scip-query 0.2.0 → 0.2.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "scip-query",
-  "version": "0.2.0",
+  "version": "0.2.1",
   "description": "Language-agnostic code intelligence CLI powered by SCIP indexes",
   "type": "module",
   "main": "dist/index.js",
@@ -8,6 +8,11 @@
   "bin": {
     "scip-query": "dist/cli.js"
   },
+  "files": [
+    "dist/**/*.js",
+    "dist/**/*.d.ts",
+    "skills/**/SKILL.md"
+  ],
   "sideEffects": false,
   "exports": {
     ".": {

package/IMPROVEMENTS.md

@@ -1,143 +0,0 @@
-# Improvement Opportunities
-
-Self-audit of the scip-query codebase using its own analysis tools. Each finding includes the command that surfaced it, what it means, and what the fix looks like.
-
----
-
-## 1. Replace 26 inline `symbolNoise` filters with `db.symbolNoise`
-
-**Found by:** `grep` on `similar` output (78-80% callee overlap driven by shared SQL boilerplate)
-
-**Problem:** The fragment `AND gs.symbol NOT LIKE '%typeLiteral%'` appears 26 times across 16 query files. The companion `AND gs.symbol NOT LIKE '%().(%'` appears in most of those too. `db.symbolNoise` was added as a reusable getter but never wired in.
-
-**Files affected:**
-- `bottlenecks.ts`, `by-kind.ts`, `call-graph.ts` (3 occurrences), `dead.ts`, `doc-coverage.ts` (3), `extract-candidates.ts` (2), `fan.ts`, `hotspots.ts`, `isolated.ts`, `members.ts`, `methods.ts`, `outline.ts`, `similar.ts` (4), `symbols.ts`, `system.ts`, `test-coverage.ts` (2), `trace.ts` (2)
-
-**Fix:** Replace each inline `AND gs.symbol NOT LIKE '%typeLiteral%' AND gs.symbol NOT LIKE '%().(%'` with `AND ${db.symbolNoise}` (or a string interpolation of the getter). This cuts ~50 lines and ensures the noise filter is defined in one place — if we need to add a new pattern (e.g., filtering out synthetic generics), it changes in one spot.
-
----
-
-## 2. Replace 32 inline `node_modules` exclusions with `db.pathExclusions`
-
-**Found by:** `grep` on `similar-files` output (100% dep-profile similarity across all query modules)
-
-**Problem:** `d.relative_path NOT LIKE 'node_modules/%'` appears 32 times across 18 files. Often paired with `.git/%` exclusions. The `db.pathExclusions` getter exists but isn't used.
-
-**Files affected:** Every query file.
-
-**Fix:** Same pattern as #1 — interpolate `${db.pathExclusions}` where applicable. Some queries use different table aliases (`def_d`, `ref_d`, `d1`, `d2`) so the getter may need a parameter for the alias, or we add alias-specific variants.
-
----
-
-## 3. Extract shared `buildFileDepGraph()` helper
-
-**Found by:** `similar-chains`, `cycles`, `deep-chains`, `similar-files` all contain identical graph-building SQL
-
-**Problem:** Four query modules build the exact same file dependency graph:
-- `cycles.ts:17-35`
-- `deep-chains.ts:16-34`
-- `similar-chains.ts:125-143`
-- `similar-files.ts:66-84`
-
-Each runs this ~18-line SQL query, builds a `Map<string, Set<string>>` adjacency list, and filters by gitignore. The code is identical except for the `scopeFilter` variable name.
-
-**Fix:** Extract a shared `buildFileDepGraph(db, scope?)` helper that returns a `Map<string, Set<string>>`. All four modules import and call it. Saves ~54 lines and ensures graph-building logic stays consistent (e.g., if we later add `.d.ts` exclusions, it changes in one place).
-
----
-
-## 4. Extract shared test-file pattern constants
-
-**Found by:** `grep` on test-pattern strings
-
-**Problem:** Test file path patterns (`%/__tests__/%`, `%.test.%`, `%.spec.%`, etc.) are defined:
-- As an array in `test-coverage.ts:7-16`
-- As individual SQL fragments in `dead.ts:34-37`
-- As individual SQL fragments in `isolated.ts:35-37`
-
-Three different representations of the same concept.
-
-**Fix:** Export a `TEST_FILE_PATTERNS` constant (and a `testFileExclusionSql(alias)` helper that generates the SQL) from a shared location. `dead.ts`, `isolated.ts`, and `test-coverage.ts` all import it.
-
----
-
-## 5. `queries/index.ts` barrel has score 136 bottleneck
-
-**Found by:** `bottlenecks` command
-
-**Problem:** The barrel re-export file (`queries/index.ts`) has fan-in=2, fan-out=68. Every query symbol is re-exported through it, so any consumer (`cli.ts`, `index.ts`) pulls the entire query surface. This is fine for a CLI tool, but if this package is used as a library, consumers pay for every query module even if they use one.
-
-**Fix (for later):** Support tree-shaking by also exporting individual query modules:
-```ts
-// Direct import for library consumers
-import { hotspots } from 'scip-query/queries/hotspots';
-```
-This needs `exports` map entries in `package.json`. Not urgent — the barrel is correct for CLI use.
-
----
-
-## 6. `cli.ts` is the highest fan-out non-barrel file (23 external symbols)
-
-**Found by:** `fan-out` command
-
-**Problem:** `cli.ts` imports from 8 internal modules and references 23 external symbols. It's a 770+ line file that defines 34 commands inline. Each command's `.action()` handler does its own `openDb()` / `queries.X()` / `console.log()` / `db.close()` dance.
-
-**Fix:** This isn't a bug — CLIs are inherently high fan-out. But if the file keeps growing, the repetitive `openDb` → query → format → `close` pattern could be extracted into a `runQuery(queryFn, formatter)` wrapper that handles the lifecycle. Each command would then be ~3 lines instead of ~15.
-
----
-
-## 7. `similar-files` shows 100% similarity across all query modules
-
-**Found by:** `similar-files --min-similarity 0.7`
-
-**Problem:** Every query file depends on the same 3 files: `db.ts`, `types.ts`, `symbol-parser.ts`. This makes the file-level similarity metric saturate at 100%. It's not a code quality issue — it's a signal that the dependency profile is too uniform to distinguish files at this level.
-
-**Implication for the tool itself:** The `similar-files` command should probably discount "universal" dependencies (files imported by >50% of the codebase) to surface more meaningful similarity. Universal deps like `types.ts` are infrastructure, not similarity signals.
-
----
-
-## 8. Callee-set queries repeat identical SQL in `similar.ts` and `extract-candidates.ts`
-
-**Found by:** `similar` command (78% overlap between those two files)
-
-**Problem:** Both `similar.ts` and `extract-candidates.ts` run the same "find all callees of a symbol within its definition range" SQL query. `similar.ts` has it in `findCallees()` (line ~120) and `getAllCalleeFingerprints()` (line ~175). `extract-candidates.ts` has it inline (line ~55).
-
-**Fix:** Extract a `getCalleesForSymbol(db, documentId, startLine, endLine, symbolId)` helper. Used by `similar.ts` (twice) and `extract-candidates.ts` (once). Also usable by `call-graph.ts` which runs a similar query.
-
----
-
-## 9. Deep chains are all rooted at `queries/index.ts` → `cli.ts`
-
-**Found by:** `deep-chains --min-depth 4`
-
-**Problem:** Every deep chain starts at `index.ts` or `cli.ts` because they're the barrel/entry points. The chains themselves are only depth 4-5, which is healthy. No action needed — this confirms the architecture is flat.
-
-**Assessment:** Not an issue. Healthy architecture signal.
-
----
-
-## 10. No circular dependencies
-
-**Found by:** `cycles` command
-
-**Assessment:** Clean. No action needed.
-
----
-
-## Summary
-
-| # | Finding | Severity | Effort | Lines saved |
-|---|---------|----------|--------|-------------|
-| 1 | Inline `symbolNoise` filters (26x) | Medium | Low | ~50 |
-| 2 | Inline `node_modules` exclusions (32x) | Medium | Low | ~30 |
-| 3 | Duplicated graph-building SQL (4 files) | Medium | Low | ~54 |
-| 4 | Duplicated test-file patterns (3 files) | Low | Low | ~15 |
-| 5 | Barrel bottleneck (tree-shaking) | Low | Medium | 0 (structure) |
-| 6 | CLI fan-out / repetitive handlers | Low | Medium | ~100 |
-| 7 | `similar-files` universal dep discount | Low | Medium | 0 (algorithm) |
-| 8 | Duplicated callee-set SQL | Medium | Low | ~30 |
-| 9 | Deep chains rooted at entry points | None | — | — |
-| 10 | No cycles | None | — | — |
-
-**Quick wins (items 1-4, 8):** ~180 lines eliminated, 5 shared helpers, ~30 minutes of work. All low-risk mechanical extractions.
-
-**Structural improvements (items 5-7):** Algorithm and architecture changes that improve the tool's own quality and the accuracy of its similarity detection. Medium effort, high value for the product.

package/PLAN.md

@@ -1,320 +0,0 @@
-# Implementation Plan: Phase 2 Commands + Agent Documentation
-
-This plan adds 10 new analysis commands and a comprehensive agent usage guide. Organized into 4 phases to keep diffs bounded and independently testable.
-
----
-
-## Phase 1: Transitive Impact + Change Planning (3 commands)
-
-These serve Use Case 1 — deep understanding for concrete planning.
-
-### 1.1 `affected <symbol>`
-
-**Purpose:** Full transitive closure of symbols that could break if a given symbol changes. Walks rdeps recursively at the symbol level (not just file level like `rdeps`).
-
-**Implementation:**
-- New file: `src/queries/affected.ts`
-- Algorithm: BFS from the target symbol through the mention graph. For each symbol that references the target, find symbols that reference *that* symbol, and so on. Track depth. Cap at configurable max depth (default 5) to avoid full-graph traversal on hub symbols.
-- Reuse: `findFirstSymbolMatch()` from `query-support.ts` to resolve the target.
-- SQL core: Recursive walk on `mentions` (role=0) → `defn_enclosing_ranges` → back to `mentions`. Each hop finds the enclosing symbol of each reference site, then finds references to *that* symbol.
-- Output type: `AffectedResult` — array of `{ symbol, shortName, file, depth }` sorted by depth then file.
-- CLI: `scip-query affected <symbol> [--max-depth N] [--scope path]`
-
-**Value:** "If I change this function's signature, what's the full blast wave?" Direct rdeps are depth 1. Their consumers are depth 2. This shows the full picture.
-
-### 1.2 `change-surface <file>`
-
-**Purpose:** Pre-change briefing: "I'm about to modify this file. What do I need to know?"
-
-**Implementation:**
-- New file: `src/queries/change-surface.ts`
-- Composes existing queries internally — not a raw SQL query but an orchestrator:
-  1. Call `symbols()` to get all symbols in the file
-  2. For each symbol, call the refs logic to count external consumers
-  3. Call `testCoverage()` to check which symbols are test-covered
-  4. Call `fanIn()` to get reference counts
-- Output type: `ChangeSurfaceResult` — per-symbol: `{ symbol, shortName, externalConsumers: number, testFiles: string[], riskLevel: 'low' | 'medium' | 'high' }` where risk = high if fan-in > 10 and no test coverage.
-- CLI: `scip-query change-surface <file>`
-
-**Value:** One command before modifying a file. Shows what's exported, who uses it, what's tested, and what's risky.
-
-### 1.3 `diff-impact`
-
-**Purpose:** Given the current git diff, compute the affected symbol set.
-
-**Implementation:**
-- New file: `src/queries/diff-impact.ts`
-- Algorithm:
-  1. Run `git diff --name-only HEAD` (via `execFileSync`) to get changed files
-  2. For each changed file, get all symbols defined in it via `symbols()` logic
-  3. For each symbol, get its fan-in count and test coverage
-  4. Aggregate: total changed symbols, total consumers affected, test coverage gaps
-- Output type: `DiffImpactResult` — `{ changedFiles, changedSymbols[], affectedConsumers[], uncoveredSymbols[], summary }`
-- CLI: `scip-query diff-impact [--base <ref>]` (default: diff against HEAD)
-- Note: Needs git available. If not in a git repo, error gracefully.
-
-**Value:** "You changed 3 files — here are the 47 symbols affected, the 12 files that consume them, and the 5 gaps in test coverage."
-
-### Phase 1 files to create:
-- `src/queries/affected.ts`
-- `src/queries/change-surface.ts`
-- `src/queries/diff-impact.ts`
-- Types added to `src/types.ts`
-- Exports added to `src/queries/index.ts`
-- CLI commands added to `src/cli.ts`
-
-### Phase 1 files to modify:
-- `src/types.ts` — add `AffectedResult`, `ChangeSurfaceResult`, `DiffImpactResult`
-- `src/queries/index.ts` — add exports
-- `src/cli.ts` — add 3 commands
-
----
-
-## Phase 2: De-bloating Commands (5 commands)
-
-These serve Use Case 2 — keeping the codebase clean.
-
-### 2.1 `drift [module]`
-
-**Purpose:** Detect pattern drift — files that don't match the typical dependency profile for their directory.
-
-**Implementation:**
-- New file: `src/queries/drift.ts`
-- Algorithm:
-  1. Build file dep profiles per directory (group files by their parent dir)
-  2. For each directory with 3+ files, compute the "median" dependency set — deps that appear in >50% of files in that dir
-  3. For each file, compute how much it deviates from the median: which expected deps are missing, which unexpected deps are present
-  4. Score deviation as a percentage. Report files with highest deviation.
-- Reuse: `buildFileDepGraph()` from `query-support.ts` for the dep edges.
-- Output type: `DriftResult` — `{ file, directory, deviationPercent, missingExpectedDeps[], unexpectedDeps[] }`
-- CLI: `scip-query drift [module] [--min-deviation N]` (default min-deviation: 30%)
-
-**Value:** Finds the files that don't follow the conventions of their neighbors. If 8 of 10 services import a validator and 2 don't, those 2 are flagged.
-
-### 2.2 `wrapper-candidates`
-
-**Purpose:** Find symbols that are only ever called through one intermediary — premature abstractions that add indirection without value.
-
-**Implementation:**
-- New file: `src/queries/wrapper-candidates.ts`
-- Algorithm:
-  1. Find all symbols with fan-in = 1 (exactly one caller)
-  2. For each, check if that single caller has fan-in > 3 (is widely used)
-  3. If so, the single-caller symbol is a wrapper candidate — it could be inlined into its caller
-  4. Also check LOC: small wrappers (< 10 LOC) are the strongest candidates
-- SQL: Subquery on `mentions` grouped by `symbol_id`, `HAVING COUNT(DISTINCT document_id) = 1`, then join to find the caller's fan-in.
-- Output type: `WrapperCandidate` — `{ symbol, shortName, file, loc, singleCaller, callerFanIn }`
-- CLI: `scip-query wrapper-candidates [--scope path] [--max-loc N]`
-
-**Value:** "This function exists only to call another function. You can inline it." Catches over-engineering.
-
-### 2.3 `passthrough-candidates`
-
-**Purpose:** Find functions that just forward to one other function without adding logic.
-
-**Implementation:**
-- New file: `src/queries/passthrough-candidates.ts`
-- Algorithm:
-  1. Find symbols with exactly 1 callee (they only call one external thing)
-  2. Filter to small functions (< 15 LOC)
-  3. These are likely passthroughs: `getUser(id) { return userRepo.findById(id); }`
-- Reuse: `getCalleeRowsForSymbol()` from `query-support.ts` to count callees.
-- Output type: `PassthroughCandidate` — `{ symbol, shortName, file, loc, forwardsTo, forwardsToFile }`
-- CLI: `scip-query passthrough-candidates [--scope path] [--max-loc N]`
-
-**Value:** Finds functions that are pure indirection. Either inline them or verify they exist for a reason (dependency inversion, testing boundary, etc.)
-
-### 2.4 `stale-abstractions`
-
-**Purpose:** Find interfaces/base classes/generics with exactly 1 implementation or 1 caller.
-
-**Implementation:**
-- New file: `src/queries/stale-abstractions.ts`
-- Algorithm:
-  1. Find type-level symbols (using `#` in the SCIP symbol — indicates class/interface/type)
-  2. For each, count cross-file references (fan-in)
-  3. Symbols with fan-in = 1 are single-consumer abstractions
-  4. Cross-reference with LOC: large single-consumer abstractions are the most wasteful
-- SQL: Filter `gs.symbol LIKE '%#%'` (type-level), then count mentions with role=0 from different documents.
-- Output type: `StaleAbstraction` — `{ symbol, shortName, file, loc, consumers: number, implementors: number }`
-- CLI: `scip-query stale-abstractions [--scope path] [--min-loc N]`
-
-**Value:** An interface with one implementation isn't an abstraction — it's indirection. A generic helper called from one place isn't reusable — it's premature.
-
-### 2.5 `complexity-hotspots`
-
-**Purpose:** Composite complexity score per symbol combining LOC, fan-in, fan-out, and callee count.
-
-**Implementation:**
-- New file: `src/queries/complexity-hotspots.ts`
-- Algorithm:
-  1. For each non-trivial symbol (>= minLoc), compute:
-     - LOC (from defn_enclosing_ranges)
-     - Fan-in (count of distinct referencing documents)
-     - Fan-out (count of distinct referenced symbols in other files)
-     - Callee count (total callees within definition range)
-  2. Score = `(LOC / 50) * (fanIn / 5) * (fanOut / 5)` (normalized so a 50-LOC function with 5 consumers and 5 callees scores ~1.0)
-  3. Sort by score descending
-- Reuse: Similar SQL patterns to `bottlenecks.ts` and `fan.ts`.
-- Output type: `ComplexityHotspot` — `{ symbol, shortName, file, loc, fanIn, fanOut, calleeCount, score }`
-- CLI: `scip-query complexity-hotspots [--scope path] [--min-loc N] [-n limit]`
-
-**Value:** The symbols with the highest scores are the ones most likely to contain bugs, be hardest to modify, and benefit most from decomposition. Combines multiple signals into one prioritized view.
-
-### Phase 2 files to create:
-- `src/queries/drift.ts`
-- `src/queries/wrapper-candidates.ts`
-- `src/queries/passthrough-candidates.ts`
-- `src/queries/stale-abstractions.ts`
-- `src/queries/complexity-hotspots.ts`
-
-### Phase 2 files to modify:
-- `src/types.ts` — add 5 result types
-- `src/queries/index.ts` — add 5 exports
-- `src/cli.ts` — add 5 commands
-
----
-
-## Phase 3: Composite Health Report (2 commands)
-
-### 3.1 `health`
-
-**Purpose:** Single command that runs all de-bloat analyses and produces a prioritized action list.
-
-**Implementation:**
-- New file: `src/queries/health.ts`
-- Algorithm: Run each analysis in sequence, aggregate results:
-  1. `dead()` → count dead symbols, total recoverable LOC
-  2. `isolated()` → count orphaned symbols
-  3. `cycles()` → count circular deps
-  4. `similarAll()` → count high-similarity pairs
-  5. `extractCandidates()` → count extraction opportunities
-  6. `wrapperCandidates()` → count wrapper symbols (new, from Phase 2)
-  7. `passthroughCandidates()` → count passthroughs (new, from Phase 2)
-  8. `staleAbstractions()` → count single-consumer types (new, from Phase 2)
-  9. `drift()` → count drifted files (new, from Phase 2)
-  10. `complexityHotspots()` → top 5 most complex symbols (new, from Phase 2)
-- Output: Grouped sections with counts and top items. A "health score" (0-100) based on weighted findings. Concrete action items sorted by effort/impact.
-- Output type: `HealthReport` — sections for each analysis, overall score, prioritized action list.
-- CLI: `scip-query health [--scope path] [--json]` (JSON mode for programmatic consumption by agents)
-
-**Value:** The difference between "powerful tool for experts" and "tool that actually gets used." One command, one report, one action list.
-
-### 3.2 `convergence <symbol1> <symbol2>`
-
-**Purpose:** Given two similar symbols (flagged by `similar`), show what a consolidated version would look like.
-
-**Implementation:**
-- New file: `src/queries/convergence.ts`
-- Algorithm:
-  1. Get callee sets for both symbols (via `getCalleeRowsForSymbol()`)
-  2. Compute shared callees (the body of the consolidated function)
-  3. Compute unique-to-A and unique-to-B (the parameterization points)
-  4. Report: "The consolidated function would call [shared callees]. A's unique behavior ([unique-to-A]) and B's unique behavior ([unique-to-B]) become parameters or strategy arguments."
-  5. Also show the file locations and LOC of both symbols for context.
-- Output type: `ConvergenceResult` — `{ symbolA, symbolB, sharedCallees[], uniqueToA[], uniqueToB[], consolidationStrategy }`
-- CLI: `scip-query convergence <symbol1> <symbol2>`
-
-**Value:** Turns a similarity finding into a concrete refactoring prescription. "These two are 75% similar" becomes "here's what the merged version looks like."
-
-### Phase 3 files to create:
-- `src/queries/health.ts`
-- `src/queries/convergence.ts`
-
-### Phase 3 files to modify:
-- `src/types.ts` — add `HealthReport`, `ConvergenceResult`
-- `src/queries/index.ts` — add 2 exports
-- `src/cli.ts` — add 2 commands
-
-### Phase 3 depends on: Phase 2 (health report calls Phase 2 commands)
-
----
-
-## Phase 4: Agent Usage Guide + Use Case Documentation
-
-### 4.1 `docs/AGENT_GUIDE.md`
-
-Comprehensive guide for AI agents (and humans) on how to use scip-query for specific goals. Structured as goal-oriented workflows, not command reference (that's already in README.md).
-
-**Sections:**
-
-#### "I need to understand how a system works before making changes"
-1. Start with `system <module>` for the full map
-2. Pick the entry point and run `call-graph <symbol>` to see what it calls and who calls it
-3. Run `deps <file>` and `rdeps <file>` to map the file-level dependency boundary
-4. Run `surface <module>` to understand the true public API (not just what's exported)
-5. Run `trace <symbol>` for any specific symbol you need to understand
-6. Run `change-surface <file>` for a pre-change briefing on anything you're about to modify
-7. Run `diff-impact` after making changes to verify the blast radius
-
-#### "I need to write a concrete implementation plan"
-1. Run `system <module>` to understand the target area
-2. Run `symbols <file>` on each file you'll modify to get line ranges and signatures
-3. Run `surface <module>` to identify the public contract you must preserve
-4. Run `refs <symbol>` for any symbol you plan to change, rename, or remove
-5. Run `affected <symbol>` for transitive impact on critical symbols
-6. Run `fan-in <symbol>` to quantify blast radius for each change
-7. Run `test-coverage <symbol>` to identify test gaps before you start
-
-#### "I want to clean up and de-bloat a codebase"
-1. Run `health` for the full prioritized report (start here)
-2. Address dead code first: `dead --min-loc 10 --skip-barrels` → safe deletions
-3. Address isolated symbols: `isolated --min-loc 5` → completely safe deletions
-4. Break cycles: `cycles` → structural fixes
-5. Reduce duplication: `similar --min-similarity 0.6` → consolidation candidates
-6. For each similar pair, run `convergence <a> <b>` to get the refactoring prescription
-7. Find extraction opportunities: `extract-candidates --min-loc 20`
-8. Remove unnecessary indirection: `wrapper-candidates`, `passthrough-candidates`
-9. Prune premature abstractions: `stale-abstractions`
-10. Fix pattern drift: `drift` → bring outlier files into line with their neighbors
-
-#### "I want to assess code quality and risk"
-1. Run `health` for the overall score
-2. Run `complexity-hotspots -n 20` for the riskiest symbols
-3. Run `bottlenecks -n 20` for coupling pressure points
-4. Run `deep-chains --min-depth 5` for architectural layering issues
-5. Run `test-coverage` for the coverage percentage
-6. Run `doc-coverage` for documentation gaps
-
-#### "I want to understand the impact of a change I already made"
-1. Run `diff-impact` to see what your changes affect
-2. Run `affected <symbol>` for any symbol you modified
-3. Run `test-coverage <symbol>` for each affected symbol to find test gaps
-
-#### Command cheat sheet
-Quick reference table: "If you want to know X, run Y."
-
-### 4.2 Update `README.md`
-
-- Add link to AGENT_GUIDE.md in the README
-- Add a "Workflows" section that links to the guide
-- Update command count and command table with new Phase 1-3 commands
-
-### Phase 4 files to create:
-- `docs/AGENT_GUIDE.md`
-
-### Phase 4 files to modify:
-- `README.md` — add workflows section, update command count
-
----
-
-## Execution Order
-
-```
-Phase 1 (impact + planning)     → 3 new commands, ~400 LOC
-Phase 2 (de-bloating)           → 5 new commands, ~500 LOC
-Phase 3 (health + convergence)  → 2 new commands, ~300 LOC
-Phase 4 (documentation)         → 1 new doc, README update
-
-Total: 10 new commands, ~1200 LOC of query logic, 1 agent guide
-```
-
-Each phase is independently testable and committable. Phase 3 depends on Phase 2. Phase 4 depends on Phases 1-3 (references all commands). Phases 1 and 2 are independent and can be built in parallel.
-
-### Shared infrastructure all phases will use:
-- `query-support.ts` — `buildFileDepGraph()`, `findFirstSymbolMatch()`, `getCalleeRowsForSymbol()`
-- `db.ts` — `pathExclusionsFor()`, `symbolNoiseFor()`, `symbolNoise`, `localSymbolPredicate`
-- `clean-signature.ts` — for any command that displays signatures
-- `symbol-parser.ts` — `shortenSymbol()` for all display output
-
-No new shared infrastructure needed. The existing helpers cover all planned commands.

package/dist/chunk-2CKGIR6G.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/queries/extract-candidates.ts"],"sourcesContent":["import type { ScipDatabase } from '../db.js';\nimport { getCalleeRowsForSymbol } from '../query-support.js';\nimport type { ExtractCandidate } from '../types.js';\nimport { shortenSymbol } from '../symbol-parser.js';\n\n/**\n * Find functions with natural extraction seams.\n *\n * A large function that references two distinct groups of symbols —\n * where group A's symbols are never co-referenced with group B's —\n * has a natural extraction boundary. The isolated cluster can likely\n * be pulled into its own function.\n *\n * We detect this by:\n * 1. Finding all callees of a function\n * 2. Building a co-occurrence graph (which callees appear in the same chunk)\n * 3. Finding connected components — disconnected components = extraction seams\n * 4. Scoring each cluster by how isolated it is from the rest\n */\nexport function extractCandidates(\n  db: ScipDatabase,\n  opts: { scope?: string; minLoc?: number; minCallees?: number; limit?: number } = {},\n): ExtractCandidate[] {\n  const { scope, minLoc = 10, minCallees = 6, limit = 20 } = opts;\n  const scopeFilter = scope ? `AND d.relative_path LIKE '%${scope}%'` : '';\n\n  // Find functions large enough to consider\n  const symbols = db.all<{\n    id: number;\n    symbol: string;\n    document_id: number;\n    start_line: number;\n    end_line: number;\n    relative_path: string;\n  }>(\n    `SELECT gs.id, gs.symbol, der.document_id, der.start_line, der.end_line, d.relative_path\n    FROM global_symbols gs\n    JOIN defn_enclosing_ranges der ON gs.id = der.symbol_id\n    JOIN documents d ON der.document_id = d.id\n    WHERE 1 = 1\n      ${db.pathExclusionsFor('d')}\n      ${db.symbolNoiseFor('gs')}\n      AND (der.end_line - der.start_line + 1) >= ?\n      ${scopeFilter}\n    ORDER BY (der.end_line - der.start_line + 1) DESC`,\n    minLoc,\n  );\n\n  const results: ExtractCandidate[] = [];\n\n  for (const sym of symbols) {\n    if (db.isIgnored(sym.relative_path)) continue;\n\n    // Skip pure type files — \"callees\" in a type file are just type references,\n    // not function calls. Splitting type files is a cosmetic choice, not an\n    // extraction opportunity.\n    const basename = sym.relative_path.split('/').pop() ?? '';\n    if (basename.includes('types')) continue;\n\n    // Get callees with their chunk locations (to build co-occurrence)\n    const calleeChunks = getCalleeRowsForSymbol(db, {\n      documentId: sym.document_id,\n      startLine: sym.start_line,\n      endLine: sym.end_line,\n      symbolId: sym.id,\n    });\n\n    // Collect unique callees\n    const calleeSet = new Set(calleeChunks.map((c) => c.symbol));\n    if (calleeSet.size < minCallees) continue;\n\n    // Build co-occurrence graph: two callees are connected if they\n    // appear in the same chunk (meaning they're used in proximity)\n    const cooccurrence = new Map<string, Set<string>>();\n    for (const callee of calleeSet) {\n      cooccurrence.set(callee, new Set());\n    }\n\n    // Group by chunk\n    const chunkToCallees = new Map<number, Set<string>>();\n    for (const cc of calleeChunks) {\n      if (!chunkToCallees.has(cc.chunkId)) chunkToCallees.set(cc.chunkId, new Set());\n      chunkToCallees.get(cc.chunkId)!.add(cc.symbol);\n    }\n\n    // Callees in the same chunk are co-occurring\n    for (const callees of chunkToCallees.values()) {\n      const arr = [...callees];\n      for (let i = 0; i < arr.length; i++) {\n        for (let j = i + 1; j < arr.length; j++) {\n          cooccurrence.get(arr[i]!)!.add(arr[j]!);\n          cooccurrence.get(arr[j]!)!.add(arr[i]!);\n        }\n      }\n    }\n\n    // Find connected components via BFS\n    const visited = new Set<string>();\n    const clusters: Set<string>[] = [];\n\n    for (const callee of calleeSet) {\n      if (visited.has(callee)) continue;\n      const cluster = new Set<string>();\n      const queue = [callee];\n      while (queue.length > 0) {\n        const current = queue.pop()!;\n        if (visited.has(current)) continue;\n        visited.add(current);\n        cluster.add(current);\n        for (const neighbor of cooccurrence.get(current) ?? []) {\n          if (!visited.has(neighbor)) queue.push(neighbor);\n        }\n      }\n      clusters.push(cluster);\n    }\n\n    // Only interesting if there are multiple clusters (= extraction seams exist)\n    if (clusters.length < 2) continue;\n\n    // Score each cluster by isolation:\n    // isolation = 1 - (edges to other clusters / total possible edges to other clusters)\n    const scoredClusters = clusters\n      .filter((c) => c.size >= 2) // single-callee clusters aren't interesting\n      .map((cluster) => {\n        const otherCallees = new Set<string>();\n        for (const c of clusters) {\n          if (c !== cluster) {\n            for (const s of c) otherCallees.add(s);\n          }\n        }\n\n        // Count cross-cluster edges\n        let crossEdges = 0;\n        for (const callee of cluster) {\n          for (const neighbor of cooccurrence.get(callee) ?? []) {\n            if (otherCallees.has(neighbor)) crossEdges++;\n          }\n        }\n\n        const maxCrossEdges = cluster.size * otherCallees.size;\n        const isolation = maxCrossEdges > 0 ? 1 - crossEdges / maxCrossEdges : 1;\n\n        return {\n          callees: [...cluster].map(shortenSymbol),\n          isolation,\n        };\n      })\n      .filter((c) => c.isolation > 0.5) // Only report well-isolated clusters\n      .sort((a, b) => b.isolation - a.isolation);\n\n    if (scoredClusters.length > 0) {\n      results.push({\n        symbol: sym.symbol,\n        shortName: shortenSymbol(sym.symbol),\n        relativePath: sym.relative_path,\n        startLine: sym.start_line,\n        endLine: sym.end_line,\n        loc: sym.end_line - sym.start_line + 1,\n        totalCallees: calleeSet.size,\n        clusters: scoredClusters,\n      });\n    }\n  }\n\n  results.sort((a, b) => b.clusters.length - a.clusters.length || b.loc - a.loc);\n  return results.slice(0, limit);\n}\n"],"mappings":";;;;;;;;AAmBO,SAAS,kBACd,IACA,OAAiF,CAAC,GAC9D;AACpB,QAAM,EAAE,OAAO,SAAS,IAAI,aAAa,GAAG,QAAQ,GAAG,IAAI;AAC3D,QAAM,cAAc,QAAQ,8BAA8B,KAAK,OAAO;AAGtE,QAAM,UAAU,GAAG;AAAA,IAQjB;AAAA;AAAA;AAAA;AAAA;AAAA,QAKI,GAAG,kBAAkB,GAAG,CAAC;AAAA,QACzB,GAAG,eAAe,IAAI,CAAC;AAAA;AAAA,QAEvB,WAAW;AAAA;AAAA,IAEf;AAAA,EACF;AAEA,QAAM,UAA8B,CAAC;AAErC,aAAW,OAAO,SAAS;AACzB,QAAI,GAAG,UAAU,IAAI,aAAa,EAAG;AAKrC,UAAM,WAAW,IAAI,cAAc,MAAM,GAAG,EAAE,IAAI,KAAK;AACvD,QAAI,SAAS,SAAS,OAAO,EAAG;AAGhC,UAAM,eAAe,uBAAuB,IAAI;AAAA,MAC9C,YAAY,IAAI;AAAA,MAChB,WAAW,IAAI;AAAA,MACf,SAAS,IAAI;AAAA,MACb,UAAU,IAAI;AAAA,IAChB,CAAC;AAGD,UAAM,YAAY,IAAI,IAAI,aAAa,IAAI,CAAC,MAAM,EAAE,MAAM,CAAC;AAC3D,QAAI,UAAU,OAAO,WAAY;AAIjC,UAAM,eAAe,oBAAI,IAAyB;AAClD,eAAW,UAAU,WAAW;AAC9B,mBAAa,IAAI,QAAQ,oBAAI,IAAI,CAAC;AAAA,IACpC;AAGA,UAAM,iBAAiB,oBAAI,IAAyB;AACpD,eAAW,MAAM,cAAc;AAC7B,UAAI,CAAC,eAAe,IAAI,GAAG,OAAO,EAAG,gBAAe,IAAI,GAAG,SAAS,oBAAI,IAAI,CAAC;AAC7E,qBAAe,IAAI,GAAG,OAAO,EAAG,IAAI,GAAG,MAAM;AAAA,IAC/C;AAGA,eAAW,WAAW,eAAe,OAAO,GAAG;AAC7C,YAAM,MAAM,CAAC,GAAG,OAAO;AACvB,eAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,KAAK;AACnC,iBAAS,IAAI,IAAI,GAAG,IAAI,IAAI,QAAQ,KAAK;AACvC,uBAAa,IAAI,IAAI,CAAC,CAAE,EAAG,IAAI,IAAI,CAAC,CAAE;AACtC,uBAAa,IAAI,IAAI,CAAC,CAAE,EAAG,IAAI,IAAI,CAAC,CAAE;AAAA,QACxC;AAAA,MACF;AAAA,IACF;AAGA,UAAM,UAAU,oBAAI,IAAY;AAChC,UAAM,WAA0B,CAAC;AAEjC,eAAW,UAAU,WAAW;AAC9B,UAAI,QAAQ,IAAI,MAAM,EAAG;AACzB,YAAM,UAAU,oBAAI,IAAY;AAChC,YAAM,QAAQ,CAAC,MAAM;AACrB,aAAO,MAAM,SAAS,GAAG;AACvB,cAAM,UAAU,MAAM,IAAI;AAC1B,YAAI,QAAQ,IAAI,OAAO,EAAG;AAC1B,gBAAQ,IAAI,OAAO;AACnB,gBAAQ,IAAI,OAAO;AACnB,mBAAW,YAAY,aAAa,IAAI,OAAO,KAAK,CAAC,GAAG;AACtD,cAAI,CAAC,QAAQ,IAAI,QAAQ,EAAG,OAAM,KAAK,QAAQ;AAAA,QACjD;AAAA,MACF;AACA,eAAS,KAAK,OAAO;AAAA,IACvB;AAGA,QAAI,SAAS,SAAS,EAAG;AAIzB,UAAM,iBAAiB,SACpB,OAAO,CAAC,MAAM,EAAE,QAAQ,CAAC,EACzB,IAAI,CAAC,YAAY;AAChB,YAAM,eAAe,oBAAI,IAAY;AACrC,iBAAW,KAAK,UAAU;AACxB,YAAI,MAAM,SAAS;AACjB,qBAAW,KAAK,EAAG,cAAa,IAAI,CAAC;AAAA,QACvC;AAAA,MACF;AAGA,UAAI,aAAa;AACjB,iBAAW,UAAU,SAAS;AAC5B,mBAAW,YAAY,aAAa,IAAI,MAAM,KAAK,CAAC,GAAG;AACrD,cAAI,aAAa,IAAI,QAAQ,EAAG;AAAA,QAClC;AAAA,MACF;AAEA,YAAM,gBAAgB,QAAQ,OAAO,aAAa;AAClD,YAAM,YAAY,gBAAgB,IAAI,IAAI,aAAa,gBAAgB;AAEvE,aAAO;AAAA,QACL,SAAS,CAAC,GAAG,OAAO,EAAE,IAAI,aAAa;AAAA,QACvC;AAAA,MACF;AAAA,IACF,CAAC,EACA,OAAO,CAAC,MAAM,EAAE,YAAY,GAAG,EAC/B,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAE3C,QAAI,eAAe,SAAS,GAAG;AAC7B,cAAQ,KAAK;AAAA,QACX,QAAQ,IAAI;AAAA,QACZ,WAAW,cAAc,IAAI,MAAM;AAAA,QACnC,cAAc,IAAI;AAAA,QAClB,WAAW,IAAI;AAAA,QACf,SAAS,IAAI;AAAA,QACb,KAAK,IAAI,WAAW,IAAI,aAAa;AAAA,QACrC,cAAc,UAAU;AAAA,QACxB,UAAU;AAAA,MACZ,CAAC;AAAA,IACH;AAAA,EACF;AAEA,UAAQ,KAAK,CAAC,GAAG,MAAM,EAAE,SAAS,SAAS,EAAE,SAAS,UAAU,EAAE,MAAM,EAAE,GAAG;AAC7E,SAAO,QAAQ,MAAM,GAAG,KAAK;AAC/B;","names":[]}

package/dist/chunk-3UOUTZQT.js

@@ -1,45 +0,0 @@
-import {
-  parseSymbol,
-  shortenSymbol
-} from "./chunk-QOV2R2WT.js";
-
-// src/queries/members.ts
-function members(db, symbolPattern) {
-  const parents = db.all(
-    `SELECT symbol FROM global_symbols WHERE symbol LIKE ?`,
-    `%${symbolPattern}%`
-  );
-  if (parents.length === 0) return [];
-  const placeholders = parents.map(() => "?").join(",");
-  const parentSymbols = parents.map((p) => p.symbol);
-  const rows = db.all(
-    `SELECT gs.symbol, der.start_line, der.end_line
-    FROM global_symbols gs
-    JOIN defn_enclosing_ranges der ON gs.id = der.symbol_id
-    WHERE gs.enclosing_symbol IN (${placeholders})
-      ${db.symbolNoise}
-    ORDER BY der.start_line`,
-    ...parentSymbols
-  );
-  return rows.map((r) => {
-    const parsed = parseSymbol(r.symbol);
-    let kind = "unknown";
-    if (!("kind" in parsed)) {
-      const sym = parsed;
-      const last = sym.descriptors[sym.descriptors.length - 1];
-      if (last) kind = last.suffix;
-    }
-    return {
-      symbol: r.symbol,
-      shortName: shortenSymbol(r.symbol),
-      startLine: r.start_line,
-      endLine: r.end_line,
-      kind
-    };
-  });
-}
-
-export {
-  members
-};
-//# sourceMappingURL=chunk-3UOUTZQT.js.map

package/dist/chunk-3UOUTZQT.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/queries/members.ts"],"sourcesContent":["import type { ScipDatabase } from '../db.js';\nimport type { MemberResult } from '../types.js';\nimport { shortenSymbol, parseSymbol } from '../symbol-parser.js';\nimport type { ScipSymbol } from '../types.js';\n\n/**\n * Find all direct children of a symbol (methods, fields, nested types).\n * Uses the enclosing_symbol relationship in global_symbols.\n */\nexport function members(db: ScipDatabase, symbolPattern: string): MemberResult[] {\n  // First find the parent symbol(s)\n  const parents = db.all<{ symbol: string }>(\n    `SELECT symbol FROM global_symbols WHERE symbol LIKE ?`,\n    `%${symbolPattern}%`,\n  );\n\n  if (parents.length === 0) return [];\n\n  // Find children whose enclosing_symbol matches any parent\n  const placeholders = parents.map(() => '?').join(',');\n  const parentSymbols = parents.map((p) => p.symbol);\n\n  const rows = db.all<{\n    symbol: string;\n    start_line: number;\n    end_line: number;\n  }>(\n    `SELECT gs.symbol, der.start_line, der.end_line\n    FROM global_symbols gs\n    JOIN defn_enclosing_ranges der ON gs.id = der.symbol_id\n    WHERE gs.enclosing_symbol IN (${placeholders})\n      ${db.symbolNoise}\n    ORDER BY der.start_line`,\n    ...parentSymbols,\n  );\n\n  return rows.map((r) => {\n    const parsed = parseSymbol(r.symbol);\n    let kind = 'unknown';\n    if (!('kind' in parsed)) {\n      const sym = parsed as ScipSymbol;\n      const last = sym.descriptors[sym.descriptors.length - 1];\n      if (last) kind = last.suffix;\n    }\n\n    return {\n      symbol: r.symbol,\n      shortName: shortenSymbol(r.symbol),\n      startLine: r.start_line,\n      endLine: r.end_line,\n      kind,\n    };\n  });\n}\n"],"mappings":";;;;;;AASO,SAAS,QAAQ,IAAkB,eAAuC;AAE/E,QAAM,UAAU,GAAG;AAAA,IACjB;AAAA,IACA,IAAI,aAAa;AAAA,EACnB;AAEA,MAAI,QAAQ,WAAW,EAAG,QAAO,CAAC;AAGlC,QAAM,eAAe,QAAQ,IAAI,MAAM,GAAG,EAAE,KAAK,GAAG;AACpD,QAAM,gBAAgB,QAAQ,IAAI,CAAC,MAAM,EAAE,MAAM;AAEjD,QAAM,OAAO,GAAG;AAAA,IAKd;AAAA;AAAA;AAAA,oCAGgC,YAAY;AAAA,QACxC,GAAG,WAAW;AAAA;AAAA,IAElB,GAAG;AAAA,EACL;AAEA,SAAO,KAAK,IAAI,CAAC,MAAM;AACrB,UAAM,SAAS,YAAY,EAAE,MAAM;AACnC,QAAI,OAAO;AACX,QAAI,EAAE,UAAU,SAAS;AACvB,YAAM,MAAM;AACZ,YAAM,OAAO,IAAI,YAAY,IAAI,YAAY,SAAS,CAAC;AACvD,UAAI,KAAM,QAAO,KAAK;AAAA,IACxB;AAEA,WAAO;AAAA,MACL,QAAQ,EAAE;AAAA,MACV,WAAW,cAAc,EAAE,MAAM;AAAA,MACjC,WAAW,EAAE;AAAA,MACb,SAAS,EAAE;AAAA,MACX;AAAA,IACF;AAAA,EACF,CAAC;AACH;","names":[]}

package/dist/chunk-4EXL2CUA.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/queries/dead.ts"],"sourcesContent":["import type { ScipDatabase } from '../db.js';\nimport { TEST_SUPPORT_PATH_PATTERNS, testFileExclusionSql } from '../query-support.js';\nimport type { DeadOptions, DeadSymbolResult, DeadSummary } from '../types.js';\nimport { shortenSymbol } from '../symbol-parser.js';\n\n/**\n * Find dead exports: symbols defined locally with no cross-file references.\n * Language-agnostic — works with any SCIP index.\n */\nexport function dead(db: ScipDatabase, opts: DeadOptions = {}): DeadSummary {\n  const {\n    scope,\n    minLoc = 1,\n    includeTests = false,\n    skipBarrels = false,\n    includeMembers = false,\n  } = opts;\n\n  const params: unknown[] = [minLoc];\n  let testFileExclusions = '';\n  let memberExclusion = '';\n\n  if (scope) {\n    params.push(`%${scope}%`);\n  }\n\n  if (!includeTests) {\n    testFileExclusions = `\n      AND ${testFileExclusionSql('d', TEST_SUPPORT_PATH_PATTERNS)}\n    `;\n  }\n\n  if (!includeMembers) {\n    memberExclusion = `AND gs.symbol NOT LIKE '%#%'`;\n  }\n\n  // Barrel file exclusion for the NOT EXISTS subquery\n  const barrelExclusions = skipBarrels\n    ? `AND ref_d.relative_path NOT LIKE '%/index.ts'\n       AND ref_d.relative_path NOT LIKE '%/index.js'\n       AND ref_d.relative_path NOT LIKE '%/mod.rs'\n       AND ref_d.relative_path NOT LIKE '%/__init__.py'`\n    : '';\n\n  const sql = `\n    SELECT\n      d.relative_path,\n      der.start_line,\n      der.end_line,\n      (der.end_line - der.start_line + 1) AS loc,\n      gs.symbol,\n      (SELECT COUNT(*) FROM mentions m2\n       JOIN chunks c2 ON m2.chunk_id = c2.id\n       WHERE m2.symbol_id = gs.id AND m2.role != 1 AND c2.document_id = d.id\n      ) AS same_file_refs\n    FROM global_symbols gs\n    JOIN defn_enclosing_ranges der ON gs.id = der.symbol_id\n    JOIN documents d ON der.document_id = d.id\n    WHERE 1 = 1\n      ${db.pathExclusionsFor('d')}\n      ${db.symbolNoiseFor('gs')}\n      AND (der.end_line - der.start_line + 1) >= ?\n      ${scope ? 'AND d.relative_path LIKE ?' : ''}\n      ${testFileExclusions}\n      ${memberExclusion}\n      AND NOT EXISTS (\n        SELECT 1\n        FROM mentions ref_m\n        JOIN chunks ref_c ON ref_m.chunk_id = ref_c.id\n        JOIN documents ref_d ON ref_c.document_id = ref_d.id\n        WHERE ref_m.symbol_id = gs.id\n          AND ref_m.role != 1\n          AND ref_d.id != d.id\n          ${barrelExclusions}\n      )\n    ORDER BY (der.end_line - der.start_line + 1) DESC, d.relative_path, der.start_line\n  `;\n\n  const rows = db.all<{\n    relative_path: string;\n    start_line: number;\n    end_line: number;\n    loc: number;\n    symbol: string;\n    same_file_refs: number;\n  }>(sql, ...params);\n\n  let deadCodeCount = 0;\n  let fileInternalCount = 0;\n  let totalLoc = 0;\n\n  const symbols: DeadSymbolResult[] = rows\n    .filter((r) => !db.isIgnored(r.relative_path))\n    .map((r) => {\n      // dead-code: zero references anywhere (not even in same file) — safe to delete\n      // file-internal: referenced within same file but never cross-file —\n      //   may be a private helper (fine) or a forgotten export (needs review)\n      const kind = r.same_file_refs === 0 ? 'dead-code' : 'file-internal';\n      if (kind === 'dead-code') deadCodeCount++;\n      else fileInternalCount++;\n      totalLoc += r.loc;\n\n      return {\n        relativePath: r.relative_path,\n        startLine: r.start_line,\n        endLine: r.end_line,\n        loc: r.loc,\n        symbol: r.symbol,\n        shortName: shortenSymbol(r.symbol),\n        sameFileRefs: r.same_file_refs,\n        kind,\n      };\n    });\n\n  return {\n    symbols,\n    totalCount: symbols.length,\n    deadCodeCount,\n    fileInternalCount,\n    totalLoc,\n  };\n}\n"],"mappings":";;;;;;;;;AASO,SAAS,KAAK,IAAkB,OAAoB,CAAC,GAAgB;AAC1E,QAAM;AAAA,IACJ;AAAA,IACA,SAAS;AAAA,IACT,eAAe;AAAA,IACf,cAAc;AAAA,IACd,iBAAiB;AAAA,EACnB,IAAI;AAEJ,QAAM,SAAoB,CAAC,MAAM;AACjC,MAAI,qBAAqB;AACzB,MAAI,kBAAkB;AAEtB,MAAI,OAAO;AACT,WAAO,KAAK,IAAI,KAAK,GAAG;AAAA,EAC1B;AAEA,MAAI,CAAC,cAAc;AACjB,yBAAqB;AAAA,YACb,qBAAqB,KAAK,0BAA0B,CAAC;AAAA;AAAA,EAE/D;AAEA,MAAI,CAAC,gBAAgB;AACnB,sBAAkB;AAAA,EACpB;AAGA,QAAM,mBAAmB,cACrB;AAAA;AAAA;AAAA,2DAIA;AAEJ,QAAM,MAAM;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,QAeN,GAAG,kBAAkB,GAAG,CAAC;AAAA,QACzB,GAAG,eAAe,IAAI,CAAC;AAAA;AAAA,QAEvB,QAAQ,+BAA+B,EAAE;AAAA,QACzC,kBAAkB;AAAA,QAClB,eAAe;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,YASX,gBAAgB;AAAA;AAAA;AAAA;AAK1B,QAAM,OAAO,GAAG,IAOb,KAAK,GAAG,MAAM;AAEjB,MAAI,gBAAgB;AACpB,MAAI,oBAAoB;AACxB,MAAI,WAAW;AAEf,QAAM,UAA8B,KACjC,OAAO,CAAC,MAAM,CAAC,GAAG,UAAU,EAAE,aAAa,CAAC,EAC5C,IAAI,CAAC,MAAM;AAIV,UAAM,OAAO,EAAE,mBAAmB,IAAI,cAAc;AACpD,QAAI,SAAS,YAAa;AAAA,QACrB;AACL,gBAAY,EAAE;AAEd,WAAO;AAAA,MACL,cAAc,EAAE;AAAA,MAChB,WAAW,EAAE;AAAA,MACb,SAAS,EAAE;AAAA,MACX,KAAK,EAAE;AAAA,MACP,QAAQ,EAAE;AAAA,MACV,WAAW,cAAc,EAAE,MAAM;AAAA,MACjC,cAAc,EAAE;AAAA,MAChB;AAAA,IACF;AAAA,EACF,CAAC;AAEH,SAAO;AAAA,IACL;AAAA,IACA,YAAY,QAAQ;AAAA,IACpB;AAAA,IACA;AAAA,IACA;AAAA,EACF;AACF;","names":[]}

package/dist/chunk-4PDAL6IL.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/queries/wrapper-candidates.ts"],"sourcesContent":["import type { ScipDatabase } from '../db.js';\nimport { testFileExclusionSql } from '../query-support.js';\nimport type { WrapperCandidate } from '../types.js';\nimport { shortenSymbol } from '../symbol-parser.js';\n\n/**\n * Find wrapper candidates: symbols called by only one other symbol.\n *\n * These are premature abstractions that add indirection without\n * providing reuse. A function with fan-in = 1 whose sole caller\n * is widely used is a strong signal of unnecessary wrapping.\n */\nexport function wrapperCandidates(\n  db: ScipDatabase,\n  opts?: { scope?: string; maxLoc?: number; limit?: number },\n): WrapperCandidate[] {\n  const { scope, maxLoc = 15, limit = 30 } = opts ?? {};\n  const scopeFilter = scope ? `AND d.relative_path LIKE '%${scope}%'` : '';\n\n  // Find all symbols with exactly 1 cross-file consumer (fan-in = 1),\n  // along with who that single caller is and the caller's own fan-in.\n  const rows = db.all<{\n    symbol: string;\n    file: string;\n    start_line: number;\n    end_line: number;\n    loc: number;\n    caller_symbol: string;\n    caller_fan_in: number;\n  }>(\n    `SELECT * FROM (\n      SELECT\n        gs.symbol,\n        d.relative_path AS file,\n        der.start_line,\n        der.end_line,\n        (der.end_line - der.start_line + 1) AS loc,\n        -- The single caller: the symbol whose definition range contains\n        -- the chunk that references our target\n        (SELECT caller_gs.symbol\n         FROM mentions ref_m\n         JOIN chunks ref_c ON ref_m.chunk_id = ref_c.id\n         JOIN defn_enclosing_ranges caller_der\n           ON caller_der.document_id = ref_c.document_id\n           AND ref_c.start_line >= caller_der.start_line\n           AND ref_c.end_line <= caller_der.end_line\n         JOIN global_symbols caller_gs ON caller_der.symbol_id = caller_gs.id\n         WHERE ref_m.symbol_id = gs.id\n           AND ref_m.role != 1\n           AND ref_c.document_id != der.document_id\n         LIMIT 1\n        ) AS caller_symbol,\n        -- Fan-in of that single caller\n        (SELECT COUNT(DISTINCT caller_ref_c.document_id)\n         FROM mentions caller_ref_m\n         JOIN chunks caller_ref_c ON caller_ref_m.chunk_id = caller_ref_c.id\n         WHERE caller_ref_m.symbol_id = (\n           SELECT caller_der2.symbol_id\n           FROM mentions ref_m2\n           JOIN chunks ref_c2 ON ref_m2.chunk_id = ref_c2.id\n           JOIN defn_enclosing_ranges caller_der2\n             ON caller_der2.document_id = ref_c2.document_id\n             AND ref_c2.start_line >= caller_der2.start_line\n             AND ref_c2.end_line <= caller_der2.end_line\n           WHERE ref_m2.symbol_id = gs.id\n             AND ref_m2.role != 1\n             AND ref_c2.document_id != der.document_id\n           LIMIT 1\n         )\n         AND caller_ref_m.role != 1\n        ) AS caller_fan_in\n      FROM global_symbols gs\n      JOIN defn_enclosing_ranges der ON gs.id = der.symbol_id\n      JOIN documents d ON der.document_id = d.id\n      WHERE 1 = 1\n        ${db.pathExclusionsFor('d')}\n        AND ${testFileExclusionSql('d')}\n        ${db.symbolNoiseFor('gs')}\n        -- Only functions/terms, not type definitions (types with # are not wrappers)\n        AND gs.symbol NOT LIKE '%#'\n        AND (der.end_line - der.start_line + 1) <= ?\n        AND (der.end_line - der.start_line + 1) >= 2\n        ${scopeFilter}\n        -- Exactly 1 cross-file consumer\n        AND (\n          SELECT COUNT(DISTINCT ref_c.document_id)\n          FROM mentions ref_m\n          JOIN chunks ref_c ON ref_m.chunk_id = ref_c.id\n          WHERE ref_m.symbol_id = gs.id\n            AND ref_m.role != 1\n            AND ref_c.document_id != der.document_id\n        ) = 1\n    ) WHERE caller_symbol IS NOT NULL AND caller_fan_in > 3\n    ORDER BY caller_fan_in DESC, loc DESC\n    LIMIT ?`,\n    maxLoc, limit,\n  );\n\n  return rows\n    .filter((r) => !db.isIgnored(r.file))\n    .map((r) => ({\n      symbol: r.symbol,\n      shortName: shortenSymbol(r.symbol),\n      file: r.file,\n      startLine: r.start_line,\n      endLine: r.end_line,\n      loc: r.loc,\n      singleCaller: r.caller_symbol,\n      singleCallerShort: shortenSymbol(r.caller_symbol),\n      callerFanIn: r.caller_fan_in,\n    }));\n}\n"],"mappings":";;;;;;;;AAYO,SAAS,kBACd,IACA,MACoB;AACpB,QAAM,EAAE,OAAO,SAAS,IAAI,QAAQ,GAAG,IAAI,QAAQ,CAAC;AACpD,QAAM,cAAc,QAAQ,8BAA8B,KAAK,OAAO;AAItE,QAAM,OAAO,GAAG;AAAA,IASd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UA6CM,GAAG,kBAAkB,GAAG,CAAC;AAAA,cACrB,qBAAqB,GAAG,CAAC;AAAA,UAC7B,GAAG,eAAe,IAAI,CAAC;AAAA;AAAA;AAAA;AAAA;AAAA,UAKvB,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,IAajB;AAAA,IAAQ;AAAA,EACV;AAEA,SAAO,KACJ,OAAO,CAAC,MAAM,CAAC,GAAG,UAAU,EAAE,IAAI,CAAC,EACnC,IAAI,CAAC,OAAO;AAAA,IACX,QAAQ,EAAE;AAAA,IACV,WAAW,cAAc,EAAE,MAAM;AAAA,IACjC,MAAM,EAAE;AAAA,IACR,WAAW,EAAE;AAAA,IACb,SAAS,EAAE;AAAA,IACX,KAAK,EAAE;AAAA,IACP,cAAc,EAAE;AAAA,IAChB,mBAAmB,cAAc,EAAE,aAAa;AAAA,IAChD,aAAa,EAAE;AAAA,EACjB,EAAE;AACN;","names":[]}

package/dist/chunk-4TYLS5XX.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/queries/clean-signature.ts"],"sourcesContent":["/**\n * Clean up the raw doc/signature string from the SCIP index.\n *\n * Shared across symbols, trace, and system queries.\n * Previously duplicated as cleanSig/cleanSignature in three files.\n */\nexport function cleanSignature(sig: string | null): string | null {\n  if (!sig || !sig.trim()) return null;\n  return sig\n    .replace(/^```\\w*\\s*/, '')\n    .replace(/\\s*```$/, '')\n    .replace(/^\\(method\\)\\s*/, '')\n    .replace(/^\\(property\\)\\s*/, '')\n    .replace(/^\\(function\\)\\s*/, '')\n    .replace(/^\\(class\\)\\s*/, '')\n    .replace(/^\\(interface\\)\\s*/, '')\n    .replace(/^\\(enum\\)\\s*/, '')\n    .replace(/^\\(type alias\\)\\s*/, '')\n    .replace(/^\\(const\\)\\s*/, '')\n    .replace(/^\\(var\\)\\s*/, '')\n    .trim() || null;\n}\n"],"mappings":";AAMO,SAAS,eAAe,KAAmC;AAChE,MAAI,CAAC,OAAO,CAAC,IAAI,KAAK,EAAG,QAAO;AAChC,SAAO,IACJ,QAAQ,cAAc,EAAE,EACxB,QAAQ,WAAW,EAAE,EACrB,QAAQ,kBAAkB,EAAE,EAC5B,QAAQ,oBAAoB,EAAE,EAC9B,QAAQ,oBAAoB,EAAE,EAC9B,QAAQ,iBAAiB,EAAE,EAC3B,QAAQ,qBAAqB,EAAE,EAC/B,QAAQ,gBAAgB,EAAE,EAC1B,QAAQ,sBAAsB,EAAE,EAChC,QAAQ,iBAAiB,EAAE,EAC3B,QAAQ,eAAe,EAAE,EACzB,KAAK,KAAK;AACf;","names":[]}

package/dist/chunk-4XHWPRAX.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/queries/complexity.ts"],"sourcesContent":["import { readFileSync } from 'node:fs';\nimport { join } from 'node:path';\nimport type { ScipDatabase } from '../db.js';\nimport { findFirstSymbolMatch, getCalleeRowsForSymbol } from '../query-support.js';\nimport type { ComplexityResult } from '../types.js';\nimport { shortenSymbol } from '../symbol-parser.js';\n\n/**\n * Per-symbol complexity analysis combining source-level branch counting\n * with index-level metrics (fan-in, fan-out, callee count).\n *\n * Branch counting uses language-aware regex. The language is read from\n * the SCIP documents table, so it works for any indexed language.\n */\nexport function complexity(\n  db: ScipDatabase,\n  symbolPattern: string,\n): ComplexityResult | null {\n  const match = findFirstSymbolMatch(db, symbolPattern);\n  if (!match) return null;\n\n  // Get language\n  const doc = db.get<{ language: string | null }>(\n    `SELECT language FROM documents WHERE relative_path = ?`,\n    match.relativePath,\n  );\n  const language = doc?.language ?? 'unknown';\n\n  // Read source for branch counting\n  const filePath = join(db.config.projectRoot, match.relativePath);\n  let source = '';\n  try {\n    const lines = readFileSync(filePath, 'utf-8').split('\\n');\n    source = lines.slice(match.startLine, match.endLine + 1).join('\\n');\n  } catch {\n    // If we can't read the file, just skip branch counting\n  }\n\n  const branches = countBranches(source, language);\n  const loc = match.endLine - match.startLine + 1;\n\n  // Callee count\n  const callees = getCalleeRowsForSymbol(db, match);\n  const uniqueCallees = new Set(callees.map((c) => c.symbol));\n\n  // Fan-in\n  const fanInRow = db.get<{ c: number }>(\n    `SELECT COUNT(DISTINCT c.document_id) AS c\n    FROM mentions m\n    JOIN chunks c ON m.chunk_id = c.id\n    WHERE m.symbol_id = ? AND m.role != 1`,\n    match.symbolId,\n  );\n\n  // Fan-out (callees in other files)\n  const fanOut = new Set(\n    callees.filter((c) => c.file !== match.relativePath).map((c) => c.symbol),\n  ).size;\n\n  return {\n    symbol: match.symbol,\n    shortName: shortenSymbol(match.symbol),\n    relativePath: match.relativePath,\n    startLine: match.startLine,\n    endLine: match.endLine,\n    loc,\n    branches,\n    cyclomaticEstimate: branches + 1,\n    calleeCount: uniqueCallees.size,\n    fanIn: fanInRow?.c ?? 0,\n    fanOut,\n  };\n}\n\n/**\n * Count branch points in source code using language-aware regex.\n * Works across all SCIP-supported languages.\n */\nfunction countBranches(source: string, language: string): number {\n  // Strip comments and strings to avoid false positives\n  const stripped = stripCommentsAndStrings(source);\n  let count = 0;\n\n  // Universal branch keywords (work across most C-family languages)\n  const universalPatterns = [\n    /\\bif\\b/g,\n    /\\belse\\s+if\\b/g,\n    /\\belse\\b/g,\n    /\\bfor\\b/g,\n    /\\bwhile\\b/g,\n    /\\bswitch\\b/g,\n    /\\bcase\\b/g,\n    /\\bcatch\\b/g,\n    /\\?\\s*[^?]/g,    // ternary (but not ??)\n    /&&/g,\n    /\\|\\|/g,\n  ];\n\n  for (const pattern of universalPatterns) {\n    const matches = stripped.match(pattern);\n    if (matches) count += matches.length;\n  }\n\n  // Language-specific patterns\n  if (language === 'python') {\n    const pyPatterns = [/\\belif\\b/g, /\\bexcept\\b/g, /\\bfinally\\b/g];\n    for (const p of pyPatterns) {\n      const m = stripped.match(p);\n      if (m) count += m.length;\n    }\n  } else if (language === 'rust') {\n    const rustPatterns = [/\\bmatch\\b/g, /=>/g, /\\bloop\\b/g];\n    for (const p of rustPatterns) {\n      const m = stripped.match(p);\n      if (m) count += m.length;\n    }\n  } else if (language === 'ruby') {\n    const rubyPatterns = [/\\belsif\\b/g, /\\bunless\\b/g, /\\brescue\\b/g, /\\bwhen\\b/g];\n    for (const p of rubyPatterns) {\n      const m = stripped.match(p);\n      if (m) count += m.length;\n    }\n  } else if (language === 'go') {\n    const goPatterns = [/\\bselect\\b/g, /\\bdefer\\b/g];\n    for (const p of goPatterns) {\n      const m = stripped.match(p);\n      if (m) count += m.length;\n    }\n  }\n\n  return count;\n}\n\n/**\n * Rough strip of comments and string literals to reduce false positives\n * in branch counting. Not perfect but good enough for estimation.\n */\nfunction stripCommentsAndStrings(source: string): string {\n  return source\n    // Block comments\n    .replace(/\\/\\*[\\s\\S]*?\\*\\//g, '')\n    // Line comments\n    .replace(/\\/\\/.*/g, '')\n    // Python/Ruby line comments\n    .replace(/#.*/g, '')\n    // Double-quoted strings\n    .replace(/\"(?:[^\"\\\\]|\\\\.)*\"/g, '\"\"')\n    // Single-quoted strings\n    .replace(/'(?:[^'\\\\]|\\\\.)*'/g, \"''\")\n    // Template literals\n    .replace(/`(?:[^`\\\\]|\\\\.)*`/g, '``');\n}\n"],"mappings":";;;;;;;;;AAAA,SAAS,oBAAoB;AAC7B,SAAS,YAAY;AAad,SAAS,WACd,IACA,eACyB;AACzB,QAAM,QAAQ,qBAAqB,IAAI,aAAa;AACpD,MAAI,CAAC,MAAO,QAAO;AAGnB,QAAM,MAAM,GAAG;AAAA,IACb;AAAA,IACA,MAAM;AAAA,EACR;AACA,QAAM,WAAW,KAAK,YAAY;AAGlC,QAAM,WAAW,KAAK,GAAG,OAAO,aAAa,MAAM,YAAY;AAC/D,MAAI,SAAS;AACb,MAAI;AACF,UAAM,QAAQ,aAAa,UAAU,OAAO,EAAE,MAAM,IAAI;AACxD,aAAS,MAAM,MAAM,MAAM,WAAW,MAAM,UAAU,CAAC,EAAE,KAAK,IAAI;AAAA,EACpE,QAAQ;AAAA,EAER;AAEA,QAAM,WAAW,cAAc,QAAQ,QAAQ;AAC/C,QAAM,MAAM,MAAM,UAAU,MAAM,YAAY;AAG9C,QAAM,UAAU,uBAAuB,IAAI,KAAK;AAChD,QAAM,gBAAgB,IAAI,IAAI,QAAQ,IAAI,CAAC,MAAM,EAAE,MAAM,CAAC;AAG1D,QAAM,WAAW,GAAG;AAAA,IAClB;AAAA;AAAA;AAAA;AAAA,IAIA,MAAM;AAAA,EACR;AAGA,QAAM,SAAS,IAAI;AAAA,IACjB,QAAQ,OAAO,CAAC,MAAM,EAAE,SAAS,MAAM,YAAY,EAAE,IAAI,CAAC,MAAM,EAAE,MAAM;AAAA,EAC1E,EAAE;AAEF,SAAO;AAAA,IACL,QAAQ,MAAM;AAAA,IACd,WAAW,cAAc,MAAM,MAAM;AAAA,IACrC,cAAc,MAAM;AAAA,IACpB,WAAW,MAAM;AAAA,IACjB,SAAS,MAAM;AAAA,IACf;AAAA,IACA;AAAA,IACA,oBAAoB,WAAW;AAAA,IAC/B,aAAa,cAAc;AAAA,IAC3B,OAAO,UAAU,KAAK;AAAA,IACtB;AAAA,EACF;AACF;AAMA,SAAS,cAAc,QAAgB,UAA0B;AAE/D,QAAM,WAAW,wBAAwB,MAAM;AAC/C,MAAI,QAAQ;AAGZ,QAAM,oBAAoB;AAAA,IACxB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAEA,aAAW,WAAW,mBAAmB;AACvC,UAAM,UAAU,SAAS,MAAM,OAAO;AACtC,QAAI,QAAS,UAAS,QAAQ;AAAA,EAChC;AAGA,MAAI,aAAa,UAAU;AACzB,UAAM,aAAa,CAAC,aAAa,eAAe,cAAc;AAC9D,eAAW,KAAK,YAAY;AAC1B,YAAM,IAAI,SAAS,MAAM,CAAC;AAC1B,UAAI,EAAG,UAAS,EAAE;AAAA,IACpB;AAAA,EACF,WAAW,aAAa,QAAQ;AAC9B,UAAM,eAAe,CAAC,cAAc,OAAO,WAAW;AACtD,eAAW,KAAK,cAAc;AAC5B,YAAM,IAAI,SAAS,MAAM,CAAC;AAC1B,UAAI,EAAG,UAAS,EAAE;AAAA,IACpB;AAAA,EACF,WAAW,aAAa,QAAQ;AAC9B,UAAM,eAAe,CAAC,cAAc,eAAe,eAAe,WAAW;AAC7E,eAAW,KAAK,cAAc;AAC5B,YAAM,IAAI,SAAS,MAAM,CAAC;AAC1B,UAAI,EAAG,UAAS,EAAE;AAAA,IACpB;AAAA,EACF,WAAW,aAAa,MAAM;AAC5B,UAAM,aAAa,CAAC,eAAe,YAAY;AAC/C,eAAW,KAAK,YAAY;AAC1B,YAAM,IAAI,SAAS,MAAM,CAAC;AAC1B,UAAI,EAAG,UAAS,EAAE;AAAA,IACpB;AAAA,EACF;AAEA,SAAO;AACT;AAMA,SAAS,wBAAwB,QAAwB;AACvD,SAAO,OAEJ,QAAQ,qBAAqB,EAAE,EAE/B,QAAQ,WAAW,EAAE,EAErB,QAAQ,QAAQ,EAAE,EAElB,QAAQ,sBAAsB,IAAI,EAElC,QAAQ,sBAAsB,IAAI,EAElC,QAAQ,sBAAsB,IAAI;AACvC;","names":[]}

package/dist/chunk-5RMYT5WH.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../src/queries/bottlenecks.ts"],"sourcesContent":["import type { ScipDatabase } from '../db.js';\nimport type { BottleneckResult } from '../types.js';\nimport { shortenSymbol } from '../symbol-parser.js';\n\n/**\n * Find coupling hubs: symbols with both high fan-in (many consumers)\n * AND high fan-out (references many other symbols).\n *\n * These are the most dangerous symbols to change — they sit at the\n * intersection of many dependency paths. Score = fanIn * fanOut.\n */\nexport function bottlenecks(\n  db: ScipDatabase,\n  opts: { limit?: number; scope?: string; minFanIn?: number; minFanOut?: number } = {},\n): BottleneckResult[] {\n  const { limit = 20, scope, minFanIn = 2, minFanOut = 2 } = opts;\n  const scopeFilter = scope ? `AND def_d.relative_path LIKE '%${scope}%'` : '';\n\n  // Use a wrapping query to filter on computed columns\n  const rows = db.all<{\n    symbol: string;\n    defined_in: string;\n    fan_in: number;\n    fan_out: number;\n  }>(\n    `SELECT * FROM (\n      SELECT\n        gs.symbol,\n        def_d.relative_path AS defined_in,\n        (SELECT COUNT(DISTINCT ref_c.document_id)\n         FROM mentions ref_m\n         JOIN chunks ref_c ON ref_m.chunk_id = ref_c.id\n         WHERE ref_m.symbol_id = gs.id AND ref_m.role != 1\n        ) AS fan_in,\n        (SELECT COUNT(DISTINCT ref_gs.id)\n         FROM mentions ref_m\n         JOIN chunks ref_c ON ref_m.chunk_id = ref_c.id\n         JOIN global_symbols ref_gs ON ref_m.symbol_id = ref_gs.id\n         JOIN defn_enclosing_ranges ref_der ON ref_gs.id = ref_der.symbol_id\n         WHERE ref_c.document_id = def_d.id\n           AND ref_m.role != 1\n           AND ref_der.document_id != def_d.id\n        ) AS fan_out\n      FROM global_symbols gs\n      JOIN defn_enclosing_ranges der ON gs.id = der.symbol_id\n      JOIN documents def_d ON der.document_id = def_d.id\n      WHERE 1 = 1\n        ${db.pathExclusionsFor('def_d')}\n        ${db.symbolNoiseFor('gs')}\n        ${scopeFilter}\n    ) WHERE fan_in >= ? AND fan_out >= ?\n    ORDER BY (fan_in * fan_out) DESC\n    LIMIT ?`,\n    minFanIn, minFanOut, limit,\n  );\n\n  return rows\n    .filter((r) => !db.isIgnored(r.defined_in))\n    .map((r) => ({\n      symbol: r.symbol,\n      shortName: shortenSymbol(r.symbol),\n      fanIn: r.fan_in,\n      fanOut: r.fan_out,\n      score: r.fan_in * r.fan_out,\n      definedIn: r.defined_in,\n    }));\n}\n"],"mappings":";;;;;AAWO,SAAS,YACd,IACA,OAAkF,CAAC,GAC/D;AACpB,QAAM,EAAE,QAAQ,IAAI,OAAO,WAAW,GAAG,YAAY,EAAE,IAAI;AAC3D,QAAM,cAAc,QAAQ,kCAAkC,KAAK,OAAO;AAG1E,QAAM,OAAO,GAAG;AAAA,IAMd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,UAsBM,GAAG,kBAAkB,OAAO,CAAC;AAAA,UAC7B,GAAG,eAAe,IAAI,CAAC;AAAA,UACvB,WAAW;AAAA;AAAA;AAAA;AAAA,IAIjB;AAAA,IAAU;AAAA,IAAW;AAAA,EACvB;AAEA,SAAO,KACJ,OAAO,CAAC,MAAM,CAAC,GAAG,UAAU,EAAE,UAAU,CAAC,EACzC,IAAI,CAAC,OAAO;AAAA,IACX,QAAQ,EAAE;AAAA,IACV,WAAW,cAAc,EAAE,MAAM;AAAA,IACjC,OAAO,EAAE;AAAA,IACT,QAAQ,EAAE;AAAA,IACV,OAAO,EAAE,SAAS,EAAE;AAAA,IACpB,WAAW,EAAE;AAAA,EACf,EAAE;AACN;","names":[]}