v0.1.3

SparseRead v0.1.3 — Codex/Pi large-workspace fix

This patch release preserves the v0.1.2 Codex and Pi integrations and fixes a
timeout observed in a large local project. After an agent switched a long file
to full_fidelity, the shared host bridge still recursively inspected its
parent collection even though the request had already selected native reading.
The bridge now skips that unnecessary parent check for this explicit native
veto. Reader algorithms, BenefitGate thresholds, and evidence format are unchanged.

Download and install

Download sparseread-source-installer-v0.1.3.zip or
sparseread-source-installer-v0.1.3.tar.gz from this release, verify the
attached SHA256SUMS, and extract the archive. From the extracted directory:

uv run --python 3.12 python scripts/install_sparseread.py \
  --platform codex --workspace /absolute/path/to/project --doctor

uv run --python 3.12 python scripts/install_sparseread.py \
  --platform pi --workspace /absolute/path/to/project --doctor

Requires Python 3.11+, uv, Node.js 22+, and the corresponding agent CLI. The
source installer fetches dependencies; it is not an offline binary. Restart or
reload the agent in the target project and review its plugin/extension trust
prompts. Codex hook trust is separate from plugin installation; the installer
does not approve hooks or Pi project extensions for you.

Verification and limits

• The full Python suite and a Codex/Pi regression for the native-veto path
  passed before release. In the affected local project, the previously timed-out
  full-fidelity decision returned native in about 3 ms after reinstall.
• Actual Codex app-server loading exposed all six MCP tools. Deterministic
  replay on two local Markdown records and one JSONL trace passed targeted
  evidence, exact raw verification, and native fallback checks. A narrow
  refine was needed to recover one exact protocol line missed by initial focus.
• The user confirmed the plugin appeared in Codex App, approved its hooks, and
  observed actual sro_preview and sro_read calls in a new App chat. App hook
  interception itself was tested through the host hook contract, not an
  automated GUI run. This is not an autonomous model-quality or token benchmark.

No private record bodies, App transcripts, or credentials are included. PyPI,
npm, and framework marketplace publishing are not part of this release; the
source-installer bundle is the supported end-user download. The original
v0.1.2 release
and its assets remain unchanged.

Installation and recovery ·
Historical v0.1.2 record replay