Mises à jour

Changelog

Toutes les versions, fonctionnalités et corrections.

Dernière

v0.6.0

Versions

Modifications

149

Release History

22 versions

0.6.0Current18 févr. 2026

HTTP API Server & Local Import Resolution

New HTTP API server alternative to MCP stdio, local import path resolution, Python stdlib detection, and symbol parent tracking.

FonctionnalitéHTTP API server — Alternative to MCP stdio with `--http-port`, endpoints: `/health`, `/stats`, `/search`, `/find-refs`, `/deps`, `/explain`

FonctionnalitéHTTP middleware — 1MB body limit, 50 concurrent requests, CORS configurable via `--cors-origin`

FonctionnalitéLocal import resolution — Resolution of local import paths to actual files on disk (JS/TS, Python, Rust, Go)

Fonctionnalité`resolved_path` column — Dependencies now store the resolved path, improving `find_refs` accuracy

FonctionnalitéPython stdlib detection — Accurate classification of Python standard vs external imports (200+ modules, binary search)

FonctionnalitéSymbol parent tracking — Symbols now include their parent (e.g., method → struct/class)

FonctionnalitéDockerfile — Multi-stage Docker image for deployment (Railway-ready)

AméliorationSchema migration v3→v4 — Automatic incremental migration (adds `resolved_path` column, no reindex required)

AméliorationAuto-indexer and CLI `index` command use local import resolution

CorrectionCorrect git clone URL in Dockerfile

CorrectionResolve clippy `module_inception` warning in HTTP tests

CorrectionBump Rust version in Dockerfile to support edition 2024 and let-chains

0.5.210 févr. 2026

Security Hardening

Major security release with fixes for ReDoS, path traversal, and input validation vulnerabilities across the codebase.

Correction[HIGH] Fixed ReDoS vulnerability - user input is now escaped with `regex::escape()` in `TextSearcher::search()` before regex compilation

Correction[HIGH] Updated `bytes` crate 1.11.0 → 1.11.1 to fix integer overflow in `BytesMut::reserve` (RUSTSEC-2026-0007)

Correction[HIGH] Text search walker now uses `hidden(true)` and `should_exclude_entry` filtering, preventing reads from `.env`, `.git/`, and other sensitive directories

Correction[MEDIUM] Fixed path traversal in `read_file_lines()` - paths are now canonicalized and verified to stay within the project root

Correction[MEDIUM] Added input validation (empty, length ≤ 500, limit ≤ 1000) to `semantiq_find_refs`, `semantiq_explain`, and `semantiq_deps` MCP handlers

Correction[MEDIUM] Added path traversal rejection (`..`) in `semantiq_deps` file path parameter

Correction[MEDIUM] `resolve_project_root()` now canonicalizes paths to normalize `..` components and symlinks

Correction[LOW] FTS5 query escaping now strips null bytes and control characters

Correction[LOW] Query expansion limited to 10 terms to prevent amplification attacks

Correction[LOW] MCP error messages sanitized to avoid leaking internal file paths

Correction[LOW] Version check HTTP response limited to 10KB to prevent memory exhaustion

Correction[LOW] Poisoned mutex recovery in `DistanceCollector` now logs warnings instead of silently continuing

0.5.031 janv. 2026

Adaptive ML Thresholds

Automatic calibration of semantic search thresholds per programming language, new CLI commands, and codebase refactoring.

FonctionnalitéAdaptive ML thresholds - automatic calibration of semantic search thresholds per programming language

FonctionnalitéBootstrap mode - collects 100% of distance observations until 500 samples

FonctionnalitéProduction mode - switches to 10% sampling after bootstrap completes

FonctionnalitéAuto-calibration - triggers automatically when bootstrap completes

FonctionnalitéPercentile-based thresholds - uses p90 for max_distance, p10 for min_similarity

FonctionnalitéPer-language calibration with fallback cascade (language → global → defaults)

FonctionnalitéNew `calibrate` CLI command - manual threshold calibration with `--dry-run` option

FonctionnalitéML stats in `stats` command - shows bootstrap progress, observations per language, calibrated thresholds

FonctionnalitéNew database tables - `distance_observations` and `threshold_calibration` for ML data

FonctionnalitéCI workflows for `dev` branch - tests, Clippy, format checks, and multi-platform builds

AméliorationRefactored `store.rs` (2108 lines → 8 modules) - better code organization with dedicated modules for files, symbols, chunks, dependencies, observations, calibrations, and tests

AméliorationRefactored `engine.rs` (1049 lines → 5 modules) - cleaner architecture with dedicated modules for search, threshold, analysis, and tests

BreakingSchema version bumped to 3 - triggers automatic reindex on upgrade

0.4.028 janv. 2026

JSON Logging & MCP Tests

Structured JSON logging throughout the codebase, MCP test coverage, and CI/security workflows.

FonctionnalitéJSON logging support - structured logging throughout the codebase

FonctionnalitéJSON logging by default for `serve` command - better integration with log aggregators

FonctionnalitéMCP tests - test coverage for MCP server functionality

FonctionnalitéCI and security workflows - automated testing and security scanning

Amélioration`init-cursor` command is now language-agnostic - works with any project type

AméliorationUpdated `deny.toml` to v2 schema

0.3.420 janv. 2026

macOS Intel Support & Optional ONNX

Restored macOS Intel (x86_64) support and made ONNX feature optional for broader platform compatibility.

FonctionnalitémacOS Intel (x86_64-apple-darwin) support restored - binary now available for Intel Macs

FonctionnalitéCI build workflow - new build.yml for testing builds on push/PR without publishing

AméliorationONNX feature now optional - `--features onnx` required on supported platforms (Apple Silicon, Linux, Windows)

AméliorationIntel Mac builds use StubEmbeddingModel (no ONNX) due to missing prebuilt binaries

AméliorationUpdated CI to use macos-15 runner for Intel Mac cross-compilation

0.3.319 janv. 2026

Search Filtering Options

New search filtering capabilities for semantiq_search with min_score, file_type, and symbol_kind parameters.

Fonctionnalité`min_score` parameter for `semantiq_search` - minimum relevance score threshold (0.0-1.0, default: 0.35)

Fonctionnalité`file_type` parameter for `semantiq_search` - filter results by file extension

Fonctionnalité`symbol_kind` parameter for `semantiq_search` - filter by symbol type (function, class, etc.)

FonctionnalitéCLI flags `--min-score`, `--file-type`, `--symbol-kind` for `semantiq search` command

FonctionnalitéSmart default exclusions - automatically excludes non-code files from search

Fonctionnalité`SearchOptions` struct in `semantiq-retrieval` with builder pattern for flexible configuration

0.3.219 janv. 2026

Gitignore Support in init-cursor

The init-cursor command now automatically handles .gitignore entries for Semantiq database files.

Fonctionnalité`.gitignore` support in `init-cursor` - automatically adds Semantiq database entries

FonctionnalitéCreates `.gitignore` if not present

FonctionnalitéUpdates existing `.gitignore` preserving original content

FonctionnalitéSkips if entries already present (no duplication)

Fonctionnalité3 new tests for `.gitignore` handling in `init_cursor.rs`

0.3.119 janv. 2026

Cursor/VS Code Init Command

New init-cursor command for automatic Cursor/VS Code configuration setup, plus centralized CLI utilities.

FonctionnalitéNew `init-cursor` command for Cursor/VS Code configuration setup

FonctionnalitéCreates `.cursor/rules/project.mdc` (general project guidelines)

FonctionnalitéCreates `.cursor/rules/semantiq.mdc` (Semantiq MCP tools usage)

FonctionnalitéCreates `.cursor/mcp.json` (MCP server configuration)

FonctionnalitéCreates `.cursorignore` (indexing exclusions)

FonctionnalitéCreates `.vscode/` config (settings, tasks, launch, extensions)

FonctionnalitéPreserves existing files (skip instead of overwrite)

AméliorationCentralized `DEFAULT_DB_NAME` and path resolution utilities in `common.rs`

AméliorationRefactored all CLI commands to use shared utilities

AméliorationCLI description now generic ("for a project" instead of "for a Rust project")

Fonctionnalité7 new unit tests for `common.rs` and `init_cursor.rs`

0.3.019 janv. 2026

Vector Search & Auto-Indexing

sqlite-vec integration for semantic vector search, automatic initial indexing, and 6 new languages.

Fonctionnalitésqlite-vec integration for vector similarity search (384-dim MiniLM-L6-v2 embeddings)

FonctionnalitéAutomatic initial indexing when MCP server starts (no more manual `semantiq index` required)

Fonctionnalité6 new languages: HTML, JSON, YAML, TOML, Bash, Elixir (total: 15 languages)

Fonctionnalitéripgrep integration for fast regex text search via `TextSearcher`

FonctionnalitéNew `search_similar_chunks()` method for semantic vector search

FonctionnalitéNew `InitialIndexResult` struct for tracking initial indexing progress

Correction"Imported by" always empty in `semantiq_deps` - rewrote `get_dependents()` to match JS/TS import paths

CorrectionImport path resolution now handles basename matching with multiple extensions

AméliorationSchema version bumped to 2 (triggers automatic reindex)

AméliorationAdded `chunks_vec` virtual table for sqlite-vec embeddings

Amélioration`start_auto_indexer()` now runs `initial_index()` before watching for changes

AméliorationImproved dependency matching with multiple LIKE patterns and post-filtering

0.2.919 janv. 2026

Arrow Function Indexing

Improved TypeScript/JavaScript function detection - arrow functions and function expressions are now correctly classified as functions.

CorrectionArrow functions (const fn = () => {}) now correctly indexed as function instead of variable

CorrectionFunction expressions (const fn = function() {}) now correctly indexed as function

AméliorationAdded is_function_variable() helper to detect functions assigned to variables

AméliorationAdded arrow_function and lexical_declaration to chunk boundaries for TypeScript/JavaScript

AméliorationBumped PARSER_VERSION to 3 (triggers automatic reindex)

0.2.818 janv. 2026

Security Hardening

Critical security improvements including checksum verification, path traversal protection, and memory exhaustion prevention.

CorrectionCRITICAL: Added SHA-256 checksum verification for ONNX model downloads (TOFU + hardcoded support)

CorrectionCRITICAL: Added path traversal protection with canonicalization in validate_path()

CorrectionHIGH: Added MAX_AST_DEPTH=500 recursion limit in parser to prevent stack overflow attacks

CorrectionHIGH: Added safe_slice() function to prevent panic on invalid byte indices

CorrectionHIGH: Changed model directory fallback from "." to system temp dir (prevents writes to unexpected locations)

CorrectionHIGH: Added pagination for get_chunks_with_embeddings() to prevent memory exhaustion DoS

CorrectionHIGH: Reduced download size limit from 500MB to 100MB

CorrectionHIGH: Added restrictive file permissions (0600 on Unix) for downloaded models and database

CorrectionMEDIUM: Added explicit symlink handling (follow_links(false)) to prevent escape from project root

AméliorationRefactored download_file() with connection timeouts (30s connect, 5min global)

AméliorationImproved checksum verification with detailed warning messages

0.2.718 janv. 2026

Update Notifications

Automatic version update notifications at server startup with 24h local caching.

FonctionnalitéAutomatic version update notification at server startup

FonctionnalitéNon-blocking background check using GitHub Releases API

FonctionnalitéLocal cache (24h) to avoid repeated API calls

Fonctionnalité--no-update-check CLI flag to disable update notifications

FonctionnalitéSEMANTIQ_UPDATE_CHECK environment variable for configuration

AméliorationUpdated author info to keyldzn

0.2.618 janv. 2026

Automatic Reindexation

Automatic reindexation when parser version changes, plus improved TypeScript/JavaScript support.

FonctionnalitéAutomatic reindexation when parser version changes (no more manual --force needed)

FonctionnalitéPARSER_VERSION constant to track parser logic changes

FonctionnalitéSupport for const/let variable extraction in TypeScript/JavaScript

FonctionnalitéGitHub Sponsors funding configuration

AméliorationVersion detection uses atomic transactions to prevent race conditions

AméliorationDocumentation updated with known limitations and setup guides

CorrectionFilter out verbose ONNX Runtime logs during indexing

0.2.418 janv. 2026

ONNX Stability Fixes

Critical fixes for ONNX model download and inference, plus improved file exclusion handling.

CorrectionModel download failing in async Tokio context (replaced reqwest::blocking with ureq)

CorrectionDownload size limit too small for 90MB ONNX model (increased to 200MB)

CorrectionONNX inference crash due to missing token_type_ids input

CorrectionEmbeddings not generated during semantiq index command

Améliorationsemantiq index now generates embeddings for all chunks

AméliorationCentralized file exclusion logic into exclusions.rs module

AméliorationAuto-indexer and FileWatcher now use shared exclusion patterns

0.2.318 janv. 2026

Semantic Search with ONNX

Semantic code search via ONNX embeddings, automatic model download, and cosine similarity matching.

FonctionnalitéONNX embedding model integration for semantic search

FonctionnalitéAutomatic model download on first run

FonctionnalitéCosine similarity search for vector matching

FonctionnalitéAlternative installation via cargo install --git

FonctionnalitéCHANGELOG.md for version history

AméliorationEmbeddings now generated automatically during auto-indexing

AméliorationSwitch from OpenSSL to rustls for better cross-compilation support

AméliorationUse ort download-binaries for automatic ONNX Runtime provisioning

BreakingRemoved macOS Intel (x86_64-apple-darwin) binary - ONNX Runtime does not support this target

0.2.217 janv. 2026

CLAUDE.md Improvements

Updated CLAUDE.md template to prioritize Semantiq MCP tools.

AméliorationImproved CLAUDE.md template to prioritize Semantiq tools over grep/Glob

0.2.117 janv. 2026

Performance & Security Fixes

Critical fixes for error handling, security, and query performance.

CorrectionError handling with proper mutex propagation

CorrectionSQL injection vulnerability via LIKE escaping

CorrectionUTF-8 safety in tree-sitter text extraction

CorrectionN+1 query pattern in get_stats() (4 queries → 1)

AméliorationShared single Arc<IndexStore> instead of 3 separate DB connections

AméliorationImproved scoring algorithm with symbol type boosting

AméliorationResults limited to 500 to prevent memory issues

AméliorationAdded PRAGMA busy_timeout=5000 for concurrent access

0.2.017 janv. 2026

NPM Package Automation

Automatic npm package version synchronization from git tags.

FonctionnalitéAutomatic npm package version update from git tag

0.1.317 janv. 2026

Project Init Command

New init command for project setup with automatic configuration.

FonctionnalitéNew semantiq init command for easy project setup

FonctionnalitéAuto-creates .claude/settings.json, CLAUDE.md, updates .gitignore

FonctionnalitéRuns initial indexation automatically

0.1.217 janv. 2026

Auto-Indexing

Real-time file updates with automatic re-indexing on changes.

FonctionnalitéAuto-indexing for real-time file updates

FonctionnalitéFileWatcher integration with create/modify/delete events

FonctionnalitéBackground task with 2-second polling

0.1.117 janv. 2026

Documentation Update

Improved npm and project documentation.

Fonctionnaliténpm README documentation

AméliorationUpdated main README with correct npm package name

0.1.017 janv. 2026

Initial Release

First public release of Semantiq with core MCP functionality.

FonctionnalitéInitial release

FonctionnalitéMCP server with 4 tools: search, find_refs, deps, explain

FonctionnalitéSupport for 9 languages via tree-sitter

FonctionnalitéSQLite storage with FTS5 search