devlensio 1.0.1 → 1.0.3

package/dist/pipeline/index.d.ts

@@ -46,6 +46,7 @@ export interface PipelineResult {
     isGithubRepo: boolean;
     gitInfo: GitInfo;
 }
+export declare function worktreeCommitHash(nodes: CodeNode[]): string;
 export declare function routesToCodeNodes(routes: (RouteNode | BackendRouteNode)[], repoPath: string): CodeNode[];
 export declare function analyzePipeline(repoPath: string, isGithubRepo: boolean, options?: PipelineOptions): Promise<PipelineResult>;
 export declare function refilterPipeline(stored: PipelineResult, thresholds: FilterThresholds): Pick<PipelineResult, "nodes" | "edges" | "stats">;

package/dist/pipeline/index.js

@@ -15,6 +15,12 @@ function generateGraphId(repoPath, isGithubRepo) {
         .digest("hex")
         .slice(0, 16);
 }
+// Stable id for repos without git: derived from node ids + code hashes, so the
+// same content overwrites one snapshot in place instead of accumulating timestamps
+export function worktreeCommitHash(nodes) {
+    const parts = nodes.map((n) => `${n.id}:${n.codeHash ?? ""}`).sort();
+    return "worktree-" + createHash("sha256").update(parts.join("|")).digest("hex").slice(0, 12);
+}
 // Gets current git state of the repo
 // Falls back gracefully if git is not initialized
 function getGitInfo(repoPath) {
@@ -28,9 +34,11 @@ function getGitInfo(repoPath) {
         return { commitHash, branch, message, hasGit: true };
     }
     catch {
-        // No git, or no commits yet — use timestamp as version key
+        // No git, or no commits yet. Placeholder resolved to a content-derived
+        // id after extraction (see worktreeCommitHash) so unchanged repos reuse
+        // one snapshot instead of growing a new timestamped one per analyze
         return {
-            commitHash: Date.now().toString(),
+            commitHash: "worktree-pending",
             branch: "unknown",
             message: "no git history",
             hasGit: false,
@@ -221,6 +229,10 @@ export async function analyzePipeline(repoPath, isGithubRepo, options) {
     }
     const allNodes = extractorResult.nodes;
     const allEdges = extractorResult.edges;
+    if (!gitInfo.hasGit) {
+        gitInfo.commitHash = worktreeCommitHash(allNodes);
+        gitInfo.message = "worktree snapshot";
+    }
     //  Step 5: Score and filter 
     console.log("\n[5/5] Scoring and filtering...");
     const scoringResult = scoreAndFilter(allNodes, allEdges, options?.thresholds);

package/dist/storage/fileStorage.d.ts

@@ -21,6 +21,7 @@ export interface CommitSummary {
     edgeCount: number;
     hasGit: boolean;
     isSummarized?: boolean;
+    isIndexed?: boolean;
 }
 export interface GraphMeta {
     graphId: string;
@@ -101,6 +102,8 @@ export declare function deleteCommit(graphId: string, commitHash: string): boole
 export declare function diffCommits(graphId: string, fromHash: string, toHash: string): NodeDiff | undefined;
 export declare function markCommitSummarized(graphId: string, commitHash: string): void;
 export declare function isCommitSummarized(graphId: string, commitHash: string): boolean;
+export declare function markCommitIndexed(graphId: string, commitHash: string): void;
+export declare function isCommitIndexed(graphId: string, commitHash: string): boolean;
 export declare function findLastSummarizedAncestor(graphId: string, commitHash: string, repoPath: string): Promise<string | undefined>;
 export declare function saveNodeSummaries(graphId: string, commitHash: string, nodeUpdates: Map<string, {
     technicalSummary: string;

package/dist/storage/fileStorage.js

@@ -182,9 +182,14 @@ export function saveGraph(result, options) {
         commits: [],
         summarizedCommits: [],
     };
-    // Replace if same commit already exists, else append
+    // Replace if same commit already exists, else append.
+    // Carry over flags the rebuilt summary does not know about (same hazard as
+    // isSummarized: buildCommitSummary starts from scratch)
     const existingCommit = meta.commits.findIndex((c) => c.commitHash === commitHash);
     if (existingCommit >= 0) {
+        const prev = meta.commits[existingCommit];
+        if (prev.isIndexed)
+            newSummary.isIndexed = true;
         meta.commits[existingCommit] = newSummary;
     }
     else {
@@ -500,6 +505,23 @@ export function isCommitSummarized(graphId, commitHash) {
         return false;
     return meta.summarizedCommits?.includes(commitHash);
 }
+// Marks a commit as having a derived .search.json index
+export function markCommitIndexed(graphId, commitHash) {
+    const meta = readMeta(graphId);
+    if (!meta)
+        return;
+    const commit = meta.commits.find((c) => c.commitHash === commitHash);
+    if (!commit || commit.isIndexed)
+        return;
+    commit.isIndexed = true;
+    writeMeta(meta);
+}
+export function isCommitIndexed(graphId, commitHash) {
+    const meta = readMeta(graphId);
+    if (!meta)
+        return false;
+    return meta.commits.find((c) => c.commitHash === commitHash)?.isIndexed === true;
+}
 // Finds the most recent ancestor commit that has been summarized.
 // Uses simple-git to walk the commit history of the repo.
 // Returns undefined if no summarized ancestor exists (= full summarization needed).
@@ -601,6 +623,8 @@ export const fileStorage = {
     diffCommits,
     markCommitSummarized,
     isCommitSummarized,
+    markCommitIndexed,
+    isCommitIndexed,
     findLastSummarizedAncestor,
     saveNodeSummaries,
     getCheckpointPath,

package/dist/storage/interface.d.ts

@@ -11,6 +11,8 @@ export interface GraphStorage {
     diffCommits(graphId: string, fromHash: string, toHash: string): NodeDiff | undefined;
     markCommitSummarized(graphId: string, commitHash: string): void;
     isCommitSummarized(graphId: string, commitHash: string): boolean;
+    markCommitIndexed(graphId: string, commitHash: string): void;
+    isCommitIndexed(graphId: string, commitHash: string): boolean;
     findLastSummarizedAncestor(graphId: string, commitHash: string, repoPath: string): Promise<string | undefined>;
     saveNodeSummaries(graphId: string, commitHash: string, nodeUpdates: Map<string, {
         technicalSummary: string;

package/extractors/rust/src/calls.rs

@@ -41,7 +41,7 @@ pub fn detect_calls(
     let mut edges = vec![];
 
     for pf in &repo.files {
-        if pf.is_test_file {
+        if pf.is_pure_test_file {
             continue;
         }
         for item in &pf.items {

package/extractors/rust/src/enrich.rs

@@ -57,7 +57,7 @@ pub fn enrich_nodes(
     let mut model_table: std::collections::HashMap<String, String> =
         std::collections::HashMap::new();
     for pf in &repo.files {
-        if pf.is_test_file {
+        if pf.is_pure_test_file {
             continue;
         }
         for item in &pf.items {

package/extractors/rust/src/imports.rs

@@ -34,7 +34,7 @@ pub fn detect_imports(
     let mut seen: std::collections::HashSet<(String, String)> = std::collections::HashSet::new();
 
     for pf in &repo.files {
-        if pf.is_test_file {
+        if pf.is_pure_test_file {
             // test files ARE leaf nodes, but they still import — emit edges
             // from the TEST node (same file id — parentFile covers it)
         }

package/extractors/rust/src/inheritance.rs

@@ -37,7 +37,7 @@ pub fn detect_inheritance(
     let mut edges = vec![];
     let mut nodes = vec![];
     for pf in &repo.files {
-        if pf.is_test_file {
+        if pf.is_pure_test_file {
             continue;
         }
         for item in &pf.items {

package/extractors/rust/src/nodes.rs

@@ -2,9 +2,11 @@
 //
 // Contract rules enforced here:
 //   - FILE id `file::rel/path.rs`; file nodes parent themselves
-//   - TEST files (any #[test] fn) = LEAF nodes: nodes=[] for them, test fns
-//     ride in metadata.testCases only (the bug that shipped in Python's
-//     parser — asserted in the bun suite)
+//   - PURE test files (tests/ integration files, or files whose every item is
+//     a #[test] fn) = LEAF TEST nodes; Rust's idiomatic inline #[cfg(test)]
+//     mod tests must NOT collapse a mixed production file — its non-test items
+//     emit normally and #[test] fns emit as FUNCTION nodes flagged
+//     metadata.isTestFunction (test bodies stay out of prod call edges there)
 //   - rawCode + codeHash on every FUNCTION/METHOD/STRUCT/ENUM/TRAIT/IMPL_BLOCK
 //   - deterministic ids (file-scoped; trait-impl method suffix for the
 //     same-name-different-trait collision class)
@@ -17,7 +19,7 @@ use serde_json::{Map, Value};
 pub fn collect_file_nodes(files: &[ParsedFile]) -> Vec<CodeNode> {
     let mut out = vec![];
     for pf in files {
-        if pf.is_test_file {
+        if pf.is_pure_test_file {
             let mut n = file_node(&pf.rel_path, pf.line_count, NODE_TEST);
             let meta = n
                 .metadata
@@ -30,25 +32,43 @@ pub fn collect_file_nodes(files: &[ParsedFile]) -> Vec<CodeNode> {
             }
             out.push(n);
         } else {
-            out.push(file_node(&pf.rel_path, pf.line_count, NODE_FILE));
+            let mut n = file_node(&pf.rel_path, pf.line_count, NODE_FILE);
+            // Mixed files keep a testCases index so consumers that never walk
+            // FUNCTION nodes still see the file's tests.
+            if !pf.test_fns.is_empty() {
+                let meta = n
+                    .metadata
+                    .entry("testCases".to_string())
+                    .or_insert_with(|| Value::Array(vec![]));
+                if let Value::Array(arr) = meta {
+                    for t in &pf.test_fns {
+                        arr.push(Value::from(t.clone()));
+                    }
+                }
+            }
+            out.push(n);
         }
     }
     out
 }
 
-/// Code nodes (FUNCTION/METHOD/STRUCT/ENUM/TRAIT/IMPL_BLOCK) — test fns are
-/// NOT emitted (leaf rule); they feed metadata.testCases only.
+/// Code nodes. PURE test files stay leaves (their fns ride in testCases).
+/// Mixed files emit ALL items; #[test] fns are flagged metadata.isTestFunction
+/// instead of being dropped, so a test function is addressable without a new
+/// node type.
 pub fn collect_code_nodes(files: &[ParsedFile]) -> Vec<CodeNode> {
     let mut out = vec![];
     for pf in files {
-        if pf.is_test_file {
+        if pf.is_pure_test_file {
             continue;
         }
         for item in &pf.items {
+            let mut n = item_node(pf, item);
             if item.is_test {
-                continue;
+                n.metadata
+                    .insert("isTestFunction".to_string(), Value::from(true));
             }
-            out.push(item_node(pf, item));
+            out.push(n);
         }
     }
     out

package/extractors/rust/src/orm_edges.rs

@@ -21,7 +21,7 @@ use std::collections::HashMap;
 pub fn table_to_model_map(repo: &ParsedRepo) -> HashMap<String, String> {
     let mut map = HashMap::new();
     for pf in &repo.files {
-        if pf.is_test_file {
+        if pf.is_pure_test_file {
             continue;
         }
         for item in &pf.items {
@@ -89,7 +89,7 @@ pub fn detect_orm_edges(
     let table_to_schema_file = table_to_schema_file_map(repo);
 
     for pf in &repo.files {
-        if pf.is_test_file {
+        if pf.is_pure_test_file {
             continue;
         }
         for item in &pf.items {

package/extractors/rust/src/parser.rs

@@ -162,7 +162,12 @@ pub struct ParsedFile {
     pub items: Vec<ParsedItem>,
     pub tables: Vec<TableDecl>,
     pub test_fns: Vec<String>,
-    pub is_test_file: bool,
+    /// True only when the file is EXCLUSIVELY tests (every top-level item is a
+    /// #[test] fn, or the file only holds inline #[cfg(test)] mods). Rust's
+    /// idiomatic inline `#[cfg(test)] mod tests` must NOT collapse a mixed
+    /// production file into a TEST leaf — only pure test files get leaf
+    /// treatment (mirrors the intent of go's _test.go rule).
+    pub is_pure_test_file: bool,
     pub line_count: i64,
 }
 
@@ -326,7 +331,7 @@ pub fn parse_file(rel_path: &str, source: &str) -> Result<ParsedFile, String> {
         items: vec![],
         tables: vec![],
         test_fns: vec![],
-        is_test_file: false,
+        is_pure_test_file: false,
         line_count: line_index.line_count() as i64,
     };
 
@@ -342,7 +347,7 @@ pub fn parse_file(rel_path: &str, source: &str) -> Result<ParsedFile, String> {
     }
     pf.ast.items = ast_items;
 
-    pf.is_test_file = !pf.test_fns.is_empty();
+    pf.is_pure_test_file = !pf.test_fns.is_empty() && pf.items.iter().all(|i| i.is_test);
     Ok(pf)
 }
 

package/extractors/rust/src/routes.rs

@@ -96,7 +96,7 @@ pub fn detect_routes(
     let mut func_lit_vars: HashMap<(String, String), String> = HashMap::new();
 
     for pf in &repo.files {
-        if pf.is_test_file {
+        if pf.is_pure_test_file {
             continue;
         }
         for item in &pf.items {
@@ -195,7 +195,7 @@ pub fn detect_routes(
     // ── actix-web / rocket: attribute macros ──
     if framework == "actix-web" || framework == "rocket" {
         for pf in &repo.files {
-            if pf.is_test_file {
+            if pf.is_pure_test_file {
                 continue;
             }
             for item in &pf.items {

package/extractors/rust/src/tests.rs

@@ -22,7 +22,7 @@ pub fn detect_tests(
 ) -> Vec<CodeEdge> {
     let mut edges = vec![];
     for pf in &repo.files {
-        if !pf.is_test_file {
+        if !pf.is_pure_test_file {
             continue;
         }
         let from = format!("file::{}", pf.rel_path);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "devlensio",
-  "version": "1.0.1",
+  "version": "1.0.3",
   "description": "Codebase intelligence engine for TypeScript/JavaScript/React/Next.js, Python, Java, Go, and Rust repositories.",
   "type": "module",
   "main": "./dist/index.js",