Skip to content

CASSIA v1.4.0

Latest

Choose a tag to compare

@ElliotXie ElliotXie released this 10 Sep 11:04
· 3 commits to main since this release

CASSIA 1.4.0 release notes

CASSIA 1.4.0 makes the command-line interface a supported package entry point
and packages the benchmark-tested agent workflows that were previously kept in
the research workspace.

Highlights

  • cassia annotate supports API providers and local Codex, Claude Code,
    Cursor, OpenCode, or custom-shell agent backends.
  • Three explicit annotation modes: one-shot, validated, and fused-boost.
  • v2-compact is the Fused Boost release default. Historical v2 remains
    available for exact legacy-prompt reproduction.
  • Stable cassia.annotation.v1 result normalization across standard and Fused
    Boost outputs.
  • cassia judge implements the blinded stable-judge-v1.1 protocol with
    rank-1 subtype determining core credit.
  • cassia guide, nested cassia help, doctor, validate, examples,
    report, and resume cover setup, preflight, and run recovery.
  • cassia boost, subcluster, and deterministic consensus workflows are
    available from the installed executable.
  • cassia agent provides an optional long-lived R/Seurat bridge with marker
    queries, cluster operations, labels, QA, finalization, and audit history.

Benchmark evidence

The release snapshot under Benchmark/cli_release/ contains fixed 50-, 100-,
and 392-case manifests and compact result records.

  • Boost-100: one-shot scored 50/100 exact and review Boost scored 71/100 exact
    on an intentionally hard, boost-enriched diagnostic set. There were 26 strict
    rescues and five strict regressions, supporting selective rather than
    unconditional Boost.
  • Full392 prompt ablation: legacy Fused Boost v2 and the compact treatment both
    scored 296/392 exact. Correct discordance was 19 versus 19 (exact McNemar
    p=1.0); estimated annotation cost was 16.3% lower for the compact arm.
  • The 50-case cross-model card is included as a capability profile. It is a
    single-run, fixed-referee comparison and should not be read as a permanent
    leaderboard.

Compatibility and packaging

  • Python 3.9 through 3.13 remain supported.
  • The wheel and source distribution include the CLI guide, agent system prompt,
    and R daemon resource.
  • Existing Python APIs remain available; the CLI is additive.

Release verification

Before upload, the release artifacts must pass:

  1. the full CLI pytest suite;
  2. wheel and source-distribution metadata/content inspection;
  3. installation into a fresh virtual environment;
  4. offline example, validate, one-shot/validated shell-agent, Fused Boost
    dry-run, Judge help, and packaged-resource smoke tests;
  5. one low-cost live OpenRouter annotation smoke test;
  6. one-cluster live annotation smoke tests through Codex CLI, Claude Code,
    Cursor Agent, and OpenCode.