WithinBase reports whether target is base itself or contained under base.

applyDedupGroups parses the DEDUP_TASK output and returns the deduped comment slice. Returns (nil, false) when the response is malformed OR when the g

buildAllowedHosts returns the default-deny allowlist used by the host guard. The set always includes the standard loopback names. If bindHost is a con

buildOpenAIParams converts the shared ChatRequest into OpenAI SDK parameters.

captureStderr swaps os.Stderr for a pipe around fn and returns what was written. Output here is tiny, so reading after the writer is closed avoids any

colorize wraps s in seq and a reset, or returns s untouched when color is off. Every escape sequence in the text output path is gated through here or

ensureMessagesSuffix appends /v1/messages to base URLs that lack a versioned path.

estimateDiffFileTokens projects the input+output token cost of reviewing a single diff (PLAN + MAIN_TASK rounds). It mirrors scan.estimateFileTokens b

executeGroupPlanPhase runs the plan phase for a file group.

expandSessionKeyInBody returns a copy of body with SessionKeyTemplateVar replaced by key in every string value, recursing into nested maps and slices.

expandSessionKeyInHeaders returns a copy of headers with every occurrence of SessionKeyTemplateVar in the values replaced by key. The input map is nev

explainError translates a Bedrock rejection into the action that fixes it. Two of these are actively misleading as the service words them: the API-key

filterLargeScans drops items whose content exceeds 80% of MaxTokens.

finalizeManifest freezes the run's coverage builder and stores the immutable manifest on the session for persistence and CLI consumption. Nil-safe. A

firstLine returns the first non-empty trimmed line of git output, so a stray trailing newline or an unexpected second line never pollutes a resolved S

formatPlanGuidance parses the PLAN_TASK JSON output into a markdown snippet suitable for embedding in MAIN_TASK. On parse failure it returns the raw c

humanTokens formats a token count as e.g. "1.2M" / "850K" / "420".

humanTokens formats a token count as e.g. "1.2M" / "850K" / "420". Mirrors internal/scan/estimate.go; kept here so agent has no scan import.

isUserEditMsg reports whether msg represents user text input (typing or paste).

loadTestTemplate returns a validated default template for runtime tests.

maybeRunPlan invokes PLAN_TASK on the file and returns a human-readable guidance string suitable for {{plan_guidance}} substitution. Returns "(no pre-

mustGetResult extracts the first result from a SARIF document, failing the test if the structure is wrong.

newCompressionRunner returns a Runner whose template forces runCompression to issue a request, plus the message slice to feed it.

newRetryObserver builds the middleware that feeds collector. It sits inside the SDK retry loop, so it runs once per real HTTP attempt and sees neithe

normalizeLine removes leading/trailing whitespace and strips any leading '+' or '-' diff marker.

objcRuleText is the exact text a sniffed .m file must resolve to. Assert on identity rather than on wording: objc.md currently ships as a placeholder

The four blind spots share one shape: the observer sees HTTP 200, the failure surfaces afterwards, and nothing downstream would notice the omission —

partitionMessages divides messages into frozen, compress, and active zones. Frozen zone is always messages[0:2]. Active zone preserves the K most rece

reloadConfigAfterSaveFailure reloads on-disk config and refreshes derived provider TUI state so the UI matches persisted data after a failed save. Ret

reloadConfigAfterSaveFailure reloads on-disk config so in-memory state matches persisted data after a failed save. Returns false when reload could not

resolveBackground determines the effective background from the flag combination. --background-file wins over --background; the commit-message fallback

resolveRepoDir resolves the repo dir for `ocr rules check`. It delegates to resolveWorkingDir(requireGit=true) so it anchors at the git top-level just

resolveWorkingDirForSession accepts an explicit --repo flag value and falls back to the current working directory. Unlike resolveRepoDir it does not r

reviseAttempt applies a client-boundary correction to the last attempt of the logical request identified by ctx. Requests without identity (scan, llm

ruleConfigSHA256 is the deterministic identity of the resolved rule configuration this run operates with: the resolver's effective rule-text layers (c

sarifInvocationFromRun builds the run.invocations entry from the manifest's terminal state and any warnings collected during the review. executionSuc

sarifResultFromComment maps a single LlmComment to a SARIF result. Field mapping follows FR-3 in the requirements spec: - Path → locations[].

setRulesCheckRepo points the rulesCheckCmd's package-level --repo flag at repo for the duration of the test, restoring it afterward. runRulesCheck rea