MCPcopy Create free account

hub / github.com/JuliusBrussee/caveman / functions

Functions12,853 in github.com/JuliusBrussee/caveman

↓ 1 callersFunctionyamlChildIndent
(lines: string[], start: number, end: number)
packages/cli/src/index.ts:14801
↓ 1 callersFunctionyamlMappingChildren
(lines: string[], start: number, end: number)
packages/cli/src/index.ts:14668
↓ 1 callersFunctionyamlParentLine
(line []byte)
engine/compressors/config.go:202
MethodAct
(_ context.Context, req ActionRequest, target Target)
browse/session_test.go:40
MethodAct
(context.Context, browse.ActionRequest, browse.Target)
browse/cmd/caveman-browse/main.go:506
MethodAct
(_ context.Context, _ browse.ActionRequest, _ browse.Target)
mcp/codemode/stdio_js_builtin_test.go:794
MethodAfterFunc
(d time.Duration, f func())
proxy/internal/gateway/cache_warm.go:84
MethodAfterFunc
(d time.Duration, f func())
proxy/internal/gateway/cache_warm_test.go:50
FunctionApply
Apply executes a plan. For ActionResize it returns the downscaled image (PNG, deterministic). For ActionDetailLow it returns the original bytes unchan
engine/image/image.go:157
MethodApplyProviderNativeTransforms
(ctx context.Context, body BodyReader, meta RequestMetadata, transformPolicy TransformPolicy)
proxy/providers/adapter.go:816
MethodApplyProviderNativeTransforms
ApplyProviderNativeTransforms is an unconditional byte-safe pass-through. Why it cannot transform: legacy generateContent explicit caching requires a
proxy/providers/gemini/explicit_cache.go:33
MethodApplyProviderNativeTransforms
ApplyProviderNativeTransforms adds one Bedrock-native cache marker to the largest stable Anthropic Claude prefix on Bedrock Runtime. Caller-managed ma
proxy/providers/bedrock/cache_points.go:42
MethodApplyProviderNativeTransforms
ApplyProviderNativeTransforms applies at most one enabled Anthropic caching strategy. OptimizerID places an explicit breakpoint on the last non-deferr
proxy/providers/anthropic/cache_breakpoints.go:26
MethodApplyProviderNativeTransforms
(ctx context.Context, body providers.BodyReader, meta providers.RequestMetadata, policy providers.TransformPol
proxy/providers/openaicompat/openaicompat.go:168
MethodApplyProviderNativeTransforms
(ctx context.Context, body providers.BodyReader, meta providers.RequestMetadata, policy providers.TransformPol
proxy/providers/openaicompat/openaicompat.go:199
MethodApplyProviderNativeTransforms
ApplyProviderNativeTransforms injects a top-level prompt_cache_key derived from the request's stable prefix (model + tools + leading system/instructio
proxy/providers/openai/cache_key.go:42
MethodApplyProviderNativeTransforms
(ctx context.Context, body providers.BodyReader, meta providers.RequestMetadata, policy providers.TransformPol
proxy/providers/azureopenai/azureopenai.go:35
MethodApplyProviderNativeTransforms
(ctx context.Context, body providers.BodyReader, meta providers.RequestMetadata, policy providers.TransformPol
proxy/providers/vertex/content_compress.go:91
MethodApplyProviderNativeTransforms
(ctx context.Context, body providers.BodyReader, meta providers.RequestMetadata, policy providers.TransformPol
proxy/internal/gateway/server_test.go:25
MethodApplyProviderNativeTransforms
(context.Context, providers.BodyReader, providers.RequestMetadata, providers.TransformPolicy)
proxy/internal/gateway/server_test.go:31
MethodAsk
Ask starts the decision for one request and returns the wait for it. The wait never outlasts routeBudget from the ask.
proxy/internal/cloudlink/link.go:540
MethodAsk
(_ context.Context, ask RouteAsk)
proxy/internal/gateway/route_test.go:75
MethodAuthenticate
(context.Context, *http.Request)
proxy/internal/cloudlink/e2e_test.go:22
MethodAuthenticate
(ctx context.Context, r *http.Request)
proxy/internal/standalone/standalone.go:72
MethodAuthenticate
(context.Context, *http.Request)
proxy/internal/store/prefix_stability_e2e_test.go:50
MethodAuthenticate
(ctx context.Context, r *http.Request)
proxy/internal/gateway/server_test.go:42
MethodAuthenticate
(context.Context, *http.Request)
proxy/internal/gateway/middleware_ops_test.go:70
MethodAuthenticate
(context.Context, *http.Request)
proxy/internal/gateway/models_metadata_route_test.go:39
MethodAuthenticate
(_ context.Context, _ *http.Request)
proxy/internal/gateway/dedup_test.go:39
MethodAutoOffered
AutoOffered reports whether Auto is in the agents' pickers: signed in with the routing module on.
proxy/internal/cloudlink/link.go:576
MethodAutoOffered
()
proxy/internal/gateway/chatgpt_auto_test.go:29
FunctionBenchmarkAskParts5MB
The link's share of the per-request work on a 5 MB body (it runs while compression does, inside the 800 ms budget): parse, ask text, declared request.
proxy/internal/cloudlink/route_effort_test.go:253
FunctionBenchmarkDeepSchema
BenchmarkDeepSchema: a schema nested 3000 levels (a tool's may come from any MCP server) costs linear time, under 5 ms a request, in both walks.
proxy/internal/translate/bench_test.go:181
FunctionBenchmarkElisionGateOnLargeDistinctPayload
The gate runs on the proxy hot path, so its cost has to stay bounded on the worst realistic input: a large payload with no repetition at all, where ev
engine/compressors/redundancy_test.go:156
FunctionBenchmarkExtractCompressible
(b *testing.B)
proxy/providers/gemini/content_compress_test.go:110
FunctionBenchmarkExtractCompressibleChat
(b *testing.B)
proxy/providers/openai/content_compress_test.go:268
FunctionBenchmarkExtractCompressibleResponses
(b *testing.B)
proxy/providers/openai/content_compress_test.go:292
FunctionBenchmarkExtractStabilizable
(b *testing.B)
proxy/providers/gemini/content_compress_test.go:122
FunctionBenchmarkExtractStabilizableChat
(b *testing.B)
proxy/providers/openai/content_compress_test.go:280
FunctionBenchmarkInspectRequest
BenchmarkInspectRequest is the request metadata read every request starts with.
proxy/internal/gateway/hotpath_bench_test.go:217
FunctionBenchmarkMiddlewarePreparation
Profile the same 100 KiB log body as the local HTTP benchmark. These component timings diagnose work; only performance.mjs measures queueing and round
proxy/internal/middleware/benchmark_test.go:19
FunctionBenchmarkNative
BenchmarkNative is the harness path's hygiene on a body with nothing to strip (every request to Anthropic's, OpenAI's or a chat API): it must cost a s
proxy/internal/translate/bench_test.go:156
FunctionBenchmarkParseUsageStream
BenchmarkParseUsageStream is the usage read of a finished Messages stream.
proxy/internal/gateway/hotpath_bench_test.go:235
FunctionBenchmarkPayloadAdversarial4MiB
BenchmarkPayloadAdversarial4MiB owns the throughput regression signal. Run it after any pattern change and compare against a saved baseline with bench
shared/platform/redact/payload_test.go:400
FunctionBenchmarkPayloadTypicalRequest
BenchmarkPayloadTypicalRequest is the shape that actually runs in production: a JSON chat request with one email in it.
shared/platform/redact/payload_test.go:414
FunctionBenchmarkProviderPrefixEvidence
BenchmarkProviderPrefixEvidence is the frozen-prefix digest every request to an adapter with prefix evidence pays.
proxy/internal/gateway/hotpath_bench_test.go:198
FunctionBenchmarkProxyHandler
BenchmarkProxyHandler is one record-mode Messages request through the whole handler against an in-process upstream, wired as the binary ships (rows fi
proxy/internal/gateway/hotpath_bench_test.go:132
FunctionBenchmarkRecord
BenchmarkRecord is one request row into a file-backed store, one caller (serial) or eight at once (parallel, run with -cpu 8).
proxy/internal/store/record_batch_test.go:28
FunctionBenchmarkRecordBatch
BenchmarkRecordBatch is the async writer's insert: rows a few at a time, one transaction each, reported per row.
proxy/internal/store/record_batch_test.go:56
FunctionBenchmarkRequestAccounting
BenchmarkRequestAccounting is the request token count record() runs after the response: a pass-through request, so both sides are the same bytes.
proxy/internal/gateway/hotpath_bench_test.go:177
FunctionBenchmarkRouteEffort5MB
The route stage's per-request work on a 5 MB Claude Code-shaped body (messages first, output_config last) whose session replays two marks.
proxy/internal/gateway/route_effort_test.go:537
FunctionBenchmarkServeRecordMode
BenchmarkServeRecordMode is the shipped wiring end to end: record mode, real SQLite spend store, in-process upstream. c=1 is one caller; c=8 is eight
proxy/internal/standalone/hotpath_bench_test.go:51
FunctionBenchmarkTransformCOGS
(b *testing.B)
proxy/providers/anthropic/cogs_bench_test.go:85
FunctionBenchmarkTranslate
BenchmarkTranslate is Request plus Serve per direction and body size; the ms/100k-tok metric is the time scaled to a 100k-token body.
proxy/internal/translate/bench_test.go:122
MethodBuildLearnSimulation
BuildLearnSimulation evaluates only content observed in the selected scan window. It never invokes retro, projects future turns, or prices tokens.
proxy/internal/store/learn_simulate.go:34
MethodCachedPrefixComponents
CachedPrefixComponents splits a direct Gemini generateContent request into `[systemInstruction, tools, content_0, ...]`, aligning with Vertex Gemini a
proxy/providers/gemini/gemini.go:117
MethodCachedPrefixComponents
CachedPrefixComponents splits a Bedrock Claude request into the components Bedrock's prompt cache keys on (system, tools/toolConfig, then each message
proxy/providers/bedrock/cache_points.go:512
MethodCachedPrefixComponents
CachedPrefixComponents is the adapter form of the package function; a token count is never cached, so it has no cached prefix.
proxy/providers/anthropic/content_compress.go:107
MethodCachedPrefixComponents
(body []byte, meta providers.RequestMetadata)
proxy/providers/openaicompat/openaicompat.go:176
MethodCachedPrefixComponents
CachedPrefixComponents mirrors Anthropic's marker-stripped prefix view on Anthropic-dialect compat mounts (so rolling frontier breakpoints never read
proxy/providers/openaicompat/openaicompat.go:225
MethodCachedPrefixComponents
(body []byte, meta providers.RequestMetadata)
proxy/providers/azureopenai/azureopenai.go:43
MethodCachedPrefixComponents
CachedPrefixComponents splits a Vertex request into the components its upstream prompt cache keys on. Both Anthropic Claude and Google Gemini families
proxy/providers/vertex/content_compress.go:80
MethodCallTool
(context.Context, string, map[string]any)
mcp/codemode/stdio_js_builtin_test.go:944
MethodCallTool
(ctx context.Context, name string, args map[string]any)
mcp/codemode/http_upstream.go:198
MethodCallTool
(ctx context.Context, name string, args map[string]any)
mcp/codemode/http_upstream.go:281
MethodCallTool
CallTool invokes tools/call on the downstream MCP server and normalizes the MCP content response into structured JSON (when the text payload is JSON)
mcp/codemode/stdio_upstream.go:221
MethodCallTool
(ctx context.Context, name string, args map[string]any)
mcp/codemode/codemode.go:180
MethodChoices
(scope string, ids []string)
proxy/internal/store/middleware_postgres.go:435
MethodChoices
Choices looks each id up in turn: SQLite runs in process, so a lookup costs no round trip.
proxy/internal/store/middleware.go:639
MethodClose
Close is a no-op.
engine/ccr/store_wasm.go:46
MethodClose
Close releases a valid connection. A quarantined connection is deliberately retained until process exit: SQLite close could replay its obsolete journa
engine/ccr/store_generation.go:187
MethodClose
()
proxy/internal/store/middleware_postgres.go:62
MethodClose
()
proxy/internal/gateway/compress_test.go:99
MethodClose
()
proxy/cmd/caveman-proxy/main.go:1330
MethodClose
()
browse/session_test.go:56
MethodClose
()
browse/cmd/caveman-browse/main.go:510
MethodClose
()
mcp/codemode/stdio_js_builtin_test.go:800
MethodClose
()
mcp/codemode/http_upstream.go:205
MethodClose
()
mcp/codemode/http_upstream.go:301
MethodClose
()
shared/platform/kms/kms_test.go:373
MethodComplete
(_ context.Context, req ModelRequest)
engine/evals/quality.go:62
MethodComplete
(_ context.Context, req ModelRequest)
engine/evals/quality.go:337
MethodComplete
(context.Context, evals.ModelRequest)
engine/evals/harness_test.go:321
MethodCompress
(in []byte)
engine/simulate_test.go:18
MethodCompress
(input []byte)
engine/compressors/axtree.go:35
MethodCompress
(input []byte)
engine/compressors/tabular.go:72
MethodCompress
(input []byte)
engine/compressors/html.go:43
MethodCompress
(input []byte)
engine/compressors/repetition.go:43
MethodCompress
(input []byte)
engine/compressors/diff.go:29
MethodCompress
(input []byte)
engine/compressors/text.go:39
MethodCompress
(input []byte)
engine/compressors/json_strategy.go:30
MethodCompress
(input []byte)
engine/compressors/toolschema.go:91
MethodCompress
(input []byte)
engine/compressors/searchresult.go:37
MethodCompress
(input []byte)
engine/compressors/code_nocgo.go:33
MethodCompress
(input []byte)
engine/compressors/code_cgo.go:68
MethodCompress
(input []byte)
engine/compressors/testreport.go:26
MethodCompress
(input []byte)
engine/compressors/log.go:81
MethodCompress
(input []byte)
engine/compressors/config.go:46
MethodCompress
(input []byte)
engine/compressors/terminal.go:58
MethodCompress
(input []byte)
engine/compressors/toolschema_annotations.go:419
← previousnext →8,201–8,300 of 12,853, ranked by callers