, which is what a scan is.
|
|
|
|
Rendered and pasted back, so the words are pixels and nothing in the file
|
|
states what they are. PyMuPDF reads no characters at all off this.
|
|
"""
|
|
import pymupdf
|
|
|
|
source = pymupdf.open(stream=_typed(), filetype="pdf")
|
|
image = source[0].get_pixmap(matrix=pymupdf.Matrix(2, 2), alpha=False).tobytes("png")
|
|
box = source[0].rect
|
|
source.close()
|
|
|
|
document = pymupdf.open()
|
|
document.new_page(width=box.width, height=box.height).insert_image(box, stream=image)
|
|
raw = document.tobytes()
|
|
document.close()
|
|
return raw
|
|
|
|
|
|
def _parse(raw: bytes, **kwargs) -> str:
|
|
sections = asyncio.run(PdfParser(**kwargs).parse(raw, "report.pdf"))
|
|
return " ".join(section.content.text for section in sections)
|
|
|
|
|
|
def _squeezed(text: str) -> str:
|
|
"""Text with its spaces taken out, which is how recognised text is checked.
|
|
|
|
The recognizer reads a word boundary from the gap between glyphs, and on a
|
|
render of a synthetic page some of those gaps are a pixel too narrow -- one
|
|
line here comes back as `revenuegrewacross`. Real scans are rendered from
|
|
paper and do not do this, so the spacing is a property of the fixture
|
|
rather than of the model, and asserting on it would pin the wrong thing.
|
|
What these tests are about is whether the words were read at all.
|
|
"""
|
|
return "".join(text.lower().split())
|
|
|
|
|
|
# -- what the models are asked -------------------------------------
|
|
|
|
|
|
def test_a_scanned_page_carries_nothing_to_read() -> None:
|
|
"""The premise of everything below. Stated as a test because a fixture that
|
|
quietly kept its text layer would make every assertion here vacuous."""
|
|
import pymupdf
|
|
|
|
with pymupdf.open(stream=_scanned(), filetype="pdf") as document:
|
|
assert document[0].get_text().strip() == ""
|
|
|
|
|
|
def test_the_models_read_the_words_off_the_picture() -> None:
|
|
with open_document(_scanned()) as document:
|
|
page = document[0]
|
|
boxes, recognised = _pipeline.text_boxes(page, _pipeline.render(page), 1)
|
|
|
|
assert recognised, "there was nothing to read, so they were recognised"
|
|
read = _squeezed(" ".join(box["text"] for box in boxes))
|
|
assert "revenuegrewacrosseveryregion" in read
|
|
assert "northernmarketremainsthelargest" in read
|
|
|
|
|
|
# -- what they buy -------------------------------------------------
|
|
|
|
|
|
def test_a_scan_is_indexed_without_a_vision_model() -> None:
|
|
"""The point of carrying the weights. Before this, a scanned PDF needed a
|
|
configured vision model and a call per page; now it needs neither."""
|
|
text = _squeezed(_parse(_scanned()))
|
|
|
|
assert "revenuegrewacrosseveryregion" in text
|
|
assert "northernmarketremainsthelargest" in text
|
|
|
|
|
|
def test_a_scans_heading_is_still_a_heading() -> None:
|
|
"""Recognised text goes through the same layout model as read text, so a
|
|
scan gets the structure a digital page gets rather than one flat block."""
|
|
sections = asyncio.run(PdfParser().parse(_scanned(), "report.pdf"))
|
|
|
|
paths = [path for s in sections if (path := s.metadata.get("heading_path"))]
|
|
assert paths, "the layout model found a heading on the picture"
|
|
assert "quarterly" in " ".join(paths[0]).lower()
|
|
|
|
|
|
# -- when they are not wanted --------------------------------------
|
|
|
|
|
|
def test_a_page_that_states_its_text_is_never_recognised(monkeypatch) -> None:
|
|
"""Recognising text the file spells out exactly costs an inference pass to
|
|
lose accuracy. The text layer wins wherever there is one."""
|
|
|
|
def explode(*args, **kwargs):
|
|
raise AssertionError("the recognizer was asked to read a page that could be read")
|
|
|
|
monkeypatch.setattr(_pipeline, "_from_ocr", explode)
|
|
|
|
assert "northern market remains the largest" in _parse(_typed()).lower()
|
|
|
|
|
|
def test_turning_it_off_leaves_the_page_for_the_vision_model() -> None:
|
|
"""The vision model is the slower and dearer of the two, and the only one
|
|
that can say what a photograph on the page shows. Unconfigured -- which the
|
|
suite guarantees -- the page is reported unread rather than indexed empty."""
|
|
from raven.knowledge._notes import collecting
|
|
|
|
with collecting() as notes:
|
|
text = _parse(_scanned(), ocr=False)
|
|
|
|
assert "revenuegrew" not in _squeezed(text)
|
|
assert any("no vision model is configured" in note for note in notes)
|
|
|
|
|
|
def test_asking_for_it_by_name_says_when_it_is_missing(monkeypatch) -> None:
|
|
"""Silence here is a reader wondering why a scan came out empty."""
|
|
from raven.knowledge._notes import collecting
|
|
|
|
monkeypatch.setattr(_pipeline, "readable", lambda: False)
|
|
|
|
with collecting() as notes:
|
|
_parse(_scanned(), ocr=True)
|
|
|
|
assert any("text recognizer is not installed" in note for note in notes)
|