baikai 0.7.1.0 → 0.7.2.0

raw patch · 12 files changed

+317/−18 lines, 12 filesPVP: major bump suggested

API removals or changes: PVP suggests a major version bump

API changes (from Hackage documentation)

Files

• CHANGELOG.md +61/−0
• baikai.cabal +1/−1
• fetch/FetchModelsCore.hs +18/−1
• src/Baikai/Models/Generated.hs +69/−0
• src/Baikai/Provider/Cli/Internal.hs +35/−3
• src/Baikai/Provider/Registry.hs +17/−5
• src/Baikai/ResponseFormat.hs +41/−0
• test/CatalogSpec.hs +4/−3
• test/CliInternalSpec.hs +48/−1
• test/PricingPolicySpec.hs +9/−2
• test/StrictEvidenceSpec.hs +1/−1
• test/SurfaceSpec.hs +13/−1

@@ -7,6 +7,67 @@  ## [Unreleased] +## [baikai 0.7.2.0] - 2026-09-30++### Added++- Curated GPT-6.1 Sol on OpenAI Responses and Claude Sonnet 5.5 on Anthropic+  Messages (`openai_gpt_6_1_sol`, `anthropic_claude_sonnet_5_5`), with endpoint+  compatibility facts, standard prices, GPT-6.1 Sol's 272K-token context tier,+  and Sonnet 5.5's one-hour cache-write rate. Sonnet 5.5 rejects forced tool+  choice locally and has no fast mode. Both passed live text and function-tool+  acceptance on 2026-09-30 through `/v1/responses` and `/v1/messages`+  ([record](docs/validation/plan-85/2026-09-30-complete.json),+  [plan 85](docs/plans/85-prove-gpt-6-1-sol-and-claude-sonnet-5-5-live-compatibility.md)).+  The focused smoke runner gains the `sol61-*` and `sonnet55-*` cases.+- `Baikai.ResponseFormat.StructuredOutputSupport`+  (`NativeJsonSchema | NoStructuredOutput`), `declaredStructuredOutput :: Api ->+  StructuredOutputSupport`, and a `structuredOutput` field on `ApiProvider`+  (default `NoStructuredOutput` in `apiProviderWith`), so a caller can ask+  whether a transport enforces a schema without calling it. Every built-in+  provider declares `NativeJsonSchema`.++## [baikai-claude 0.7.1.0] - 2026-09-30++Requires `baikai >=0.7.2`.++### Added++- `claude -p` honours a `JsonSchema` response format (IR-11): it receives+  `--json-schema '<schema>'` and the response text is the tool's validated+  `structured_output`. A missing `structured_output` is a `DecodeFailure`; a+  `claude` too old for the flag yields an `InvalidRequest` error with the exit+  code rather than unconstrained text. `JsonObject`, `name` and `strict` are not+  forwarded; requests without a schema render the same argument vector as+  before. Both Claude providers declare `structuredOutput = NativeJsonSchema`.++### Fixed++- Anthropic adaptive `ThinkingHigh` now sends+  `output_config.effort: "high"` instead of omitting the field. Claude Opus 5.5+  defaults to `medium`, so it previously ran `ThinkingHigh` at medium effort;+  other adaptive models default to `high` and behave as before. Evidence for+  these calls records `effortText = "high"` and no longer carries+  `effort_omitted`, so strict evidence mode no longer refuses them.+  `Baikai.Evidence.EffortOmitted` stays exported, and older records still decode.++## [baikai-openai 0.7.1.0] - 2026-09-30++Requires `baikai >=0.7.2`.++### Added++- `codex exec` honours a `JsonSchema` response format (IR-11): the schema is+  written to a temporary file passed as `--output-schema <file>` and deleted+  however the call ends. A `codex` too old for the flag yields an+  `InvalidRequest` error with the exit code rather than unconstrained text.+  `JsonObject`, `name` and `strict` are not forwarded; requests without a schema+  render the same argument vector as before. All three OpenAI providers declare+  `structuredOutput = NativeJsonSchema`.+- `codexCliCommandWith`, which renders the `codex exec`+  vector with a given `--output-schema` file. `codexCliCommand` is unchanged+  and never renders the flag.+ ## [baikai 0.7.1.0] - 2026-09-23  ### Added

@@ -1,6 +1,6 @@ cabal-version: 3.4 name: baikai-version: 0.7.1.0+version: 0.7.2.0 synopsis: Unified Haskell interface for multiple AI providers description:   baikai provides a unified, provider-agnostic Haskell interface for working

@@ -267,6 +267,7 @@   Map.union     ( Map.fromList         [ ("gpt-6-astra", Just (CatalogResponsesCompat astraResponsesFacts)),+          ("gpt-6.1-sol", Just (CatalogResponsesCompat gpt61ResponsesFacts)),           ("gpt-6-sol", Just (CatalogResponsesCompat gpt6ResponsesFacts)),           ("gpt-6-luna", Just (CatalogResponsesCompat gpt6ResponsesFacts))         ]@@ -304,6 +305,11 @@     -- https://developers.openai.com/api/docs/models/gpt-6-sol     -- https://developers.openai.com/api/docs/models/gpt-6-luna     gpt6ResponsesFacts = astraResponsesFacts+    -- 2026-09-29: as Astra, tool calling requires Responses; efforts low+    -- through max (no none or minimal); sampling removed; 30m cache TTL.+    -- https://developers.openai.com/api/docs/models/gpt-6.1-sol+    -- https://developers.openai.com/api/docs/guides/latest-model+    gpt61ResponsesFacts = astraResponsesFacts     -- 2026-09-07: native tools require Responses; only modern 30m cache TTL.     -- https://developers.openai.com/api/docs/guides/latest-model     astraResponsesFacts =@@ -355,6 +361,12 @@       -- This is the finding: the retired prefix table did not know this id       -- and sent it budget_tokens — same source.       ("claude-sonnet-5", adaptiveNoSampling),+      -- 2026-09-29: adaptive thinking on by default; forced tool choice and+      -- nondefault sampling return 400; no fast mode. Baikai never sends+      -- the rejected "disabled" thinking type.+      -- https://platform.claude.com/docs/en/models/sonnet-5-5/whats-new-sonnet-5-5+      -- https://platform.claude.com/docs/en/build-with-claude/fast-mode+      ("claude-sonnet-5-5", adaptiveNoSampling & #supportsForcedToolChoice .~ False),       -- 2026-08-27: as claude-opus-4-6 — budget deprecated but functional,       -- sampling accepted; baikai prefers the non-deprecated shape — same       -- source. Plan 40 left this membership to a live check that never@@ -639,11 +651,14 @@   where     c = m ^. #cost --- | Provider documentation verified 2026-09-23. These rules supplement base+-- | Provider documentation verified 2026-09-23; GPT-6.1 Sol and Sonnet 5.5+-- verified 2026-09-29. These rules supplement base -- models.dev rates, which do not describe the full request billing policy. -- https://developers.openai.com/api/docs/models/gpt-6-astra -- https://developers.openai.com/api/docs/models/gpt-6-sol -- https://developers.openai.com/api/docs/models/gpt-6-luna+-- https://developers.openai.com/api/docs/models/gpt-6.1-sol+-- https://platform.claude.com/docs/en/models/sonnet-5-5/overview -- https://platform.claude.com/docs/en/models/fable-5-1/overview -- https://platform.claude.com/docs/en/models/opus-5-5/overview -- https://platform.claude.com/docs/en/build-with-claude/fast-mode@@ -654,6 +669,8 @@       (("anthropic", "claude-opus-5"), Model.PricingPolicy [] (Just 10)),       (("anthropic", "claude-opus-4-8"), Model.PricingPolicy [] (Just 10)),       (("openai", "gpt-6-astra"), Model.PricingPolicy [Model.InputPriceTier 272000 (Model.ModelCost 20 75 2 25)] Nothing),+      (("anthropic", "claude-sonnet-5-5"), Model.PricingPolicy [] (Just 4)),+      (("openai", "gpt-6.1-sol"), Model.PricingPolicy [Model.InputPriceTier 272000 (Model.ModelCost 4 15 0.2 5)] Nothing),       (("openai", "gpt-6-sol"), Model.PricingPolicy [Model.InputPriceTier 272000 (Model.ModelCost 4 15 0.4 5)] Nothing),       (("openai", "gpt-6-luna"), Model.PricingPolicy [Model.InputPriceTier 272000 (Model.ModelCost 0.2 0.75 0.02 0.25)] Nothing),       (("anthropic", "claude-fable-5-1"), Model.PricingPolicy [] (Just 20))

@@ -509,6 +509,41 @@             }     } +anthropic_claude_sonnet_5_5 :: Model+anthropic_claude_sonnet_5_5 =+  emptyModel+    { modelId = "claude-sonnet-5-5",+      name = "Claude Sonnet 5.5",+      api = AnthropicMessages,+      provider = "anthropic",+      baseUrl = "https://api.anthropic.com",+      reasoning = True,+      input = [InputText, InputImage],+      cost =+        ModelCost+          { inputCost = 2 % 1,+            outputCost = 10 % 1,+            cacheReadCost = 1 % 5,+            cacheWriteCost = 5 % 2+          },+      fastModeCost = Nothing,+      pricingPolicy = Just (PricingPolicy [] (Just (4 % 1))),+      contextWindow = 1000000,+      maxOutputTokens = 128000,+      headers = Map.empty,+      compat =+        CompatAnthropicMessages+          defaultAnthropicMessagesCompat+            { supportsLongCacheRetention = True,+              supportsCacheControlOnTools = True,+              sendSessionAffinityHeaders = False,+              thinkingStyle = AnthropicThinkingAdaptive,+              supportsSamplingParameters = False,+              supportsFastMode = False,+              supportsForcedToolChoice = False+            }+    }+ deepseek_deepseek_chat :: Model deepseek_deepseek_chat =   emptyModel@@ -1009,6 +1044,38 @@       compat = CompatNone     } +openai_gpt_6_1_sol :: Model+openai_gpt_6_1_sol =+  emptyModel+    { modelId = "gpt-6.1-sol",+      name = "GPT-6.1 Sol",+      api = OpenAIResponses,+      provider = "openai",+      baseUrl = "https://api.openai.com",+      reasoning = True,+      input = [InputText, InputImage],+      cost =+        ModelCost+          { inputCost = 2 % 1,+            outputCost = 10 % 1,+            cacheReadCost = 1 % 10,+            cacheWriteCost = 5 % 2+          },+      fastModeCost = Nothing,+      pricingPolicy = Just (PricingPolicy [InputPriceTier 272000 (ModelCost (4 % 1) (15 % 1) (1 % 5) (5 % 1))] Nothing),+      contextWindow = 1050000,+      maxOutputTokens = 128000,+      headers = Map.empty,+      compat =+        CompatOpenAIResponses+          defaultOpenAIResponsesCompat+            { supportedReasoningEfforts = Just [ThinkingLow, ThinkingMedium, ThinkingHigh, ThinkingXHigh, ThinkingMax],+              supportsSamplingParameters = False,+              supportsLongCacheRetention = False,+              supportsPromptCacheOptions = True+            }+    }+ openai_gpt_6_astra :: Model openai_gpt_6_astra =   emptyModel@@ -1270,6 +1337,7 @@     anthropic_claude_sonnet_4_5,     anthropic_claude_sonnet_4_6,     anthropic_claude_sonnet_5,+    anthropic_claude_sonnet_5_5,     deepseek_deepseek_chat,     deepseek_deepseek_reasoner,     openai_gpt_4_1,@@ -1290,6 +1358,7 @@     openai_gpt_5_6_terra,     openai_gpt_5_mini,     openai_gpt_5_nano,+    openai_gpt_6_1_sol,     openai_gpt_6_astra,     openai_gpt_6_luna,     openai_gpt_6_sol,

@@ -22,6 +22,9 @@     ClaudeCliReport (..),     decodeClaudeCliResult, +    -- * Failures a CLI reports about its own arguments+    unsupportedFlagError,+     -- * What baikai knows about the process it launched     ExecutableIdentity (..),     executableIdentity,@@ -41,7 +44,7 @@   ) import Baikai.Context (Context) import Baikai.Cost (Cost (..), providerReportedBasis, zeroCost, zeroCostBreakdown)-import Baikai.Error (BaikaiError, decodeError)+import Baikai.Error (BaikaiError (..), ErrorCategory (..), decodeError, processError) import Baikai.Evidence (EvidenceStrength (..), Observed (..), deriveStrength, usageEnvelope) import Baikai.Message   ( AssistantPayload (..),@@ -447,7 +450,10 @@     reportedModel :: !(Maybe Text),     -- | The token counts and reported cost, when the tool included a     -- usage block.-    usage :: !(Maybe Usage)+    usage :: !(Maybe Usage),+    -- | The validated value of the tool's internal @StructuredOutput@+    -- tool, present when the run was given @--json-schema@.+    structuredOutput :: !(Maybe Value)   }   deriving stock (Eq, Show, Generic) @@ -483,14 +489,40 @@       body <- o .: "result"       failed <- o .: "is_error"       session <- o .:? "session_id"+      structured <- o .:? "structured_output"       pure         ClaudeCliReport           { result = body,             isError = failed,             sessionId = session,             reportedModel = KeyMap.lookup "modelUsage" o >>= soleModelUsageKey,-            usage = claudeUsage o+            usage = claudeUsage o,+            structuredOutput = structured           }++-- | Recognise a CLI's refusal of a flag it does not know.+--+-- Given the tool name, the flag baikai sent, the exit code and the+-- decoded stderr, this returns an 'InvalidRequest' error (keeping the+-- exit code) exactly when stderr is the tool's argument parser rejecting+-- that flag: commander's @unknown option '<flag>'@ (@claude@) or clap's+-- @unexpected argument '<flag>'@ (@codex@). Any other failure, including+-- one about the flag's /value/, is 'Nothing' and stays a+-- 'ProcessFailure'.+unsupportedFlagError :: Text -> String -> Int -> Text -> Maybe BaikaiError+unsupportedFlagError tool flag code stderr+  | any (`Text.isInfixOf` stderr) spellings =+      Just ((processError code msg) {category = InvalidRequest})+  | otherwise = Nothing+  where+    quoted = "'" <> Text.pack flag <> "'"+    spellings = ["unknown option " <> quoted, "unexpected argument " <> quoted]+    msg =+      tool+        <> " does not accept "+        <> Text.pack flag+        <> "; upgrade it or unset Options.responseFormat: "+        <> stderr  -- | The model @claude@ reported as having consumed tokens. --

@@ -15,7 +15,7 @@ -- error-shaped 'Response' in the 'Baikai.Error.ProviderUnavailable' -- category. module Baikai.Provider.Registry-  ( ApiProvider (apiTag, stream, complete, describeThinking, strengthCeiling),+  ( ApiProvider (apiTag, stream, complete, describeThinking, strengthCeiling, structuredOutput),     apiProviderWith,     describeApi,     ProviderRegistry,@@ -51,6 +51,7 @@ import Baikai.Options (Options, emptyOptions) import Baikai.Options qualified as Options import Baikai.Response (Response (..), errorResponse, flattenAssistantBlocks, flattenAssistantText, responseError)+import Baikai.ResponseFormat (StructuredOutputSupport (..)) import Baikai.StopReason (StopReason (..)) import Baikai.Stream.Event (AssistantMessageEvent) import Control.Exception qualified as Exception@@ -111,7 +112,16 @@     -- declare 'Evidence.EvidenceRequestedOnly', and will still fail a     -- strict caller at the terminal — see     -- @docs\/adr\/0014-strict-evidence-means-a-record-exists.md@.-    strengthCeiling :: !Evidence.EvidenceStrength+    strengthCeiling :: !Evidence.EvidenceStrength,+    -- | Whether this provider forwards a 'Baikai.ResponseFormat.JsonSchema'+    -- request to a mechanism the host enforces: a static declaration a+    -- caller can read before dispatch.+    --+    -- Like 'strengthCeiling', declare only what the provider delivers.+    -- The built-in providers take their value from+    -- 'Baikai.ResponseFormat.declaredStructuredOutput'; a caller-supplied+    -- transport that honours schemas sets 'NativeJsonSchema' itself.+    structuredOutput :: !StructuredOutputSupport   }   deriving stock (Generic) @@ -127,8 +137,9 @@ -- 'describeThinking' defaults to reporting that nothing was requested -- and nothing translated, which is honest for a transport with no -- reasoning controls; 'strengthCeiling' defaults to--- 'Evidence.EvidenceRequestedOnly', matching @declaredStrength (Custom _)@.--- Override either by record update.+-- 'Evidence.EvidenceRequestedOnly', matching @declaredStrength (Custom _)@;+-- 'structuredOutput' defaults to 'NoStructuredOutput'. Override any of+-- them by record update. apiProviderWith ::   Api ->   (Model -> Context -> Options -> Stream IO AssistantMessageEvent) ->@@ -140,7 +151,8 @@       stream = producer,       complete = completer,       describeThinking = \_ _ -> Evidence.noThinkingRequested,-      strengthCeiling = Evidence.EvidenceRequestedOnly+      strengthCeiling = Evidence.EvidenceRequestedOnly,+      structuredOutput = NoStructuredOutput     }  -- | How an 'Api' tag reads in a dispatch failure.

@@ -1,3 +1,4 @@+{-# LANGUAGE LambdaCase #-} {-# LANGUAGE OverloadedRecordDot #-}  -- | Provider-agnostic structured-output preference.@@ -10,9 +11,14 @@   ( ResponseFormat (..),     JsonSchemaFormat (name, schema, strict),     jsonSchemaFormat,++    -- * Which transports enforce a schema+    StructuredOutputSupport (..),+    declaredStructuredOutput,   ) where +import Baikai.Api (Api (..)) import Data.Aeson   ( FromJSON (parseJSON),     ToJSON (toJSON),@@ -96,3 +102,38 @@               (jsonSchemaFormat schemaName schemaDoc) {strict = fromMaybe False isStrict}           )       other -> fail ("unknown ResponseFormat tag: " <> show other)++-- | Whether a transport forwards a 'JsonSchema' request to a mechanism+-- the host enforces.+--+-- Answerable without spawning or calling anything: read it from a+-- registered provider (@Baikai.Provider.Registry.structuredOutput@) or+-- from a model's transport tag ('declaredStructuredOutput').+data StructuredOutputSupport+  = -- | 'JsonSchema' reaches the provider and the provider enforces it.+    NativeJsonSchema+  | -- | 'Baikai.Options.responseFormat' is ignored by this transport.+    NoStructuredOutput+  deriving stock (Eq, Ord, Show, Enum, Bounded, Generic)++-- | The structured-output support of each built-in transport.+--+-- 'NativeJsonSchema' covers 'JsonSchema' only. The HTTP providers send+-- the schema in the request body; the subscription CLIs receive it as a+-- flag (@claude -p --json-schema@, @codex exec --output-schema@) and+-- 'JsonObject', 'strict' and 'name' are not forwarded to them. An+-- installed CLI too old to know the flag still rejects it at run time;+-- that call returns an error-shaped response in the+-- 'Baikai.Error.InvalidRequest' category rather than unconstrained text.+--+-- A 'Custom' transport is 'NoStructuredOutput' here because this table+-- cannot know what a caller-registered provider does; that provider's+-- own @structuredOutput@ field is the authoritative answer.+declaredStructuredOutput :: Api -> StructuredOutputSupport+declaredStructuredOutput = \case+  AnthropicMessages -> NativeJsonSchema+  OpenAIChatCompletions -> NativeJsonSchema+  OpenAIResponses -> NativeJsonSchema+  AnthropicMessagesCli -> NativeJsonSchema+  OpenAICompletionsCli -> NativeJsonSchema+  Custom _ -> NoStructuredOutput

@@ -60,7 +60,7 @@             c.supportsSamplingParameters @?= False             c.supportedReasoningEfforts @?= Just [ThinkingLow, ThinkingMedium, ThinkingHigh, ThinkingXHigh, ThinkingMax]           _ -> assertFailure "Astra needs explicit OpenAI endpoint facts",-      testCase "Sol and Luna select Responses with explicit endpoint facts" $+      testCase "Sol, Luna, and GPT-6.1 Sol select Responses with explicit endpoint facts" $         mapM_           ( \mid -> do               [api m | m <- allModels, modelId m == mid] @?= [OpenAIResponses]@@ -72,7 +72,7 @@                   c.supportedReasoningEfforts @?= Just [ThinkingLow, ThinkingMedium, ThinkingHigh, ThinkingXHigh, ThinkingMax]                 _ -> assertFailure "GPT-6 Sol/Luna need explicit OpenAI Responses facts"           )-          ["gpt-6-sol", "gpt-6-luna"],+          ["gpt-6-sol", "gpt-6-luna", "gpt-6.1-sol"],       testCase "regenerating from data/models produces no diff" $         withSystemTempDirectory "baikai-catalog-spec" $ \tmpDir -> do           let regenPath = tmpDir <> "/Generated.hs"@@ -119,7 +119,8 @@     ("claude-opus-5-5", (AnthropicThinkingAdaptive, False, False, True)),     ("claude-sonnet-4-5", (AnthropicThinkingBudget, True, True, False)),     ("claude-sonnet-4-6", (AnthropicThinkingAdaptive, True, True, False)),-    ("claude-sonnet-5", (AnthropicThinkingAdaptive, False, True, False))+    ("claude-sonnet-5", (AnthropicThinkingAdaptive, False, True, False)),+    ("claude-sonnet-5-5", (AnthropicThinkingAdaptive, False, False, False))   ]  assertFacts :: Model -> IO ()

@@ -12,6 +12,7 @@ import Baikai import Baikai.Provider.Cli.Internal import Control.Lens ((&), (.~), (^.))+import Data.Aeson qualified as Aeson import Data.ByteString (ByteString) import Data.ByteString qualified as BS import Data.ByteString.Char8 qualified as BS8@@ -35,6 +36,7 @@     [ promptTests,       codexParserTests,       claudeParserTests,+      unsupportedFlagTests,       executableIdentityTests,       evidenceHelperTests     ]@@ -267,7 +269,52 @@       testCase "malformed stdout is a decode error rather than an exception" $         case decodeClaudeCliResult "not json" of           Left _ -> pure ()-          Right r -> assertFailure ("expected a decode error, got: " <> show r)+          Right r -> assertFailure ("expected a decode error, got: " <> show r),+      -- Shape recorded from Claude Code 2.1.285 run with --json-schema:+      -- the validated value arrives beside a compact copy in "result".+      testCase "a --json-schema run's structured_output decodes" $+        case decodeClaudeCliResult+          "{\"type\":\"result\",\"is_error\":false,\"stop_reason\":\"tool_use\",\+          \\"result\":\"{\\\"items\\\":[]}\",\"structured_output\":{\"items\":[]}}" of+          Left err -> assertFailure ("expected the document to decode: " <> show err)+          Right r -> do+            r ^. #result @?= "{\"items\":[]}"+            r ^. #structuredOutput @?= Just (Aeson.object ["items" Aeson..= ([] :: [Aeson.Value])]),+      testCase "a run without --json-schema has no structured output" $+        case decodeClaudeCliResult "{\"result\":\"pong\",\"is_error\":false}" of+          Left err -> assertFailure ("expected a bare object to decode: " <> show err)+          Right r -> r ^. #structuredOutput @?= Nothing+    ]++-- ============================================================+-- Unsupported-flag classification+-- ============================================================++unsupportedFlagTests :: TestTree+unsupportedFlagTests =+  testGroup+    "unsupportedFlagError"+    [ testCase "commander's unknown option (claude) is InvalidRequest with the exit code" $+        case unsupportedFlagError "claude" "--json-schema" 1 "error: unknown option '--json-schema'\n" of+          Nothing -> assertFailure "expected the rejection to be recognised"+          Just e -> do+            e ^. #category @?= InvalidRequest+            e ^. #exitCode @?= Just 1+            assertBool+              ("message names the tool and flag: " <> show (e ^. #message))+              ("claude does not accept --json-schema" `Text.isPrefixOf` (e ^. #message)),+      testCase "clap's unexpected argument (codex) is InvalidRequest with the exit code" $+        case unsupportedFlagError "codex" "--output-schema" 2 "error: unexpected argument '--output-schema' found\n" of+          Nothing -> assertFailure "expected the rejection to be recognised"+          Just e -> do+            e ^. #category @?= InvalidRequest+            e ^. #exitCode @?= Just 2,+      testCase "an unreadable schema file is not an unsupported flag" $+        unsupportedFlagError "codex" "--output-schema" 1 "Failed to read output schema file /tmp/x.json: No such file"+          @?= Nothing,+      testCase "a rejection naming a different flag is not this flag" $+        unsupportedFlagError "claude" "--json-schema" 1 "error: unknown option '--bogus-flag'"+          @?= Nothing     ]  -- ============================================================

@@ -23,8 +23,8 @@ tests =   testGroup     "Pricing policy"-    [ testCase "Sol and Luna prices switch the complete request above 272K" $-        forM_ [(Models.openai_gpt_6_sol, M.ModelCost 2 10 (1 / 5) (5 / 2), M.ModelCost 4 15 (2 / 5) 5), (Models.openai_gpt_6_luna, M.ModelCost (1 / 10) (1 / 2) (1 / 100) (1 / 8), M.ModelCost (1 / 5) (3 / 4) (1 / 50) (1 / 4))] $ \(m, standard, highRates) -> do+    [ testCase "Sol, Luna, and GPT-6.1 Sol prices switch the complete request above 272K" $+        forM_ [(Models.openai_gpt_6_sol, M.ModelCost 2 10 (1 / 5) (5 / 2), M.ModelCost 4 15 (2 / 5) 5), (Models.openai_gpt_6_luna, M.ModelCost (1 / 10) (1 / 2) (1 / 100) (1 / 8), M.ModelCost (1 / 5) (3 / 4) (1 / 50) (1 / 4)), (Models.openai_gpt_6_1_sol, M.ModelCost 2 10 (1 / 10) (5 / 2), M.ModelCost 4 15 (1 / 5) 5)] $ \(m, standard, highRates) -> do           forM_ [(272000, standard), (272001, highRates)] $ \(n, expected) -> do             let u = U.zeroUsage & #inputTokens .~ (n - 2000) & #cacheReadTokens .~ 1000 & #cacheWriteTokens .~ 1000 & #outputTokens .~ 100             resolveRates Nothing m u @?= Right expected@@ -38,6 +38,13 @@         (computeCostAtSpeed m SpeedFast u).usd @?= 10         (computeCostForService (Just CacheRetentionLong) Nothing m observedFast).usd @?= 16         Set.member (C.UnsupportedSpeed "fast") (computeCostForService (Just CacheRetentionLong) Nothing m observedFast).basis.estimateReasons @?= False,+      testCase "Sonnet 5.5 prices short and long cache writes and has no fast rates" $ do+        let m = Models.anthropic_claude_sonnet_5_5+            u = U.zeroUsage & #cacheWriteTokens .~ 1000000+        (computeCostWith (Just CacheRetentionShort) m u).usd @?= 5 / 2+        (computeCostWith (Just CacheRetentionLong) m u).usd @?= 4+        (computeCost m (U.zeroUsage & #cacheReadTokens .~ 1000000)).usd @?= 1 / 5+        Set.member (C.UnsupportedSpeed "fast") (computeCostAtSpeed m SpeedFast u).basis.estimateReasons @?= True,       testCase "requested tiers never substitute for observed service" $ do         let unknown = N.normalizeUsage N.InclusiveInput (N.ReportedUsage (Just 1000) (Just 0) (Just 0) (Just 0) Nothing)             standard = U.observeBilling [U.BillingServiceTier "default"] unknown

@@ -180,7 +180,7 @@         (EffortCollapsedToToggle ThinkingMax)         "bare on/off toggle",       refusesDowngrade-        "an adaptive high sends no effort field at all"+        "an omitted effort field is indistinguishable from the provider default"         (EffortOmitted ThinkingHigh)         "indistinguishable on the wire",       refusesDowngrade

@@ -64,11 +64,23 @@             logCfg = callLogConfig "/dev/null"         provider ^. #apiTag @?= Custom "probe"         provider ^. #strengthCeiling @?= EvidenceRequestedOnly+        provider ^. #structuredOutput @?= NoStructuredOutput         req ^. #attempt @?= 2         req ^. #runId @?= "r"         tool ^. #name @?= "t"         tool ^. #parameters @?= Aeson.Null         embedding ^. #modelId @?= "e"         logCfg ^. #path @?= "/dev/null"-        logCfg ^. #enabled @?= True+        logCfg ^. #enabled @?= True,+      testCase "declaredStructuredOutput: every built-in transport enforces a schema, Custom does not" $ do+        let builtIns =+              [ AnthropicMessages,+                OpenAIChatCompletions,+                OpenAIResponses,+                AnthropicMessagesCli,+                OpenAICompletionsCli+              ]+        map declaredStructuredOutput builtIns @?= map (const NativeJsonSchema) builtIns+        declaredStructuredOutput (Custom "x") @?= NoStructuredOutput+        [minBound .. maxBound] @?= [NativeJsonSchema, NoStructuredOutput]     ]