Release 1.4.5

Changes

• Python 3.10 test compatibility fixed by using tomli as fallback for tomllib (available from Python 3.11+).
• Model metadata registry made conservative and typed using frozen dataclass; stale entries removed/downgraded, modern model-family inference added.
• Model metadata helpers now exported from package root: ModelMetadata, get_model_info, get_model_metadata, infer_model_metadata.
• Model metadata inference no longer treats every o... model name as OpenAI; only gpt-* and known reasoning prefixes (o1*, o3*, o4*) infer OpenAI.
• README diagrams refactored to use standalone SVG assets in docs/assets/ instead of raw inline SVG for improved GitHub rendering.
• Release workflow validated for trusted publishing; tag/version validation moved after Python setup.
• Validation completed: Python 3.10+ compatibility, linting, type checking, and test suite all pass.

v1.4.0

• Add ensure_model_ready_detailed() function with structured ReadinessResult
  return type for richer readiness diagnostics (success, phase, message, source,
  elapsed_seconds, error). Async wrapper async_ensure_model_ready_detailed()
  also available.
• Make start_service() warmup timeout configurable via warmup_timeout
  parameter (default: 30.0 seconds) and CLI --warmup-timeout argument.
• Document platform-specific installer policies in API reference: macOS prefers
  brew, Linux prefers apt, Windows prefers choco, with script-based fallbacks.
• Refactor ensure_model_ready() to delegate to ensure_model_ready_detailed()
  internally for code reuse and consistency.

v1.2.2

1.2.2\n\n- Add a broader Ollama administration surface for local model operations.\n- Add platform-aware install backend detection with brew, apt, and choco support plus script fallback.\n- Add structured remote catalog, download lifecycle tracking, and explicit model readiness helpers for Ollama.\n- Export and document the new Ollama admin helpers across the package root, API docs, install docs, and usage docs.\n- Add focused unit coverage for the new Ollama administration helpers.\n- Set asyncio_default_fixture_loop_scope = function in pytest.ini to remove the local pytest-asyncio deprecation warning.

v1.2.1

modelito v1.2.1

• Fixed package version fallback so runtime version reporting stays aligned with package metadata.
• Exported estimate_remote_timeout_details and aligned API docs with package exports.
• Corrected OllamaConnector usage docs to match the current constructor signature.
• Resolved mypy typing issue in normalization helpers.
• Normalized changelog/release documentation and marked legacy v1.0.3 release docs as archived historical records.

Validation

• ruff check . passed
• mypy modelito --ignore-missing-imports passed
• pytest -q passed (91 passed, 3 skipped)
• python -m build passed

modelito 1.0.8

Bump to 1.0.8

• Ollama helpers: expose resolve_ollama_command() documented path fallbacks.
• run_ollama_command(..., env=...) and start_detached_ollama_serve(..., env=...) added; helpers ensure PYTHONPATH contains the repository root for Python entrypoints.
• start_service() now returns 0 when it successfully starts the server even if no model is configured.

v1.0.6

1.0.5 - 2026-04-21

• Bump version and release notes placeholder.

1.0.6 - 2026-04-21

• Fix: resolve several stray git-merge conflict markers that caused import-time SyntaxErrors.
• Fix: restore estimate_remote_timeout behavior and include matched_model_override in diagnostic details.
• Docs: update API docs to document with_source optional argument and returned diagnostic tuple.

Changelog

All notable changes to this project will be documented in this file.

1.0.0 - 2026-04-19

• Major release: break compatibility in favor of a cleaner, modern API.
• Add typed Message/Response dataclasses and expanded provider Protocols.
• Streaming stream() surface implemented for OpenAI, Claude, Gemini, Ollama, and Grok.
• SDK-aware integrations for OpenAI (async/streaming/embed), Claude and Gemini.
• Ollama HTTP streaming implemented with robust fallback behavior.
• Integration tests added (gated by environment variables).
• CI/type/lint updates and general cleanup.

0.2.2 - 2026-04-18

• Bump version to 0.2.2.
• All tests passing (24 passed, 1 skipped).

v0.1.3 - 2026-04-18

• Merge fix/ollama-compat: Ollama compatibility improvements (CLI fallback, improved list_models).
• Documentation: update README.md and docs/USAGE.md with clearer overview and quickstart.
• Chore: small post-merge fix to modelito/ollama.py.

0.2.0 - Unreleased

• Export fixes and API consistency (install_service, ensure_ollama_running_verbose).
• OllamaProvider now attempts to use the local Ollama HTTP API when available and falls back to a deterministic shim.
• OllamaProvider now attempts to use the local Ollama HTTP API when available and falls back to the Ollama CLI (when present) before using a deterministic shim.
• Add unit tests covering SDK detection for OpenAIProvider, ClaudeProvider, GeminiProvider, and OllamaProvider.
• CI workflow updated to run package tests only for more stable runs.
• Documentation and README updated to reflect a production-ready library and optional extras.

0.1.0 - Unreleased

• Initial extraction of core helpers: tokenizer, timeout, connector, config, ollama_service
• Add tests, CI, and publish workflow

0.1.1 - Release

• Add packaging metadata and setuptools discovery in pyproject.toml
• Export runtime __version__ from the package
• Build wheel and sdist artifacts
• Add concise usage docs and install/build instructions

v1.0.3

modelito v1.0.3 — Audit and small fixes

Release date: 2026-04-19

Summary

modelito v1.0.3 focuses on a deeper repository audit, CI hardening, and a few small, user-facing fixes.

Highlights

• Audit: Performed a deeper code & docs audit and added AUDIT_REPORT.md with findings and recommendations.
• CI: scripts/check_no_legacy_dicts.py now targets only docs/examples/tests to avoid false positives in code.
• Fix: modelito/openai.py deterministic fallback now uses the flattened messages list for consistent summarization.
• Packaging & release: bumped package to 1.0.3 and published to PyPI.
• Tooling: unit tests, ruff, and mypy ran clean in the release validation.

Breaking changes

None. Backward-compatible changes only.

Upgrade

Run one of these commands to upgrade:

pip install -U modelito
# or pin to this release
pip install modelito==1.0.3

Links

• GitHub Release: https://github.com/krahd/modelito/releases/tag/v1.0.3
• PyPI: https://pypi.org/project/modelito/1.0.3/
• Audit report: https://github.com/krahd/modelito/blob/main/AUDIT_REPORT.md

Notes

• Provider integration tests (OpenAI/Anthropic/Ollama) require credentials and a self-hosted runner; these were not executed in this release run.
• Report regressions or issues at https://github.com/krahd/modelito/issues/new

Thank you to everyone who contributed and tested this release.

modelito v1.0.2

Changelog

All notable changes to this project will be documented in this file.

1.0.0 - 2026-04-19

• Major release: break compatibility in favor of a cleaner, modern API.
• Add typed Message/Response dataclasses and expanded provider Protocols.
• Streaming stream() surface implemented for OpenAI, Claude, Gemini, Ollama, and Grok.
• SDK-aware integrations for OpenAI (async/streaming/embed), Claude and Gemini.
• Ollama HTTP streaming implemented with robust fallback behavior.
• Integration tests added (gated by environment variables).
• CI/type/lint updates and general cleanup.

0.2.2 - 2026-04-18

• Bump version to 0.2.2.
• All tests passing (24 passed, 1 skipped).

v0.1.3 - 2026-04-18

• Merge fix/ollama-compat: Ollama compatibility improvements (CLI fallback, improved list_models).
• Documentation: update README.md and docs/USAGE.md with clearer overview and quickstart.
• Chore: small post-merge fix to modelito/ollama.py.

0.2.0 - Unreleased

• Export fixes and API consistency (install_service, ensure_ollama_running_verbose).
• OllamaProvider now attempts to use the local Ollama HTTP API when available and falls back to a deterministic shim.
• OllamaProvider now attempts to use the local Ollama HTTP API when available and falls back to the Ollama CLI (when present) before using a deterministic shim.
• Add unit tests covering SDK detection for OpenAIProvider, ClaudeProvider, GeminiProvider, and OllamaProvider.
• CI workflow updated to run package tests only for more stable runs.
• Documentation and README updated to reflect a production-ready library and optional extras.

0.1.0 - Unreleased

• Initial extraction of core helpers: tokenizer, timeout, connector, config, ollama_service
• Add tests, CI, and publish workflow

0.1.1 - Release

• Add packaging metadata and setuptools discovery in pyproject.toml
• Export runtime __version__ from the package
• Build wheel and sdist artifacts
• Add concise usage docs and install/build instructions

v1.0.1

Chore: strict mypy fixes, docs cleanup, CI improvements, add TestPyPI workflow. See CHANGELOG.

v0.2.2

Bump version to 0.2.2.

All tests passing (24 passed, 1 skipped).

See CHANGELOG.md for details.