v0.10.4 — Visible Chunk Failures & Hypergraph Incremental Feed Fix

Bugfix release reported by a downstream integration: an incremental-feed crash in hypergraph KAs, and silent chunk-extraction failures.

🐛 Hypergraph Incremental Feed Crash

Feeding a second document into an existing hypergraph KA crashed with AttributeError: 'key_extractor' — _update_data_state referenced an attribute that only exists as node_key_extractor/edge_key_extractor. First feeds worked (empty-state guard), every subsequent feed failed.

👁️ Visible Chunk Extraction Failures

Per-chunk LLM extraction failures were converted to None and silently skipped — sources could be ingested with missing chunks and callers had no way to tell.

• Every feed_text() / parse() run now records failures: read ka.extraction_failures → [{"chunk_index": ..., "stage": ..., "error": ...}]
• he feed / he parse print a warning when chunks were dropped

⚙️ on_error Strategy

New constructor option on every AutoType:

• "skip" (default): current behavior — record and continue with surviving chunks
• "raise": abort the feed/parse on the first failing chunk

ka = Template.create("method/graph_rag", on_error="raise")

Full changelog: v0.10.3...v0.10.4