v0.0.4 — Cache retrieval fix + full value retrieval

What's Changed

Bug Fix

• max_size was accepted but ignored in get_cached_result — now properly forwarded to the cache backend

New Feature

• full=True parameter on get_cached_result — bypasses preview truncation and returns the complete cached value

Dependencies

• mcp-refcache floor raised to >=0.2.1
• Dependency refresh (fastmcp 3.x, langfuse 4.x)

Infrastructure

• Simplified Docker to single multi-stage build (removed base image)
• Fixed release workflow 403 on GHCR push

Tests

• 7 new regression tests (413 total, all passing)

Security

• pip-audit via uv run for accurate lockfile auditing
• Accepted exception: CVE-2026-25990 (Pillow, pinned by langchain-nomic)

Full Changelog: v0.0.3...v0.0.4

v0.0.3 - Semantic Comment Search

New MCP Tools

Search Tools

• semantic_search_comments - Search comments using natural language with auto-indexing
• semantic_search_all - Unified search across transcripts AND comments

Index Management

• get_indexed_videos - List videos in the semantic search index
• delete_indexed_video - Remove content from the semantic search index

Infrastructure

• Added content_type metadata field to all indexed documents
• New CommentChunker class for processing YouTube comments
• Extended TranscriptIndexer with comment indexing methods

Testing

• 406 tests (104 new)
• 73%+ code coverage maintained

See CHANGELOG.md for full details.

v0.0.2 - Semantic Transcript Search

✨ New Features

Semantic Transcript Search

Search YouTube video transcripts using natural language queries!

New MCP Tools:

• warmup_semantic_search - Pre-load embedding model (~270MB) to avoid timeout on first search
• semantic_search_transcripts - Search transcripts with auto-indexing
• index_channel_transcripts - Optional bulk pre-warming for faster searches

Example

# Search Vimjoyer's channel for NixOS topics
results = await semantic_search_transcripts(
    query="how to configure garbage collection",
    channel_ids=["UCuAXFkgsw1L7xaCfnd5JJOw"],
)

# Returns timestamped URLs for direct playback
print(results["results"][0]["timestamp_url"])
# https://www.youtube.com/watch?v=DnA4xNTrrqY&t=231

🔧 Technical Details

• Embeddings: Nomic Embed Text v1.5 with Matryoshka dimensionality (512 dims)
• Vector Store: ChromaDB with optimized HNSW settings
• Chunking: Token-based (256 tokens) with timestamp preservation
• Tests: 343 passing (was 178)

📦 Installation

pip install yt-api-mcp==0.0.2

# Or with uvx
uvx yt-api-mcp stdio

Full Changelog

See CHANGELOG.md

v0.0.1 - Fix stdio mode for MCP clients

🔧 Bug Fix Release

Fixed

• stdio mode protocol fix: Startup info now writes to stderr instead of stdout, fixing MCP client handshake issues with Zed and Claude Desktop

Added

• Aligned executable name: Added yt-api-mcp as primary executable (matching package name)
• yt-mcp remains as legacy alias for backwards compatibility

Installation

# PyPI (recommended)
pip install yt-api-mcp==0.0.1

# Or with uvx
uvx yt-api-mcp stdio

# Legacy alias still works
uvx --from yt-api-mcp yt-mcp stdio

Full Changelog

See CHANGELOG.md

YouTube MCP Server v0.0.0 - First Experimental Release

🎉 First Experimental Release

This is the initial release of the YouTube MCP server - an experimental version to test both the implementation and release workflow.

✨ Features

• 16 YouTube Tools: Search, metadata, transcripts, comments, live streaming
• Intelligent Caching: Multi-tier caching reduces API quota by ~75%
• Docker Support: Production-ready images (229MB)
• 178 Tests: 76% code coverage

📦 Installation

# PyPI
pip install yt-api-mcp

# Or with uvx (recommended)
uvx yt-api-mcp stdio

# Docker
docker pull ghcr.io/l4b4r4b4b4/yt-mcp:0.0.0

⚠️ Version 0.0.0 Notice

This is an experimental first release. Version 0.0.0 signals 'expect issues' - we're testing the release workflow alongside the implementation.

📖 Full Changelog

See CHANGELOG.md for complete details.

🙏 Feedback Welcome

Found an issue? Open a GitHub issue