RAGWire v1.2.9

RAGWire 1.2.9 — Release Notes

Agent & LLM Friendliness

• Added llms.txt — follows the llmstxt.org standard; available at the repo root and served at https://laxmimerit.github.io/RAGWire/llms.txt. Contains the complete API reference, annotated config.yaml, filter syntax rules, metadata schema, and agent integration pattern — all in a single plain-text file that any AI coding tool can fetch and use.
• Added llms-full.txt — full API reference in plain text (MkDocs-syntax-free), with both example files and config templates appended. Linked from llms.txt.
• Improved README.md Quick Start — now includes explicit filter examples, multi-year OR logic, agent-controlled filtering pattern, and exact doc.metadata key names with types.

New Types

• IngestStats / IngestError — typed TypedDict return types for ingest_documents() and ingest_directory(). Agents and IDEs now get full autocomplete on the stats dict (total, processed, skipped, failed, chunks_created, errors). Fully backwards-compatible.

Better Error Messages

• Missing embeddings section — raises a clear ValueError with a config snippet and valid provider list instead of a cryptic downstream error.
• Missing vectorstore.url — raises a ValueError with a config example and the Docker command to start Qdrant locally.
• LLM provider not installed — ImportError now includes the exact pip install "ragwire[provider]" command rather than a raw ModuleNotFoundError.
• Unsupported LLM/embedding provider — error messages now list all valid options and include a config snippet.