buildAnthropicParams converts the shared ChatRequest into Anthropic SDK parameters.

newQuietHandle silences stdout for machine-readable formats (json, sarif) or when audience=="agent"; otherwise the returned handle is a no-op restorer

outputRetryReportText renders the frozen retry report as the terminal summary. It is a run result, not a warning, so it goes to the same writer as the

resolveWorkingDir returns (absPath, isGitRepo, err). When requireGit is true, returns an error if the directory is not a git repo. When false, returns

runCompression performs three-zone memory compression on the given messages, summarizing the compress zone while preserving the active zone intact. Re

ApplyLanguage injects a language directive into all system-role messages across MAIN_TASK, PLAN_TASK (if set), and MEMORY_COMPRESSION_TASK.

CommentsForPath returns a copy of comments whose Path matches the given path.

IsUserIncluded reports whether the given path matches any user include pattern. The check is case-insensitive: both path and pattern are lowercased. R

LoadDefault parses the embedded task_template.json and resolves prompt file references.

LoadScanDefault parses the embedded scan_template.json.

NewDiffMap creates a frozen, read-only DiffMap from a plain map.

PromptTokenLimit returns tokenWarningThreshold (80%) of maxTokens. It is shared by the agent and scan pre-flight gates, their large-input filters, and

RecordAttempt appends one observed HTTP attempt for m. startedAt and endedAt bracket the real HTTP call: the observer takes them immediately before a

RegisterAll registers tools from the MCP client into the tool registry. When allowedTools is non-empty, only tools whose names appear in the list are

RequestMetaFromContext returns the RequestMeta attached to ctx, if any. It is exported so the packages that stamp identity — internal/llmloop and int

ResolveEndpointWithModelOverride is a compatibility wrapper around ResolveEndpointWithOptions for callers that only override the model.

SealSelected closes the selected set. After it returns, RegisterSelected fails and the coverage denominator is fixed; per-item Mark* calls continue to

Tool returns which tool this provider implements.

ValidateScanOptions verifies that the previous session was a full-file scan.

buildResponsesParams converts the shared ChatRequest into Responses API parameters. Mapping notes: - Multiple system messages are concatenated into I

dispatchSubtasks groups items into batches per the configured strategy, then processes batches sequentially while running files within each batch conc

executeReviewFilter runs the REVIEW_FILTER_TASK to remove comments that are provably incorrect based solely on the diff. Errors are logged and silentl

freezeOne freezes c and returns the single listed request.

newFakeLLM returns a server that succeeds on the first attempt for every file. rateLimitOnce and hardFail are set by the caller before use.

retryTestRepo builds a two-file repo with one reviewable commit range (HEAD~1..HEAD), each file carrying its own marker in the second commit only.

sanitizeReason is the single, best-effort redaction+truncation floor applied to every reason the builder stores, so no caller path can write an unreda

startFakeLLM starts srv and points the OCR_LLM_* endpoint resolution at it, with HOME/XDG_CONFIG_HOME redirected so the developer's real config and se

AnyToAttr converts an arbitrary value to an OTel attribute.KeyValue.

BuildReLocationMessages renders the re-location prompt for cm against d. Returns nil when the task template is absent or empty, which the caller treat

DiscoverRepos walks the sessions root and returns one entry per subdirectory.

ErrorEvent emits an error event with error status.

Finalize performs the hard-validated close of the run and returns the immutable manifest. In order it: sweeps any item still selected into failed (col

IsPathExcluded returns true when relPath matches any of the supplied gitignore-style patterns or any default excluded directory prefix (see ExcludedDi

NewProvider creates a Provider for range mode: from..to (via merge-base).

RecordLLMResult sets the outcome of an LLM request on the span.

RecordUsage adds the prompt/completion/cache tokens reported by an LLM response to the runner's aggregate counters. Used by callers (plan phase in age

ResolveIdentity computes the identity a review with these args would record, without creating anything: no session, manifest, runner or LLM call. Resu

RunManifest returns the frozen v1 coverage result for review runs. Scan remains legacy and returns nil.

SetPendingFailureCause records the classification Finalize must use when it sweeps items that never received a Mark*. It records no run_failure and tr

ValidateProtocol accepts the three canonical protocol names and rejects everything else.

captureStderr captures everything written to os.Stderr during fn. Mirrors captureStdout; used to assert structured usage emitted on the failure path.

emitFailureUsage writes a best-effort structured usage record to stderr when a review fails, so the outer caller still sees the cost of the failed att

filterLargeDiffs drops diffs whose diff content alone consumes more than 80% of MaxTokens.

loadLLMRuntime loads tool defs from toolConfigPath, reads the app config from the user's default config path (applying the configured language to tpl

mapResponsesResponse converts the SDK Response into the shared ChatResponse. Text output is read via the SDK's OutputText() helper (it walks all outpu

peekSession reads the first record, all review_item records (for comment count), and the last record of a JSONL file.

recordWarning adds a non-fatal warning to the agent's warning list.

retryCodesMiddleware returns an HTTP middleware that forces the SDK to retry responses whose status code is in the given set, by injecting the x-shoul

End-to-end coverage of the #368 P5 wiring: a real review run against the fake Anthropic server, driven through runReview so review_cmd.go's own Freeze

validateReviewRefs rejects ref-option injection (#112): any --from/--to/ --commit value must be a real commit ref and must not start with '-'.

CanonicalPath returns an absolute path with symlinks resolved.

IsUserExcluded reports whether the given path matches any user exclude pattern. The check is case-insensitive: both path and pattern are lowercased.

LoadDetail returns the summary plus per-file item records for one session.

LoadSummary loads a single session's Summary. Errors when the session file is missing or unreadable.

MarkReused records that the item was reused from a parent session checkpoint.

MarkWaived records that the user explicitly waived this selected item, which must still be in the pre-terminal (selected) state. reason is required an

ParseComments extracts LlmComment entries from tool call arguments without writing to the Collector. Returns parsed comments and an error message (emp

ParseHunks parses raw unified diff text for a single file into a slice of Hunks. Lines before the first @@ header (file-level headers like "diff --git

ReLocateComment calls the LLM to regenerate a precise existing_code snippet when text-based matching fails, then retries ResolveComment with the new s

RecordReviewItemReused records that this run reused a checkpoint from another session.

ReviseLastAttempt rewrites the last attempt of m as an error. It is the client-boundary correction for what the observer cannot see: an error surfaci

SetFinalManifest stores the frozen manifest the agent produced. It is embedded into session_end by Finalize and returned to the CLI via FinalManifest.

StripMarkdownFences removes ```json and ``` wrappers some models add around structured outputs. Exposed so callers (e.g. agent's review-filter post-st