[refactor] Semantic function clustering: consolidate duplicated log parsers, reinvented utils, and redundant engine overrides

[github-actions]

🔧 Semantic Function Clustering Analysis

Analysis of github/gh-aw — read-only semantic + naming-pattern clustering across the pkg/ tree.

Overview

Analyzed 986 non-test .go files (~15.1k func declarations), concentrated in pkg/workflow (423 files) and pkg/cli (337 files). The codebase is generally well-factored — most repeated function names are legitimate interface implementations (RenderMCPConfig, GetSecretValidationStep, ...) that already delegate to shared builders, //go:build js||wasm stub pairs, or thin wrappers that already call the canonical util (sliceutil.SortedKeys, stringutil.Truncate, repoutil.SplitRepoSlug). Those were checked and excluded as false positives.

What remains is a small set of genuine, actionable duplicates and reinvented utilities. Each item below was confirmed by reading both function bodies.

Key findings (by impact)

#	Category	Location(s)	Action
1	Duplicate log parser	gemini_logs.go:22 ↔ antigravity_logs.go:22	Extract one shared stats-JSONL parser
2	Duplicate type	GeminiResponse ↔ AntigravityResponse	Single shared statsJSONResponse
3	Redundant overrides	Get{LogFileForParsing,DefaultDetectionModel} on Gemini/Antigravity	Delete — identical to BaseEngine default
4	Reinvented git helpers	pr_command.go:259	Call existing getCurrentBranch / createAndSwitchBranch
5	Reinvented sliceutil.SortedKeys	logs_report_firewall.go:74, codemod_dependabot_permissions.go:242	Delete, call util
6	Duplicate byte formatters	console/format.go:6 ↔ console/progress_shared.go:6	Collapse to one formatter
7	Reinvented repo-slug split	update_version_labels.go:124, dispatch_workflow_validation.go:176	Route through repoutil.SplitRepoSlug
8	Reinvented stdlib/util	intent/resolver.go:170 cloneStrings; parser/schema_suggestions.go:231 wrappers	slices.Clone / call stringutil.*

Priority 1 — high-confidence, low-risk

1 & 2. Gemini and Antigravity log parsing are near-identical duplicates

• GeminiEngine.ParseLogMetrics — pkg/workflow/gemini_logs.go:22
• AntigravityEngine.ParseLogMetrics — pkg/workflow/antigravity_logs.go:22

Both share the same LogMetrics{} init, the same strings.SplitSeq line loop, the same token extraction from Stats["models"] (input_tokens/output_tokens as float64), and the same Stats["tools"] → toolCallCounts aggregation. Antigravity merely extracted the inner logic into parseAntigravityResponseLine / applyAntigravityResponseMetrics.

Their response structs are byte-for-byte identical:

• GeminiResponse — pkg/workflow/gemini_logs.go:14
• AntigravityResponse — pkg/workflow/antigravity_logs.go:14

type GeminiResponse struct {
    Response string         `json:"response"`
    Stats    map[string]any `json:"stats"`
}

Recommendation: Introduce a shared statsJSONResponse type and a parseStatsJSONLLogMetrics(logContent, ...) helper; have both engines delegate. Also route the hand-rolled for toolName, count := range toolCallCounts { append ToolCallInfo{...} } loops (gemini_logs.go:80, antigravity_logs.go:53) through the existing FinalizeToolMetrics(...) helper that PiEngine.ParseLogMetrics (pkg/workflow/pi_logs.go:100) already uses.

3. Redundant engine overrides identical to the BaseEngine default

Both engines embed BaseEngine and override two methods with the exact value the base already returns:

• GetLogFileForParsing → constants.AgentStdioLogPath
  • pkg/workflow/gemini_logs.go:99, pkg/workflow/antigravity_logs.go:113
  • Base default: pkg/workflow/agentic_engine.go:374
• GetDefaultDetectionModel → ""
  • pkg/workflow/gemini_logs.go:105, pkg/workflow/antigravity_logs.go:119
  • Base default: pkg/workflow/agentic_engine.go:361

Recommendation: Delete all four overrides; inheritance yields identical behavior. (Verify no interface-satisfaction requirement forces the explicit override before removing.)

4. pr_command.go reimplements existing git.go helpers

applyPatchToRepo (pkg/cli/pr_command.go:259) inlines exec.Command("git", "branch", "--show-current") — duplicating getCurrentBranch (pkg/cli/git.go:445) — and at pr_command.go:294 inlines exec.Command("git", "checkout", "-b", branchName) with an identical error wrap — duplicating createAndSwitchBranch (pkg/cli/git.go:465).

Recommendation: Behavior-preserving reuse of the existing getCurrentBranch() / createAndSwitchBranch() helpers.

5. Reinvented sliceutil.SortedKeys

• convertDomainsToSortedSlices (pkg/cli/logs_report_firewall.go:74) manually ranges each map[string]struct{} then slices.Sorts — exactly sliceutil.SortedKeys (pkg/sliceutil/sliceutil.go:63).
• sortedRemainingPermissionKeys (pkg/cli/codemod_dependabot_permissions.go:242) is a pure passthrough: keys := sliceutil.SortedKeys(remaining); return keys.

Recommendation: Replace both with direct sliceutil.SortedKeys(...) calls and delete the wrappers. (Its sibling sortedMissingPermissionKeys does a PermissionScope→string conversion and is justified — keep it.)

Priority 2 — medium-confidence cleanups

6. Two byte-size formatters in pkg/console

• FormatFileSize(int64) — pkg/console/format.go:6 → "1.5 KB" (spaced, up to TB)
• formatBytes(int64) — pkg/console/progress_shared.go:6 → "1.5KB" (compact, up to GB)

Same package, same job, only cosmetic output differences.

Recommendation: Collapse into one formatter (add a compact bool param if the progress-bar format is deliberate) so byte formatting lives in one place.

7. Local owner/repo splitters duplicating repoutil.SplitRepoSlug

repoutil.SplitRepoSlug (pkg/repoutil/repoutil.go:15) is canonical. Local reimplementations:

• splitOwnerRepo — pkg/cli/update_version_labels.go:124 (SplitN/2 + empty checks, bool return)
• parseRepoSlugLiteral — pkg/workflow/dispatch_workflow_validation.go:176 (same split + whitespace rejection)

pkg/cli/engine_secrets.go:696 splitRepoSlug already delegates to repoutil — the pattern to follow.

Recommendation: Route both through repoutil.SplitRepoSlug (adapt error→bool), keeping parseRepoSlugLiteral's whitespace strictness as a thin pre-check.

8. Reinvented stdlib / stringutil

• cloneStrings — pkg/intent/resolver.go:170: hand-rolled make+copy; replace with stdlib slices.Clone (note the nil vs non-nil-empty return difference).
• FindClosestMatches / LevenshteinDistance — pkg/parser/schema_suggestions.go:231,241: one-line wrappers shadowing the identical public stringutil.FindClosestMatches / stringutil.LevenshteinDistance. Call stringutil.* at the sites (keep only if the added debug log is valued).

Also near-duplicate (lower priority)

• ghAPIGet / ghAPIGetArray (pkg/cli/outcome_eval.go:225,248) differ only in unmarshal target (map vs []map). Extract a shared ghAPIGetRaw(endpoint, repo) ([]byte, error) or a generic ghAPIGet[T].
• GenerateCopilotInstallerSteps (pkg/workflow/copilot_installer.go:11) ↔ GenerateAntigravityInstallerSteps (pkg/workflow/antigravity_installer.go:12): same installer-step shape; extract generateCliInstallerStep(...). Parameterize carefully — Copilot has an extra ExpressionPattern security branch + GH_HOST env.
• Outlier: durationStringToSeconds (pkg/workflow/mcp_renderer.go:58) is a generic time converter in an MCP-rendering file, used once — candidate to move to a time util (low priority).

Explicitly ruled out (not duplicates)

False positives verified and excluded

• Repeated engine methods (RenderMCPConfig, GetSecretValidationStep, GetModelEnvVarName, GetExecutionSteps, GetSquidLogsSteps, ...) — interface polymorphism already delegating to shared builders.
• validate* / docker* / getDefaultGHHost "duplicates" — //go:build js||wasm vs !js&&!wasm stub pairs (intentional).
• AddMetrics, Error, Unwrap, UnmarshalJSON, String name collisions — distinct interface impls on different types.
• Already-delegating wrappers (correct design): dedupeAllowedTools→sliceutil.Deduplicate, mergeUniqueLogged→sliceutil.MergeUnique, truncateSHAForLog→stringutil.Truncate, engine_secrets.splitRepoSlug→repoutil.SplitRepoSlug, resolveImportPath/checkCleanWorkingDirectory method→free-func wrappers.
• No stray local min/max/contains/unique/fileExists reimplementations remain in non-test code.
• The ~90 codemod_*.go files are one-codemod-per-file by design; template structural similarity was not flagged.

Suggested next actions

• P1: Merge Gemini/Antigravity ParseLogMetrics + response struct + drop redundant overrides (rejig docs #1–3)
• P1: Reuse getCurrentBranch/createAndSwitchBranch in pr_command.go (Add workflow: githubnext/agentics/weekly-research #4); delete SortedKeys reinventions (Add workflow: githubnext/agentics/weekly-research #5)
• P2: Unify console byte formatters (add cli flag to guard dropping a agentic workflow instructinos file #6); route repo-slug splitters through repoutil (Weekly Research Report: AI Workflow Automation Landscape and Market Opportunities - August 2025 #7); slices.Clone + drop parser fuzzy-match wrappers (Add workflow: githubnext/agentics/weekly-research #8)
• Update/verify tests after each consolidation

Analysis metadata

• Non-test Go files analyzed: 986 · Dominant packages: pkg/workflow (423), pkg/cli (337)
• Detection method: function-signature clustering (duplicate-name + naming-pattern) with per-finding body-diff confirmation
• False-positive filtering: interface impls, build-tag stubs, and existing delegating wrappers excluded
• Analysis date: 2026-07-02

References: §28555945946

Generated by 🔧 Semantic Function Refactoring · 663 AIC · ⌖ 16 AIC · ⊞ 9.2K · ◷

• expires on Jul 3, 2026, 4:19 PM UTC-08:00

[github-actions]

🍪 Issue Monster selected this for Copilot

I've identified this issue as a good candidate for automated resolution and requested assignment to the Copilot coding agent.

If assignment succeeds, the Copilot coding agent will analyze the issue and create a pull request with the fix.

Om nom nom! 🍪

🍪 Om nom nom by Issue Monster · 12.9 AIC · ⌖ 22.3 AIC · ⊞ 8K · ◷