All notable changes to ModelMesh are documented here.
- 21-question FAQ covering all major library features (middleware, error handling, proxy, storage, observability, streaming, discovery, multi-pool, hot-reload, browser usage)
- Programmatic connector registration — all 6 connector types support
"instance"injection via API register_connector()function for runtime registration of custom connector classes- Budget-aware rotation —
on_budget_exceeded: rotatepool config option - CONTRIBUTING.md — contributor setup guide
- Makefile — unified task runner (
make install,make test,make lint,make docker-up) modelmesh.example.yaml— annotated config template for Docker/YAML setup- Root
package.jsonwith npm workspaces for monorepo support - Samples
package.json+tsconfig.json— TypeScript samples can now be run directly .editorconfig— consistent formatting across editorsscripts/bump-version.sh— atomic version sync between Python and TypeScript- Docker healthcheck in Dockerfile and docker-compose.yaml
- GitHub Issue Templates — structured bug report and feature request forms
- GitHub Pull Request Template — checklist for tests, docs, changelog
- SECURITY.md — vulnerability disclosure policy
- VS Code Dev Container (
.devcontainer/) — one-click dev environment - Dependabot — automated dependency updates for Python, npm, Docker, Actions
- CodeQL security scanning — weekly + PR security analysis
- Test coverage reporting — pytest-cov + Jest coverage → Codecov CI workflow
- Prettier + ESLint — TypeScript code formatting and linting
- mypy type checking — Python type checking in CI
- TypeDoc configuration — auto-generated TypeScript API docs
- Troubleshooting Guide (
docs/guides/Troubleshooting.md) - npm publish workflow — GitHub Actions → npm with provenance
- PyPI publish workflow — GitHub Actions → PyPI with trusted publishing
- GitHub Release workflow — auto-create releases from tags with assets
- Docker multi-platform build — linux/amd64 + linux/arm64 images
- Integration test suite (
tests/integration/) — end-to-end routing tests - Performance benchmark suite (
tests/benchmarks/) — routing latency tests - Pre-commit hooks (
.pre-commit-config.yaml) — ruff, prettier, gitleaks - Starter templates (
tools/starter-template/) — Python and TypeScript project starters - License compliance check (
scripts/check-licenses.sh)
- Dockerfile
COPYpath corrected (was referencing root-levelpyproject.toml) install-python.shnow runspip installfrom correct directorytest-all.shnow runsnpm installbefore TypeScript tests.env.exampleclarified — only one provider key required (not all three)- 553 internal doc links fixed (
.html→.mdextension consistency) - Cross-reference audit: all docs have proper "See also" links
- FAQ expanded from 10 to 21 questions
- Docker Compose now includes healthcheck and restart policy
- README expanded with "Running Samples" section and CONTRIBUTING link
- Initial release of ModelMesh Lite
- Python and TypeScript libraries with zero core dependencies
- OpenAI-compatible API interface
- 8 rotation strategies (stick-until-failure, round-robin, cost-first, latency-first, etc.)
- 22 pre-shipped provider connectors
- Capability-based routing with hierarchical taxonomy
- Budget enforcement with daily/monthly limits
- Mock client for testing
- Docker proxy deployment
- 54 pre-shipped connectors across 6 types
- CDK (Connector Development Kit) with base classes
- Comprehensive documentation suite (14 docs + 5 CDK docs + 8 guides)
- 1,879 tests (1,166 Python + 713 TypeScript)