LlamaIndex 0.2.1 — channel attribution

Updates the LlamaIndex package homepage and documentation links with channel attribution, and restores tag-based PyPI publishing from the migrated brand repository.

Runtime code, tool schemas, and dependencies are unchanged (line endings normalized). Validation: 71 network-disabled tests, Ruff, wheel/sdist build, Twine checks, and fresh installation with 17 tool schemas; all five PR checks passed.

Source: #6
Package: https://pypi.org/project/llama-index-tools-flashdata/0.2.1/

This release tags the LlamaIndex integration only. PyPI distribution is performed by the dedicated Publish LlamaIndex integration workflow.

Haystack 0.2.1 — channel attribution

Updates the Haystack package homepage and documentation with channel attribution, corrects package availability information, and makes all 15 README guide links work from PyPI. Adds dedicated tag-only PyPI publishing from the brand repository.

Validation: 81 network-disabled tests, Ruff, wheel/sdist builds, Twine checks, and fresh installation with 17 components, tool schemas and pipeline serialization. All five PR checks passed. Runtime logic, schemas and dependencies are unchanged; only package version metadata and line endings differ.

Source: #7
Package: https://pypi.org/project/flashdata-haystack/0.2.1/

This tag covers the Haystack integration only. The dedicated Publish Haystack integration workflow distributes the package to PyPI.

flashdata-mcp v0.3.0

FlashData MCP 0.3.0

Transcript and captions accept one language code or 1–10 distinct codes. Omit language
to prefer en-orig, then en, returning one track. Lists preserve successful tracks and
per-language errors; existing single-language calls remain compatible.

MCP text defaults to Markdown. Explicit text/JSON presentation preserves structured
fields, source links and timestamps. Other API contracts remain unchanged.

This release includes the combined npm package, 15 task packages and MCPB bundles,
the AI SDK adapter, and updated transcript/captions Activepieces packages. Python
task CLIs, Haystack and LlamaIndex distributions are prepared for separate PyPI upload.

Registry uploads to npm and PyPI are manual. Download the release archives to install
the verified candidates directly; existing registry versions and client pins remain
unchanged until the uploads are verified. Hosted MCP deployment is separate.