diff --git a/.github/knowledge-model-policy.json b/.github/knowledge-model-policy.json index f5029653..db1c8d15 100644 --- a/.github/knowledge-model-policy.json +++ b/.github/knowledge-model-policy.json @@ -28,29 +28,23 @@ "suspended": false, "source": "https://platform.claude.com/docs/en/models/opus-5-5/overview" }, - "gpt-6-astra": { - "aliases": [], - "efforts": ["xhigh", "max"], - "suspended": false, - "source": "https://developers.openai.com/api/docs/models/gpt-6-astra" - }, - "gpt-6.1-sol": { + "claude-fable-5-2": { "aliases": [], "efforts": ["xhigh", "max"], "suspended": false, - "source": "https://developers.openai.com/api/docs/models/gpt-6.1-sol" + "source": "https://platform.claude.com/docs/en/models/fable-5-2/whats-new-fable-5-2" }, - "gpt-6.1-terra": { + "gpt-6.2-astra": { "aliases": [], "efforts": ["xhigh", "max"], "suspended": false, - "source": "https://developers.openai.com/api/docs/models/gpt-6.1-terra" + "source": "https://developers.openai.com/api/docs/models/gpt-6.2-astra" }, - "claude-fable-5-2": { + "gpt-6.2-sol": { "aliases": [], "efforts": ["xhigh", "max"], "suspended": false, - "source": "https://platform.claude.com/docs/en/models/fable-5-2/whats-new-fable-5-2" + "source": "https://developers.openai.com/api/docs/models/gpt-6.2-sol" } }, "approval_owners": ["Dong1017"], diff --git a/docs/knowledge-model-policy.md b/docs/knowledge-model-policy.md index 5fde3075..c91d3ead 100644 --- a/docs/knowledge-model-policy.md +++ b/docs/knowledge-model-policy.md @@ -75,32 +75,42 @@ conflicting declarations and stale coverage fail. PR text is never shell input. | Owner label | Canonical ID | Explicit aliases | Allowed effort | | --- | --- | --- | --- | | Claude Opus 5.5 | `claude-opus-5-5` | `anthropic.claude-opus-5-5` (Bedrock) | `xhigh`, `max` | -| GPT-6-Astra | `gpt-6-astra` | None | `xhigh`, `max` | -| GPT-6.1 Sol | `gpt-6.1-sol` | None | `xhigh`, `max` | -| GPT-6.1 Terra | `gpt-6.1-terra` | None | `xhigh`, `max` | | Claude Fable 5.2 | `claude-fable-5-2` | None | `xhigh`, `max` | +| GPT-6.2 Astra | `gpt-6.2-astra` | None | `xhigh`, `max` | +| GPT-6.2 Sol | `gpt-6.2-sol` | None | `xhigh`, `max` | IDs and effort support were checked on 2026-10-21 against official -[Claude IDs](https://platform.claude.com/docs/en/models/opus-5-5/overview), -[Claude effort](https://platform.claude.com/docs/en/build-with-claude/effort), -[GPT-6 Astra](https://developers.openai.com/api/docs/models/gpt-6-astra) and -[GPT-6.1 Sol](https://developers.openai.com/api/docs/models/gpt-6.1-sol) docs. +[Claude IDs](https://platform.claude.com/docs/en/models/opus-5-5/overview) and +[Claude effort](https://platform.claude.com/docs/en/build-with-claude/effort) docs. Actual client settings still need truthful declaration. OpenAI uses `reasoning.effort`; Claude uses `output_config.effort`. Identity is resolved first, then effort uses explicit set membership. Default, `high`, `ultra` and unknown settings do not qualify. Aliases are never inferred. Suspension also denies aliases. -GPT-6.1 Terra's canonical ID and support for `xhigh` and `max` were checked on -2026-11-01 against its [official model documentation](https://developers.openai.com/api/docs/models/gpt-6.1-terra). -The policy admits only those two efforts for Terra; the model's other supported -efforts do not qualify for knowledge authorship. - Claude Fable 5.2's canonical ID and support for `xhigh` and `max` were checked on 2026-11-14 against its [official model documentation](https://platform.claude.com/docs/en/models/fable-5-2/whats-new-fable-5-2). The policy admits only those two efforts for Fable 5.2; other documented efforts (`low`, `medium`, `high`) do not qualify for knowledge authorship. The drifting `fable` alias is not admitted. +GPT-6.2 Astra's and GPT-6.2 Sol's canonical IDs were checked on 2026-11-18 against +their official model documentation ([GPT-6.2 Astra](https://developers.openai.com/api/docs/models/gpt-6.2-astra), +[GPT-6.2 Sol](https://developers.openai.com/api/docs/models/gpt-6.2-sol)), and their +support for `xhigh` and `max` against the official [model catalog](https://developers.openai.com/api/docs/models). +The policy admits only those two efforts for both models; their other documented +efforts (`low`, `medium`, `high`) do not qualify for knowledge authorship. Neither +model has an admitted alias: `gpt-6.2`, `astra` and `sol` are not admitted. + +Following [#487](https://github.com/InferMatrixLab/InferRSI/issues/487), the +repository no longer admits GPT-6-Astra (`gpt-6-astra`), GPT-6.1 Sol +(`gpt-6.1-sol`) or GPT-6.1 Terra (`gpt-6.1-terra`). This is a repository +admission decision, not a claim about the vendor's support for those models. Their +entries were removed, so a declaration naming them fails like any other unapproved +model. Merged contributions are not re-evaluated. An open PR is assessed against +the current list at its next refresh; if it still declares a removed model, its +affected work must be re-authored with an admitted model and truthfully declared, +or it needs an [explicit exception](#explicit-exceptions-and-policy-review). + ## Explicit exceptions and policy review The approval owner is `@Dong1017`. Their own contributions require a separately diff --git a/tests/test_knowledge_model_policy.py b/tests/test_knowledge_model_policy.py index 2e371697..a846d104 100644 --- a/tests/test_knowledge_model_policy.py +++ b/tests/test_knowledge_model_policy.py @@ -62,7 +62,7 @@ def block(value, tag=DECLARATION): return f"```{tag}\n{json.dumps(value)}\n```" -def attribution(model="gpt-6.1-sol", effort="xhigh"): +def attribution(model="gpt-6.2-sol", effort="xhigh"): return {"tool": "Example agent", "model_id": model, "effort": effort, "role": "author"} @@ -86,8 +86,14 @@ def approval(paths, kind="exception", reason="Bounded exception", login="policy- return {"id": identity, "user": {"login": login}, "body": block(value, APPROVAL)} -@pytest.mark.parametrize("model", ["gpt-6-astra", "gpt-6.1-sol", "gpt-6.1-terra", "claude-opus-5-5", - "anthropic.claude-opus-5-5", "claude-fable-5-2"]) +def test_default_attribution_is_admitted(policy): + # Failure-asserting tests rely on the default author being the compliant one. + assert model_errors([attribution()], policy) == [] + assert assess(pr(), {KNOWLEDGE}, policy)["state"] == "success" + + +@pytest.mark.parametrize("model", ["claude-opus-5-5", "anthropic.claude-opus-5-5", + "claude-fable-5-2", "gpt-6.2-astra", "gpt-6.2-sol"]) @pytest.mark.parametrize("effort", ["xhigh", "max"]) def test_verified_models_and_explicit_aliases_qualify(policy, model, effort): result = assess(pr(declared([attribution(model, effort)])), {KNOWLEDGE}, policy) @@ -95,9 +101,18 @@ def test_verified_models_and_explicit_aliases_qualify(policy, model, effort): assert result["description"] == "Declaration compliance only" +@pytest.mark.parametrize("model", ["gpt-6-astra", "gpt-6.1-sol", "gpt-6.1-terra"]) +@pytest.mark.parametrize("effort", ["xhigh", "max"]) +def test_removed_models_are_rejected_at_every_effort(policy, model, effort): + authors = [attribution(model, effort)] + errors = model_errors(authors, policy) + assert errors == ["Material AI author 1: model is unapproved or suspended"] + assert assess(pr(declared(authors)), {KNOWLEDGE}, policy)["state"] == "failure" + + @pytest.mark.parametrize("effort", ["low", "medium", "high", "ultra", "maximum", "unknown", - "XHIGH", "xhigh "]) -@pytest.mark.parametrize("model", ["gpt-6.1-sol", "gpt-6.1-terra", "claude-fable-5-2"]) + "XHIGH", "xhigh ", "none", "minimal", "MAX", " max"]) +@pytest.mark.parametrize("model", ["claude-fable-5-2", "gpt-6.2-astra", "gpt-6.2-sol"]) def test_effort_is_exact_set_membership(policy, model, effort): assert model_errors([attribution(model=model, effort=effort)], policy) @@ -105,15 +120,19 @@ def test_effort_is_exact_set_membership(policy, model, effort): @pytest.mark.parametrize("model", ["gpt-6-sol", "gpt-6-astra-latest", "GPT-6.1 Sol", "unknown", "claude-opus-5-5-extra", "gpt-6.1-terra-latest", "GPT-6.1 Terra", "fable", "claude-fable-5-2-latest", - "Claude Fable 5.2"]) + "Claude Fable 5.2", "gpt-6.2", "astra", "sol", + "GPT-6.2 Astra", "GPT-6.2 Sol", "GPT-6.2-SOL", "gpt-6.2-Astra", + " gpt-6.2-astra", "gpt-6.2-sol ", "gpt-6.2-astra-latest", + "gpt-6.2-terra"]) def test_unknown_models_and_inferred_aliases_fail(policy, model): assert assess(pr(declared([attribution(model)])), {KNOWLEDGE}, policy)["state"] == "failure" @pytest.mark.parametrize("canonical, identities", [ ("claude-opus-5-5", ("claude-opus-5-5", "anthropic.claude-opus-5-5")), - ("gpt-6.1-terra", ("gpt-6.1-terra",)), ("claude-fable-5-2", ("claude-fable-5-2",)), + ("gpt-6.2-astra", ("gpt-6.2-astra",)), + ("gpt-6.2-sol", ("gpt-6.2-sol",)), ]) def test_suspension_overrides_canonical_and_alias(policy, canonical, identities): policy["models"][canonical]["suspended"] = True @@ -121,6 +140,20 @@ def test_suspension_overrides_canonical_and_alias(policy, canonical, identities) assert model_errors([attribution(model)], policy) +@pytest.mark.parametrize("bad", [attribution("unapproved-model"), attribution("gpt-6.1-sol"), + attribution("sol"), attribution("GPT-6.2-Sol"), + attribution("gpt-6.2-astra", "high")]) +def test_one_noncompliant_author_fails_a_mixed_declaration(policy, bad): + good = attribution("gpt-6.2-astra", "max") + assert model_errors([good], policy) == [] + for authors in ([good, bad], [bad, good]): + assert assess(pr(declared(authors)), {KNOWLEDGE}, policy)["state"] == "failure" + # Only the noncompliant author is reported. + index = authors.index(bad) + 1 + assert [error.split(":")[0] for error in model_errors(authors, policy)] == [ + f"Material AI author {index}"] + + def test_every_material_author_must_qualify(policy): value = declared([attribution(), attribution("unapproved-model")]) assert assess(pr(value), {KNOWLEDGE}, policy)["state"] == "failure"