v0.1.0-rc6

[WaylandYang]

Two weeks in which extraction stopped writing into a schema and started writing what documents say, and alignment learned to turn that back into typed facts. The ledger, the interface and the reading of files all moved with it.

A document is read in its own words

Extraction writes an open graph — the subject, the relation phrase as written, the object, the quote it came from — and nothing is bent to fit a class on the way in (0044, #710, #731, #736).

• The quote is the first thing a statement carries, so the model copies the sentence before it writes from it (The quote comes first #748)
• A statement reads on its own, and a thing is named by a proper name or a term that means the same in any document (A statement reads on its own and a thing is named by a term #743)
• A statement says its mood: a plan, a forecast and a requirement are not facts about the world, and never become typed facts (A statement says its mood #745, The design says what a mood is #746)
• A table keeps its headings, its sections and its caption, in HTML, in Word, in a spreadsheet and in a CSV (A table keeps its headings, its sections and its caption #744, A table is a table in every format #750)
• A memory waits in its own words rather than being written straight into the graph (A memory waits in its own words #735)

The ontology is a view over that, not a gate in front of it

• A kind word binds to a class and a phrase binds to a property, each by two votes with the candidates reversed, and a person's decision is never overwritten (A kind word binds to a class #741, A phrase binds to a property #751)
• A bound statement becomes a typed fact through the same gate as any other fact, with its sources recorded; several statements about one triple make one row, not one each (A bound statement becomes a typed fact #752, One typed fact for many statements #753)
• What the aligner could not decide has a queue, beside the other review queues (The aligner's open questions have a queue #754)
• Editing a class or property definition reconsiders the negative bindings it may now fit, not only the new ones (Reconsider negative bindings when existing definitions change #775, thanks @Maya-Kid)
• The ontology table can create classes and properties, not only edit them (Give the ontology table view a way to create classes and properties #756, thanks @siu91)

Time

• A document is dated from its own text, every time mention is interpreted by the model and computed by code, and upload time is never a document's date (0045, 0045 resolves a time mention against its document #724, A time mention is resolved against its document #740)
• A timeline closes on how a start was got, not on a number. The resolution grade travels from the mention to the statement to the typed row; a successor whose start could not be anchored closes nothing and opens no conflict, and AUTO_CLOSE_MIN_CONFIDENCE is gone from the engine (A timeline closes on how a date was got #762)

A file is read by something that says how it read it

• A chunk carries its origin — stated, OCR'd, transcribed, described — the model that produced it, and an anchor back into the original bytes (A chunk says where its words came from #719, A reader has a card and an origin has a label #722)
• Scans and images go to a layout service; a recording is read with its speakers, and an unlabelled transcript degrades rather than lying (A scan is read by a layout service #720, A recording is read with its speakers #721)
• A PDF whose text layer uses a CJK CMap no longer panics the reader (Fix GBK-encoded PDF text extraction #761, thanks @Floating-Y)
• A WebDAV file whose name contains an XML entity is no longer silently skipped (Keep XML references from truncating WebDAV file paths #765, thanks @Maya-Kid)

What a call to the endpoint survives

• 408, 502, 503, 504 and a request that never sent are transient and retried; 500 is not, and neither is a read timeout (What a long document found #757)
• Extraction streams, so the timeout measures silence rather than thinking. A stream that stops mid-answer is a failure, not a short answer (What a long document found #757), and the two live-stream paths now check the same thing (Reject model streams that end before completing their turn #785, thanks @Maya-Kid)
• A multi-byte character split across two network chunks survives (Preserve Unicode when streamed model replies split a character #764, thanks @Maya-Kid)
• The token ceiling is ours rather than the endpoint's default, and a reply cut at it says so in the drop table (The token ceiling is ours, and a cut reply says so #786, thanks @basil-k-aji-dev)
• Embedding responses that carry an index are paired by it, not by arrival order (Keep indexed embedding responses paired with their input texts #784, thanks @Maya-Kid)
• A passage that could not be extracted is a row in the drop table, not silence (What a long document found #757)

Measured, and left alone

Capping the model's reasoning halves the thinking and costs a sixth of a filing's table figures, 885 statements against 729, and nearly double the misworded rate. The lever is documented and not pulled (#759). Two runs of the same four filings under one configuration differ by about 7% of their statements and half a point of misworded, which is the band any claim has to clear (#769).

Interface

• A follow-up answer keeps streaming when the previous connection finishes closing (Keep late chat callbacks from finishing a follow-up answer #763, thanks @Maya-Kid)
• A statement whose object is a value shows the value; the extraction rail beside a document widens with the screen and separates its items (An extracted value has an object #772)
• A calendar date that does not exist is refused even when a clock time is supplied (Reject impossible calendar dates before converting clock inputs #766, thanks @Maya-Kid)
• Removing a name asks first (What a long document found #757), the stop button keeps a stop-sized square (The stop button keeps a stop-sized square #777), the panel's tab strip starts where its rows do (The tab strip starts where the rows do #771)
• Closing a fact clears it from the unconfirmed queue (Closing a fact clears it from the unconfirmed queue #732, thanks @siu91)

Reads through MCP

Schema reads follow result chunks and are not truncated by the row cap (#737, thanks @rollroyces). Rule and document read failures are tool errors rather than empty successes, and the retrieval limit is applied after visibility filtering (#727, #728, #733, #734, thanks @Maya-Kid).

---

Thank you @Maya-Kid, @siu91, @rollroyces, @Floating-Y and @basil-k-aji-dev.