Skip to content

v0.9.38

Choose a tag to compare

View all tags
@github-actions github-actions released this 26 Sep 00:29
· 2348 commits to main since this release
v0.9.38
44bf44a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Maintenance release

Stella includes reliability and maintenance improvements.

Packages

  • @stll/ai-catalog: Add MODEL_OUTPUT_TOKEN_LIMITS and getOutputTokenLimit: the most output tokens one response of each offered model may carry. An id the catalog does not list has no limit.
  • @stll/cli: New list capabilities: lists.verifications.list lists a document's verifications, newest first, with claim counts per verdict state; lists.verifications.latest.list reads the latest verification of up to 200 document files at once.
  • @stll/cli: New list capabilities: lists.items.fact-details.update sets a fact item's date, evidence kind, medium, confidence, interpretation note and scoring hold; lists.verifications.create starts checking a document against a list's facts; lists.verifications.get reads the claims found and their verdicts; lists.verifications.claim-reviews.create and lists.verifications.claim-reviews.bulk.create record reviewer decisions on those claims.
  • @stll/business-registries, @stll/cli, @stll/money, @stll/template-conditions, @stll/ui: Narrow values with type guards instead of type assertions, and give internal CLI helpers names that describe what they hold. mapEntityStatus now reads only the codes its mapping declares, so an inherited object key maps to unknown.

Changes

🚀 Features

  • AVT matter view for list verification (beta) (#3779)
  • api: verify documents against a legal list's facts (#3776)
  • add approved Chrome browser control (#2436)

🐛 Bug Fixes

  • chat: keep approval cards in step with their calls (#3880)
  • chat: keep a denied call denied in the live view (#3875)
  • chat: settle open tool calls when a turn stops (#3866)
  • chat: bound model calls by each model's output limit (#3872)
  • chat: merge split assistant copies in the interrupt snapshot (#3858)
  • api: name the font in dev-seed DOCX packages (#3874)
  • web: open search locations in the file tree (#3859)
  • web: open folder links in the file tree (#3857)

🧪 Tests

  • chat: check rendered cards against recorded conversations (#3879)
  • chat: check the live view against a reload after every send (#3865)

📦 Dependencies

  • deps: bump folio to 0.49 (#3869)

⚙️ Miscellaneous

  • release v0.9.38 (#3881)
  • report failures on best-effort paths (#3878)
  • dev: bind local compose services to loopback (#3876)
  • lint: check database awaits in loops with the type checker (#3867)
  • api: open Postgres-gated test databases through one helper (#3871)
  • case-law: cover the relations the public corpus readers use (#3868)
  • api: pass worker database handles explicitly (#3863)
  • ci: run a model turn in the staging smoke (#3864)
  • observability: grade request failures alongside existing output (#3860)
  • observability: one reader for failure evidence (#3856)
  • api: read org settings, audiences and edit sessions through the request scope (#3862)
  • api: batch independent database and network loops (#3861)
  • share duplicated code through its owners (#3855)
  • clear more small ratchet entries (#3853)
  • search: run search reads through the request scope (#3854)
  • ci: run image checks on pull requests (#3850)
Assets 9
Loading