argus-code-review 0.2.1__tar.gz → 0.2.2__tar.gz

{argus_code_review-0.2.1 → argus_code_review-0.2.2}/.gitignore

@@ -12,3 +12,4 @@ build/
 .env.local
 .env.*.local
 !.env.example
+.argus/

{argus_code_review-0.2.1 → argus_code_review-0.2.2}/CHANGELOG.md

@@ -7,6 +7,22 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.2.2] - 2026-09-11
+
+### Added
+
+- `docs/BENCH.md`, documenting the reviewer bench's config-driven platform,
+  model, and caching selection for leaf-reviewer roles, including the override
+  chain, valid values, platform credentials, and a worked Gemini configuration.
+  README and `docs/BUILDING_A_REVIEW_LOOP.md` now point to the new guide.
+
+### Fixed
+
+- Updated the `gemini-mini` alias from the superseded `gemini-3-flash-preview`
+  model to the stable `gemini-3.8-flash` release. The approved-model policy
+  table's Google default now matches, and a regression test pins the alias to
+  the approved value.
+
 ## [0.2.1] - 2026-09-10
 
 ### Added
@@ -187,7 +203,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   packaged set.
 - `argus --version`, `argus prompts list`, and `argus prompts export`.
 
-[Unreleased]: https://github.com/redesignhealth/argus-review/compare/v0.2.1...HEAD
+[Unreleased]: https://github.com/redesignhealth/argus-review/compare/v0.2.2...HEAD
+[0.2.2]: https://github.com/redesignhealth/argus-review/compare/v0.2.1...v0.2.2
 [0.2.1]: https://github.com/redesignhealth/argus-review/compare/v0.2.0...v0.2.1
 [0.2.0]: https://github.com/redesignhealth/argus-review/compare/v0.1.5...v0.2.0
 [0.1.5]: https://github.com/redesignhealth/argus-review/compare/v0.1.4...v0.1.5

{argus_code_review-0.2.1 → argus_code_review-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: argus-code-review
-Version: 0.2.1
+Version: 0.2.2
 Summary: Self-orchestrated PR review agent using LangGraph + Claude Agent SDK
 Project-URL: Repository, https://github.com/redesignhealth/argus-review
 Project-URL: Issues, https://github.com/redesignhealth/argus-review/issues
@@ -437,6 +437,8 @@ see the [write-up](https://www.redesignhealth.com/content/agentic-code-review-ha
   contract for self-hosters.
 - [`docs/BUILDING_A_REVIEW_LOOP.md`](docs/BUILDING_A_REVIEW_LOOP.md) — how to
   wrap Argus in a review → fix → re-review automation loop gated on CI.
+- [`docs/BENCH.md`](docs/BENCH.md) — configure the static reviewer bench to
+  choose platforms and models for leaf reviewers.
 - [`docs/PRECHECKS.md`](docs/PRECHECKS.md) — the deterministic, non-LLM gate
   that runs before the LLM pipeline: CI-status routing signal and custom
   semgrep rules mined from Argus's own recurring findings.

{argus_code_review-0.2.1 → argus_code_review-0.2.2}/README.md

@@ -385,6 +385,8 @@ see the [write-up](https://www.redesignhealth.com/content/agentic-code-review-ha
   contract for self-hosters.
 - [`docs/BUILDING_A_REVIEW_LOOP.md`](docs/BUILDING_A_REVIEW_LOOP.md) — how to
   wrap Argus in a review → fix → re-review automation loop gated on CI.
+- [`docs/BENCH.md`](docs/BENCH.md) — configure the static reviewer bench to
+  choose platforms and models for leaf reviewers.
 - [`docs/PRECHECKS.md`](docs/PRECHECKS.md) — the deterministic, non-LLM gate
   that runs before the LLM pipeline: CI-status routing signal and custom
   semgrep rules mined from Argus's own recurring findings.

{argus_code_review-0.2.1 → argus_code_review-0.2.2}/argus/llm/models.py

@@ -105,7 +105,12 @@ ALIAS_MAP: Final[dict[str, str]] = {
     # TODO: upgrade to stable gemini-3.1-pro (non-preview) when GA;
     # re-eval by 2026-11-01.
     "gemini-frontier": "gemini-3.1-pro-preview",
-    "gemini-mini": "gemini-3-flash-preview",
+    # gemini-3-flash-preview has been superseded by gemini-3.8-flash, the
+    # stable (non-preview) release -- same reasoning as gemini-frontier's
+    # move off a deprecated preview above: prefer the stable successor over
+    # a preview model once one exists. test_llm_models.py pins this exact
+    # value as a regression guard.
+    "gemini-mini": "gemini-3.8-flash",
 }
 
 # Short-lived pins for evals / preview models. Anything in here is a known

{argus_code_review-0.2.1 → argus_code_review-0.2.2}/argus/prompts/pr-review-specialist-llm-patterns.md

@@ -9,7 +9,7 @@ Only approved model families. Check model strings in code against this table:
 |----------|----------|---------|
 | Anthropic | claude-opus-4, claude-sonnet-4, claude-haiku-4 (use `CLAUDE_MINI`), claude-fable-5 (use `CLAUDE_FRONTIER`), claude-opus-5 (use `CLAUDE_OPUS`), claude-sonnet-4-6 (use `CLAUDE_DEFAULT`) families; claude-sonnet-5 (the previous `CLAUDE_DEFAULT` pin) remains valid as an explicit `ARGUS_SPECIALIST_MODEL`/`--specialist-model` override value, not as a new hardcoded default | claude-sonnet-4-6 |
 | OpenAI | gpt-5.4 family (gpt-5.4, gpt-5.4-mini) and gpt-5.6 family | gpt-5.6-luna |
-| Google | gemini-3 family | gemini-3-flash |
+| Google | gemini-3 family | gemini-3.8-flash |
 
 Flag any use of: o3, o1, gpt-4 family, gpt-5-mini, claude-3/3.5 family, gemini-1.5/2.5 family.
 

{argus_code_review-0.2.1 → argus_code_review-0.2.2}/tests/test_llm_models.py

@@ -14,6 +14,7 @@ from argus.llm.models import (
     CLAUDE_DEFAULT,
     EXPERIMENTAL_MODELS,
     GEMINI_FRONTIER,
+    GEMINI_MINI,
     GPT_FRONTIER,
     GPT_MINI,
     estimate_cost_usd,
@@ -163,6 +164,20 @@ class TestEstimateCostUsd:
         cost = estimate_cost_usd(GPT_MINI, input_tokens=1000, output_tokens=500)
         assert cost > 0.0
 
+    def test_gemini_mini_pinned_to_approved_model(self) -> None:
+        """Regression guard for gemini-mini, alongside gpt-frontier and
+        gpt-mini's guards above: gemini-3-flash-preview has been superseded by
+        gemini-3.8-flash, the stable (non-preview) release on the approved model
+        list (see pr-review-specialist-llm-patterns.md). Pin the alias to that
+        approved value so an accidental regression to a preview model or
+        unapproved string is caught here instead of at review time.
+
+        Also exercises estimate_cost_usd(GEMINI_MINI, ...) directly."""
+        assert ALIAS_MAP["gemini-mini"] == "gemini-3.8-flash"
+        assert GEMINI_MINI == "gemini-3.8-flash"
+        cost = estimate_cost_usd(GEMINI_MINI, input_tokens=1000, output_tokens=500)
+        assert cost > 0.0
+
 
 class TestResolveOverrides:
     def test_resolve_claude_default_honors_env_override(