This repository was archived by the owner on Mar 13, 2026. It is now read-only.
Releases: LeahArmstrong/codebase_index
Releases · LeahArmstrong/codebase_index
v0.3.2
Immutable
release. Only release title and notes can be modified.
Bug Fixes
- Per-association error resilience — A single broken association (e.g. missing target class) no longer aborts the entire model. Warnings are collected and surfaced in the extraction summary. (#24)
- Dynamic concern directory discovery —
ConcernExtractornow globsapp/**/concerns/to find all concern directories, including deeply nested ones. (#25) - JsonSnapshotStore fallback — On MySQL-only hosts where
sqlite3isn't in the Gemfile, temporal snapshots now fall back to JSON files instead of silently failing. - MCP parameter type coercion — String values from MCP clients are now coerced to integers where expected, preventing handler crashes.
extenddependency tracking —extended modules are now tracked as dependencies in model and concern extractors (previously onlyincludewas detected).- Array-form WHERE support —
EmbeddedExecutornow handles array-form WHERE clauses for JSON column queries. - Configurable embedded sql/query tools — Read-only sql/query tools in embedded mode are now opt-in via
read_tools_enabled, with SqlValidator + SafeContext defense-in-depth. - Notion sync columns metadata —
ModelExtractornow writes richcolumnsmetadata (name/type/null/default) soNotion::Exporter#sync_model_columnsactually has data to sync.
Full Changelog: v0.3.1...v0.3.2
v0.3.1
Immutable
release. Only release title and notes can be modified.
Full Changelog: v0.3.0...v0.3.1
v0.3.0
Immutable
release. Only release title and notes can be modified.
Highlights
- Redis/SolidCache caching layer for retrieval pipeline
- Engine classification — framework vs application tagging
- 10 new docs — Context7-optimized FAQ, Troubleshooting, Architecture, Extractor Reference, MCP Cookbook, and more
- README streamlined from 620 to 325 lines
- Bug fixes — vendor path leak across 9 extractors, Prism compat, schema_sha for structure.sql
What's Changed
Added
- Redis/SolidCache caching layer with TTL, namespace isolation, and nil-caching
- Engine classification (
:framework/:application) based on install path - Graph analysis staleness tracking (
generated_at+graph_sha) - Docker setup guide (
docs/DOCKER_SETUP.md) - Context7 documentation suite (10 new docs +
context7.json)
Fixed
- Vendor path leak in source file resolution across 9 extractors
- Prism cross-version compatibility
schema_shasupportsdb/structure.sqlfallback- ViewComponent extractor skips framework internals with no source
- HTTP connection reuse and retry in embedding providers
- DependencyGraph
to_hcache pollution - MCP tool counts corrected (27 index / 31 console)
Changed
- README streamlined from 620 to 325 lines
- Internal rake tasks hidden from
rails -T - Simplification sweep — dead code removal, shared helpers
Performance
- Critical hotspots fixed across extraction, storage, and retrieval
Full Changelog: v0.2.1...v0.3.0
v0.2.1
Immutable
release. Only release title and notes can be modified.
Full Changelog: v0.2.0...v0.2.1
v0.2.0
Immutable
release. Only release title and notes can be modified.
What's Changed
- Add CODEOWNERS, issue routing, and improved PR template by @LeahArmstrong in #15
- Fix duplicate URI warning in gemspec by @LeahArmstrong in #16
- Add embedded console MCP server by @LeahArmstrong in #17
Full Changelog: v0.1.0...v0.2.0
v0.1.0 — Initial Release
Immutable
release. Only release title and notes can be modified.
CodebaseIndex v0.1.0
Rails codebase extraction and indexing for AI-assisted development. Uses runtime introspection to produce version-accurate representations with inlined concerns, resolved callback chains, and schema-aware associations.
Highlights
- 34 extractors covering models, controllers, services, jobs, mailers, routes, middleware, GraphQL, Pundit, state machines, events, caching, factories, and more
- 2 MCP servers — 27-tool index server (reads pre-extracted JSON) + 31-tool console server (bridges to live Rails)
- Full retrieval pipeline — query classification, hybrid search, RRF ranking, token-budgeted context assembly
- Backend agnostic — pgvector, Qdrant, or SQLite for vectors; MySQL or PostgreSQL for metadata; OpenAI or Ollama for embeddings
- Behavioral depth — callback side-effect analysis, resolved Rails config introspection, optional pre-computed request flow maps
- Notion export — sync extraction data to Notion databases
Getting Started
gem 'codebase_index'See Getting Started for the full walkthrough.