Skip to content

v1.0.5

Choose a tag to compare

@vohongtho vohongtho released this 25 Jul 15:39

Highlights

  • Added plain code-intel analyze auto-incremental mode with sticky embeddings
  • Introduced stable repository IDs, unique naming, and repo management commands
  • Added routed portal settings with authenticated config APIs and admin updates
  • Improved vulnerability scanning precision, analyzer accuracy, and HTML parsing support
  • Tightened CLI startup/version behavior, search relevance, and exclusion controls
  • Reduced dependency risk and refreshed docs, workflows, and tests

What’s new

Auto-incremental analyze and sticky embeddings

  • Plain code-intel analyze now auto-attempts incremental reindexing when prior repo metadata is valid and safety checks pass
  • CLI output now distinguishes auto-incremental runs from full-analysis fallback and explains fallback reasons
  • code-intel analyze --embeddings now remembers semantic-search preference per repository in .code-intel/meta.json
  • Later plain, --incremental, and --force analyze runs auto-enable remembered embeddings unless --skip-embeddings is passed
  • Embedding metadata now tracks readiness, staleness, provider, model, and vector dimensions for automatic rebuild and compatibility checks

Stable repository identity and repo management

  • Persisted repositories now have stable id fields while keeping user-facing unique name values and mutable path values
  • Added migration for legacy registry entries without IDs and deterministic duplicate-name repair
  • code-intel analyze now supports --name <name> with explicit create, rename, and relink semantics
  • Added code-intel repo list|show|rename|relink for repository identity management
  • MCP, HTTP, and group membership flows now resolve repos by stable ID as well as names and paths

Portal settings and config management

  • Added authenticated routed settings pages under /settings/:section with browser back/forward support
  • Added a Settings entry in the portal header profile menu above Sign Out
  • Added masked config read and validated config update APIs at /api/v1/config
  • Portal now supports editing LLM, embeddings, analysis, server, authentication, updates, and telemetry settings
  • Global config reads are available to authenticated viewers; updates require the admin role
  • Logout now clears cached settings state so direct revisits to routed settings URLs re-enter auth cleanly

Vulnerability scan precision and analyzer accuracy

  • Reduced confirmed false positives in security-signal detection, including SSRF, SQL injection, and command-injection cases
  • Improved generic-tier vulnerability scanning with case-insensitive sink matching and ORM raw-query method detection
  • Reduced generic XSS false positives on bare output statements that are not HTML-rendering sinks
  • Added finding tier metadata and confidence scaling for generic-heuristic results
  • Secret scanning now catches common bare and camelCase sensitive names and runs entropy detection independently
  • Coverage-gap detection now recognizes PHPUnit, Python, and Ruby test-file conventions
  • Flow tracing now excludes test and fixture targets from production-entry analysis
  • Health scoring is now normalized by repository size and exposes normalization metadata

Search, parsing, and CLI improvements

  • Added HTML language detection, grammar wiring, and packaged tree-sitter-html.wasm support
  • Natural-language symbol search now filters low-value terms and applies deterministic reranking across names, paths, kinds, and content
  • Search work is bounded for large repos, with deterministic tie-breaking and cache invalidation on graph changes
  • Search output now includes copyable qualified selectors for exact inspect and impact follow-ups
  • inspect <symbol> now reports ambiguity instead of silently choosing the first duplicate match
  • Added --json output for search and inspect
  • Fixed noisy code-intel --version startup behavior, especially on Windows, and added a fast bootstrap path for version-only output

Exclusions, dependencies, docs, and release hardening

  • Added --skip-folders and --skip-files for transient per-run analyze exclusions
  • Added .codeintelignore.local for personal untracked exclusions and expanded .codeintelignore to support file patterns
  • Unified exclusion layers across defaults, ignore files, and CLI flags with cached pattern matching and verbose match reporting
  • Forced @hono/node-server to 2.0.11 in the workspace dependency graph to clear a moderate audit finding inherited via @modelcontextprotocol/sdk
  • Updated README, core README, changelog, package metadata, release validation, and publish workflow behavior
  • Added broad unit, integration, and perf coverage for analyze, settings, search, storage, parsing, security, and registry changes

Documentation

  • Synced README.md and code-intel/core/README.md with v1.0.5 behavior
  • Added changelog coverage for release details and migration notes
  • Refreshed release validation and publish workflow support for the new release path

Suggested short version

1.0.5 adds plain analyze auto-incremental mode, sticky embeddings, stable repo identity, routed portal settings, better scan precision, stronger analyzer accuracy, HTML parsing support, and tighter CLI/search behavior.