MCPcopy Create free account

hub / github.com/JuliusBrussee/caveman / functions

Functions12,853 in github.com/JuliusBrussee/caveman

FunctionTestImportedOpenAIUsagePricesOverlappingDetailsOnce
(t *testing.T)
proxy/internal/store/trial_feature_test.go:447
FunctionTestImportedUsageHonorsDistinctReasoningRate
(t *testing.T)
proxy/internal/store/trial_feature_test.go:462
FunctionTestInboundKeyCannotChangeResolvedVertexPrincipal
(t *testing.T)
proxy/providers/vertex/credentials_test.go:11
FunctionTestIndexLockIsSharedAndStaleLocksBreak
(t *testing.T)
proxy/internal/pool/pool_test.go:309
FunctionTestInferredWindowOnCatalogModelIsNotAFallback
(t *testing.T)
proxy/internal/store/learn_v2_p0_test.go:442
FunctionTestInitializeNativePersistenceCorruptCCRForcesRecordPassThrough
(t *testing.T)
proxy/cmd/caveman-proxy/main_test.go:14
FunctionTestInitializeNativePersistenceHealthyCCRRetainsRequestedMode
(t *testing.T)
proxy/cmd/caveman-proxy/main_test.go:30
FunctionTestInitializeNegotiatesSupportedProtocolVersion
(t *testing.T)
mcp/server_test.go:489
FunctionTestInitializeOffersItsOwnVersionToNewerClients
A client asking for a version this adapter does not implement must get the version it DOES implement, not an error. Erroring here dropped the server f
mcp/server_test.go:510
FunctionTestInitializeUsesSuppliedBuildVersion
(t *testing.T)
mcp/server_test.go:466
FunctionTestInjectRetrieveToolAppendsToExistingAnthropicTools
(t *testing.T)
proxy/internal/gateway/retrieve_tool_test.go:52
FunctionTestInjectRetrieveToolStaticOpenAIPreservesMessageBytes
(t *testing.T)
proxy/internal/gateway/retrieve_tool_test.go:31
FunctionTestInjectedContextPayload
The predicate's contract, stated directly: every rule is anchored at the payload start, so a human block that quotes one of these is still mined.
proxy/internal/store/recurring_test.go:84
FunctionTestInsertedUsageCannotSelfPromoteOrPoisonTotals
(t *testing.T)
proxy/internal/store/trial_feature_test.go:422
FunctionTestInspectContentBlindDigestRejectsDrift
(t *testing.T)
shared/platform/cacheguard/cacheguard_test.go:122
FunctionTestInspectFailsClosedOnUnknownBoundaryOrAdapter
(t *testing.T)
shared/platform/cacheguard/cacheguard_test.go:55
FunctionTestInspectFreezesFirstPrefixAndRejectsDrift
(t *testing.T)
shared/platform/cacheguard/cacheguard_test.go:12
FunctionTestInspectObjectMatchesFullDecode
TestInspectObjectMatchesFullDecode: the targeted decode accepts exactly what json.Unmarshal into map[string]any accepts and yields the same metadata.
proxy/providers/inspect_test.go:125
FunctionTestInspectRequestFailsPricingClosedForNonTokenCharges
(t *testing.T)
proxy/providers/adapter_contract_test.go:57
FunctionTestInspectRequest_GuardrailChargeFailsPricingClosed
(t *testing.T)
proxy/providers/bedrock/bedrock_test.go:291
FunctionTestInspectRequest_MantleModelFromBody
(t *testing.T)
proxy/providers/bedrock/bedrock_test.go:267
FunctionTestInspectRequest_MantleRejectsNonAnthropicMessageModel
(t *testing.T)
proxy/providers/bedrock/bedrock_test.go:281
FunctionTestInspectRequest_ModelFromPath
(t *testing.T)
proxy/providers/bedrock/bedrock_test.go:248
FunctionTestInspectRequest_ModelFromPathAndStream
(t *testing.T)
proxy/providers/vertex/vertex_test.go:125
FunctionTestInspectStoreNeverCreatesOrWrites
(t *testing.T)
engine/ccr/store_repair_test.go:203
FunctionTestInspectStoreReportsAFullStore
(t *testing.T)
engine/ccr/store_repair_test.go:229
FunctionTestInstallIDDistinguishesInstallsSharingAHomePath
An install id keyed on the caveman home PATH is the same string on every machine where the user has the same name, so the dashboard adds three hosts i
proxy/internal/cloudlink/link_test.go:1462
FunctionTestInstallIDPrefersTheDeviceIDTheCLIWrote
A deviceId the CLI already persisted stays authoritative: it is what the entitlement device_hash is taken from, so the two must not drift apart.
proxy/internal/cloudlink/link_test.go:1491
FunctionTestInstanceIdentityIsLoopbackOnly
The identity header exists so the local CLI can match a run-state file it can already read. On a shared listener it only hands unauthenticated /health
proxy/cmd/caveman-proxy/status_test.go:263
FunctionTestInstanceIdentityIsPublishedOnlyOnHealth
(t *testing.T)
proxy/cmd/caveman-proxy/status_test.go:49
FunctionTestInstanceIdentityOnAWildcardBindIsForLoopbackCallersOnly
A wildcard bind (CAVEMAN_LISTEN=0.0.0.0:<port>) is still the user's own runtime: the CLI and status reach it through loopback, so a loopback caller ge
proxy/cmd/caveman-proxy/status_test.go:299
FunctionTestInstanceProbeNeverFollowsRedirectOrUsesEnvironmentProxy
(t *testing.T)
proxy/internal/runstate/runstate_test.go:282
FunctionTestInstanceProbeRejectsClosedListener
(t *testing.T)
proxy/internal/runstate/runstate_test.go:311
FunctionTestInstanceProbeRequiresExactLiveTokenWithoutSendingIt
(t *testing.T)
proxy/internal/runstate/runstate_test.go:243
FunctionTestInterruptedDeleteIsRevokedAndResumes
P1: sessions/delete commits the revocation before it purges in batches. A delete that fails midway already answers "deleted" everywhere, and its retry
proxy/internal/middleware/hardening_test.go:637
FunctionTestInterruptedStreamsAbortClientFramingWithoutReplay
(t *testing.T)
proxy/internal/gateway/reliability_test.go:158
FunctionTestInvalidUTF8ToolOutputRoundTripsByteExact
(t *testing.T)
proxy/internal/nativeruntime/runtime_test.go:1275
FunctionTestInvariantHighPrecisionDecimalStatesNoNumericRange
The guard has to hold the line end to end, not just in isolation: a column of high-precision decimals must leave the numeric track entirely rather tha
engine/compressors/invariants_test.go:1123
FunctionTestInvariantNumberRefusesUnorderableLiterals
A float64 may carry a field's min/max only if it can ORDER two literals written to that field's precision. Past 2^53 that fails for integers (snowflak
engine/compressors/invariants_test.go:1073
FunctionTestIsBlockedSeparatesRefusalFromNetworkFailure
IsBlocked separates the guard's policy refusal from a network failure, so a caller can say "refused" instead of "upstream down" (#1440). It has to see
shared/platform/ssrf/ssrf_test.go:885
FunctionTestIsGDScriptRejectsOtherLanguages
(t *testing.T)
engine/compressors/code_gdscript_test.go:192
FunctionTestIsProductionValueFoldsCaseAndWhitespace
(t *testing.T)
shared/platform/runtimeenv/runtimeenv_test.go:5
FunctionTestIsReadOnlyToolCompoundTokensAndQuickJSPrototypeHardening
(t *testing.T)
mcp/codemode/stdio_js_builtin_test.go:974
FunctionTestIsRecoveryToolName
(t *testing.T)
proxy/providers/adapter_contract_test.go:35
FunctionTestIsSensitiveHeader
(t *testing.T)
shared/platform/redact/redact_test.go:216
FunctionTestIsVertexHost
--- Host validation (defense-in-depth) ----------------------------------
proxy/providers/vertex/vertex_test.go:194
FunctionTestJSONCompressorCanonicalElisionNote
(t *testing.T)
engine/compressors/json_test.go:130
FunctionTestJSONCompressorDeterministic
(t *testing.T)
engine/compressors/json_test.go:117
FunctionTestJSONCompressorIdempotent
(t *testing.T)
engine/compressors/json_test.go:52
FunctionTestJSONCompressorKeepsMiddleSignal
(t *testing.T)
engine/compressors/json_test.go:92
FunctionTestJSONCompressorMalformedPassesThrough
(t *testing.T)
engine/compressors/json_test.go:250
FunctionTestJSONCompressorOutputIsValidJSON
(t *testing.T)
engine/compressors/json_test.go:14
FunctionTestJSONCompressorPreservesErrorSubtree
(t *testing.T)
engine/compressors/json_test.go:35
FunctionTestJSONCompressorQueryBias
(t *testing.T)
engine/compressors/json_test.go:219
FunctionTestJSONCompressorShortArrayUnchanged
(t *testing.T)
engine/compressors/json_test.go:257
FunctionTestJSONElisionCollapsesCyclingCategoricalFields
A field cycling through a few values is not an anomaly. Before the rarity rule, element length took three levels, the MAD collapsed onto the two adjac
engine/compressors/json_test.go:297
FunctionTestJSONElisionKeepsFrequentNumericOutlierCluster
A common value on a real numeric field is still an outlier. 60 timeouts at exactly 503 / 30000ms share one value, so a rarity rule must not apply to t
engine/compressors/json_test.go:361
FunctionTestJSONElisionKeepsMinorityStateClusterByLength
A minority state is not a cycling level. 12% of rows in a fixed-width rollback_required state share one length, far from the jittered normal rows, wit
engine/compressors/json_test.go:386
FunctionTestJSONElisionKeepsRealAnomalyInCyclingArray
The rarity rule must not hide a genuine numeric outlier inside a cycling array: one row whose latency dwarfs the rest is still force-kept. bytes_out o
engine/compressors/json_test.go:336
FunctionTestJSONElisionRefusesArrayOfDistinctRecords
The array counterpart of the bibliography: rows that each say something different are a list of distinct entities, and dropping most of them deletes t
engine/compressors/redundancy_test.go:124
FunctionTestJSONElisionStillCollapsesRepeatedRecords
A uniform array of the same event repeated is the shape elision exists for, and it must keep working.
engine/compressors/redundancy_test.go:138
FunctionTestJSONMarkerCarriesElementInvariants
(t *testing.T)
engine/compressors/invariants_test.go:495
FunctionTestJSONSchemaGrader
(t *testing.T)
engine/evals/harness_test.go:102
FunctionTestJSRunnerChainingAndSandboxIsolation
(t *testing.T)
mcp/codemode/stdio_js_builtin_test.go:163
FunctionTestJoinRecoveryHandlesBounded
TestJoinRecoveryHandlesBounded pins the row bound: every turn re-substitutes every earlier block, so the disclosed handle list must dedupe and stay sh
proxy/internal/gateway/compress_test.go:1147
FunctionTestKMSCallTransportReadAndResponseLimits
(t *testing.T)
shared/platform/kms/kms_test.go:375
FunctionTestKeepAliveWarmsWhileAsksCome
Asks start a keep-alive that warms every warmEvery without other contact and stops warmFor after the last ask.
proxy/internal/cloudlink/warm_test.go:164
FunctionTestKeepNumericExtremaIgnoresNonFinite
keepNumericExtrema latches minIndex/maxIndex on the first numeric cell it sees and then only displaces them on a strict comparison. A NaN compares fal
engine/compressors/invariants_test.go:1027
FunctionTestKeepaliveIsAnsweredBeforeTheTokenGate
The keepalive beacon from older CLIs carries no credential and changes nothing, so it is answered in front of the inbound token gate. A 401 here would
proxy/cmd/caveman-proxy/status_test.go:357
FunctionTestKeepaliveMarksActivityWithoutOpeningASession
Keepalive is the CLI's heartbeat for older proxies that still idle-exit. It must move lastActivity without touching session accounting: a beat is acti
proxy/internal/nativeruntime/runtime_test.go:1155
FunctionTestKnownClasses
(t *testing.T)
engine/safety/safety_test.go:9
FunctionTestLargeRetrieveResponseReplaysOriginalAndClearsClaims
(t *testing.T)
proxy/internal/gateway/compress_test.go:963
FunctionTestLargeSessionDeletesUnderTheDefaultDeadline
P1: a session far larger than one transaction can delete in 500ms is deleted by one request under the default limits: delete has retrieve's deadline a
proxy/internal/middleware/hardening_test.go:674
FunctionTestLargeWriteBelowPageCapReportsPageCap
(t *testing.T)
engine/ccr/store_full_test.go:200
FunctionTestLastEffortOutsideTheContractGoesEmpty
A last effort outside the contract's values goes as "".
proxy/internal/cloudlink/route_effort_test.go:315
FunctionTestLateSpliceFailureKeepsSubstitutions
TestLateSpliceFailureKeepsSubstitutions: when assembling this turn's new compression fails, the request must still re-send the replacements earlier tu
proxy/internal/gateway/cache_prefix_invariant_test.go:1568
FunctionTestLateUsageChunkIsKept
A usage chunk that comes well after finish_reason (but within lateUsageGrace) is still read, by every caller of a chat upstream.
proxy/internal/translate/wire_gaps_test.go:974
FunctionTestLearnApplyHonorsRepoFilter
TestLearnApplyHonorsRepoFilter: apply must rebuild the same plan the user saw. A sink that only exists in repo beta cannot be applied under --repo alp
proxy/cmd/caveman-proxy/learn_wire_test.go:239
FunctionTestLearnBehaviorDumbzoneAndSubagent
TestLearnBehaviorDumbzoneAndSubagent proves real-transcript behavioral detection.
proxy/internal/store/trial_feature_test.go:566
FunctionTestLearnCapabilitiesAnswersWithoutAStore
(t *testing.T)
proxy/cmd/caveman-proxy/learn_wire_test.go:351
FunctionTestLearnConfigSinksAndCavemem
TestLearnConfigSinksAndCavemem proves config-tax + CLAUDE.md-weight detection and that reducible sinks become stored cavemem learnings.
proxy/internal/store/trial_feature_test.go:527
FunctionTestLearnExperimentLifecycleEndToEnd
(t *testing.T)
proxy/cmd/caveman-proxy/learn_wire_test.go:96
FunctionTestLearnExperimentStartSkipsFlagValuesForLabel
(t *testing.T)
proxy/cmd/caveman-proxy/learn_wire_test.go:338
FunctionTestLearnExportWritesPrivacySafeDigest
(t *testing.T)
proxy/cmd/caveman-proxy/learn_wire_test.go:173
FunctionTestLearnHTMLReport
TestLearnHTMLReport proves the static offline report shape.
proxy/internal/store/trial_feature_test.go:671
FunctionTestLearnHTMLSavingsSection
TestLearnHTMLSavingsSection proves the report renders the single savings card — saved-so-far and could-have-saved panels — token-only, with neutral at
proxy/internal/store/learn_savings_test.go:123
FunctionTestLearnHonestyAndScoreTransparency
TestLearnHonestyAndScoreTransparency proves Class B softening, score transparency, inferred basis, and no currency on the local path.
proxy/internal/store/trial_feature_test.go:603
FunctionTestLearnIsReadOnly
TestLearnIsReadOnly proves the analyzer never edits user config.
proxy/internal/store/trial_feature_test.go:642
FunctionTestLearnPlanCarriesWrapMeasured
TestLearnPlanCarriesWrapMeasured proves the plan build wires the block in and names its basis in a caveat — the HTML test below hand-constructs struct
proxy/internal/store/learn_savings_test.go:86
FunctionTestLearnPortfolioGroupingConfidenceAndBestNextMoveTieBreak
(t *testing.T)
proxy/internal/store/learn_portfolio_test.go:25
FunctionTestLearnReconcileComparesBilledAgainstMeasured
(t *testing.T)
proxy/cmd/caveman-proxy/learn_wire_test.go:214
FunctionTestLearnRecurringBelowThreshold
TestLearnRecurringBelowThreshold proves fail-closed: a heavy block in one session and a tiny block in three sessions produce no recurring sink.
proxy/internal/store/trial_feature_test.go:827
FunctionTestLearnRecurringContextSink
TestLearnRecurringContextSink proves cross-session re-paste mining emits a recurring_context sink whose fix is cavemem_offload, and that the recurring
proxy/internal/store/trial_feature_test.go:741
FunctionTestLearnRecurringEvidenceNoRawBody
TestLearnRecurringEvidenceNoRawBody proves the analyzer stores fingerprints and locators, never the raw block body (extends the no-raw-prompts rule).
proxy/internal/store/trial_feature_test.go:787
FunctionTestLearnRepoAggregationAndPreDetectionFilter
(t *testing.T)
proxy/internal/store/learn_portfolio_test.go:81
FunctionTestLearnReportListsEveryMemoryHealthFinding
(t *testing.T)
proxy/internal/store/detect_memory_health_test.go:304
FunctionTestLearnReportRendersTrendsSection
(t *testing.T)
proxy/internal/store/learn_trends_test.go:220
FunctionTestLearnRetroCountsEachSegmentOnce
(t *testing.T)
proxy/internal/store/learn_retro_test.go:230
FunctionTestLearnRetroDedupsUsageByMessageID
Claude Code writes one API response as several JSONL lines (one per content block), each repeating the same message.usage. The scan must count the res
proxy/internal/store/learn_retro_test.go:390
FunctionTestLearnRetroEmitsMeasuredBlock
--- tests ------------------------------------------------------------------
proxy/internal/store/learn_retro_test.go:167
← previousnext →9,601–9,700 of 12,853, ranked by callers