<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>
170 lines
5.8 KiB
Python
170 lines
5.8 KiB
Python
"""Image uploads: one picture, described by a model, indexed as one piece."""
|
|
|
|
from __future__ import annotations
|
|
|
|
import base64
|
|
|
|
import pytest
|
|
|
|
from raven.knowledge._naive_chunker import NaiveChunker
|
|
from raven.knowledge._vision import VisionError
|
|
from raven.knowledge.parser import ELEMENTS, LAYOUT_TYPE, READING_ORDER, LayoutType
|
|
from raven.knowledge.parser.image_parser import ImageParser
|
|
|
|
PNG = base64.b64decode(
|
|
"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk+M9QDwADhgGAWjR9awAAAABJRU5ErkJggg=="
|
|
)
|
|
|
|
|
|
class _Vision:
|
|
max_figures = 64
|
|
|
|
def __init__(self, text: str = "A bar chart of revenue by region.", error: Exception | None = None) -> None:
|
|
self.text = text
|
|
self.error = error
|
|
self.seen: list[tuple[int, str]] = []
|
|
|
|
async def describe(self, image: bytes, *, mime: str = "", context_above: str = "", context_below: str = "") -> str:
|
|
self.seen.append((len(image), mime))
|
|
if self.error is not None:
|
|
raise self.error
|
|
return self.text
|
|
|
|
|
|
# -- what it claims ------------------------------------------------
|
|
|
|
|
|
def test_it_claims_the_raster_formats_a_decoder_can_open() -> None:
|
|
assert "image/png" in ImageParser.supported_media_types
|
|
assert "image/jpeg" in ImageParser.supported_media_types
|
|
|
|
|
|
def test_svg_and_heic_are_left_alone() -> None:
|
|
"""SVG is XML, which a text parser reads better than a model would; HEIC
|
|
needs a Pillow plugin nothing here installs. Offering either would be
|
|
offering an upload that fails on arrival."""
|
|
claimed = " ".join(ImageParser.supported_media_types)
|
|
|
|
assert "svg" not in claimed
|
|
assert "heic" not in claimed
|
|
|
|
|
|
def test_the_picker_gets_the_extensions_people_type() -> None:
|
|
assert ImageParser.supported_extensions() == [
|
|
".bmp",
|
|
".gif",
|
|
".jpeg",
|
|
".jpg",
|
|
".png",
|
|
".tif",
|
|
".tiff",
|
|
".webp",
|
|
]
|
|
|
|
|
|
# -- what it produces ----------------------------------------------
|
|
|
|
|
|
async def test_the_description_is_the_section() -> None:
|
|
vision = _Vision()
|
|
|
|
sections = await ImageParser(vision).parse(PNG, "chart.png")
|
|
|
|
assert len(sections) == 1
|
|
assert sections[0].content.text == "A bar chart of revenue by region."
|
|
assert sections[0].source == "chart.png"
|
|
assert sections[0].metadata[READING_ORDER] == 0
|
|
assert sections[0].metadata[LAYOUT_TYPE] == LayoutType.FIGURE
|
|
|
|
|
|
async def test_the_section_says_it_is_a_figure_over_its_whole_text() -> None:
|
|
"""Which is what the chunker reads to mean "this stands alone" -- so the
|
|
span has to cover everything, not the first line."""
|
|
vision = _Vision(text="A" * 500)
|
|
|
|
sections = await ImageParser(vision).parse(PNG, "chart.png")
|
|
|
|
spans = sections[0].metadata[ELEMENTS]
|
|
assert len(spans) == 1
|
|
assert spans[0]["layout_type"] == LayoutType.FIGURE
|
|
assert (spans[0]["char_start"], spans[0]["char_end"]) == (0, 500)
|
|
|
|
|
|
async def test_the_media_type_travels_from_the_filename() -> None:
|
|
"""A BMP or a TIFF is not in the magic-byte table the preparation step
|
|
sniffs with, so the name is what carries the format through."""
|
|
vision = _Vision()
|
|
|
|
await ImageParser(vision).parse(PNG, "scan.tiff")
|
|
|
|
assert vision.seen[0][1] == "image/tiff"
|
|
|
|
|
|
async def test_a_path_is_read_from_disk(tmp_path) -> None:
|
|
picture = tmp_path / "chart.png"
|
|
picture.write_bytes(PNG)
|
|
vision = _Vision()
|
|
|
|
sections = await ImageParser(vision).parse(str(picture), "chart.png")
|
|
|
|
assert sections[0].content.text
|
|
assert vision.seen[0][0] == len(PNG)
|
|
|
|
|
|
# -- one picture, one chunk ----------------------------------------
|
|
|
|
|
|
async def test_a_long_description_is_still_one_chunk() -> None:
|
|
"""The requirement this parser exists under: an image is one chunk, however
|
|
much the model had to say about it. A description cut in half is two pieces
|
|
of one picture, and neither says what it is a picture of."""
|
|
vision = _Vision(text="The chart shows revenue. " * 300)
|
|
|
|
sections = await ImageParser(vision).parse(PNG, "chart.png")
|
|
chunks = await NaiveChunker(chunk_size=64).chunk(sections)
|
|
|
|
assert len(chunks) == 1
|
|
|
|
|
|
async def test_it_is_not_merged_with_anything_either() -> None:
|
|
"""The other half of standing alone: a query that matches the picture gets
|
|
the picture, not a paragraph that happened to sit beside it."""
|
|
from raven.knowledge._types import Section, TextBlock
|
|
|
|
sections = await ImageParser(_Vision()).parse(PNG, "chart.png")
|
|
sections.append(Section(content=TextBlock(text="Unrelated prose."), source="notes.txt", metadata={}))
|
|
|
|
chunks = await NaiveChunker(chunk_size=4096, image_context_size=0).chunk(sections)
|
|
|
|
assert [chunk.content.text for chunk in chunks] == [
|
|
"A bar chart of revenue by region.",
|
|
"Unrelated prose.",
|
|
]
|
|
|
|
|
|
# -- when it cannot ------------------------------------------------
|
|
|
|
|
|
async def test_no_vision_model_is_a_failure_with_a_way_out(monkeypatch) -> None:
|
|
"""Not an empty section: a document that silently holds nothing is worse
|
|
than a row that says why it is empty."""
|
|
monkeypatch.setattr("raven.knowledge._vision.load_vision_model", lambda: None)
|
|
|
|
with pytest.raises(VisionError) as caught:
|
|
await ImageParser().parse(PNG, "chart.png")
|
|
|
|
assert "no vision model is configured" in str(caught.value)
|
|
assert "Settings" in str(caught.value), "and where to set one"
|
|
|
|
|
|
async def test_a_model_that_answers_nothing_is_a_failure_too() -> None:
|
|
with pytest.raises(VisionError):
|
|
await ImageParser(_Vision(text=" ")).parse(PNG, "chart.png")
|
|
|
|
|
|
async def test_the_endpoint_failure_is_passed_through() -> None:
|
|
"""So the row names the endpoint's own words rather than "could not
|
|
parse"."""
|
|
with pytest.raises(VisionError, match="rate limited"):
|
|
await ImageParser(_Vision(error=VisionError("rate limited"))).parse(PNG, "chart.png")
|