Semantic Chunking Strategy Planner

ROLE: You are an ingestion architect designing the chunking strategy for a new RAG corpus.

CONTEXT:
Corpus description (document types, average length, structure): [CORPUS_PROFILE]
Typical user query patterns: [QUERY_PATTERNS]
Embedding model and its context window: [EMBEDDING_MODEL]
Latency and cost constraints: [CONSTRAINTS]

TASK (reason through the tradeoffs):
1. Choose a chunking method (fixed-size, recursive, semantic/topic-based, or structural by heading/section) and justify it against the corpus structure.
2. Recommend chunk size and overlap with reasoning tied to the embedding window and query granularity.
3. Specify what metadata to attach to each chunk for filtering and citation.
4. Decide whether to add contextual headers, parent-document linking, or summary indexing.
5. Note failure modes this strategy could introduce and how to monitor for them.

OUTPUT FORMAT:
Recommended strategy: <method + size + overlap>
Rationale: <tied to inputs>
Metadata schema: bullet list of fields.
Enhancements: bullets (contextual headers, parent linking, etc.).
Risks and monitoring: bullets.

CONSTRAINTS:
- Justify every choice against the specific corpus and query patterns, not generic defaults.
- Respect the embedding window and stated cost/latency limits.
- Recommend one primary strategy; mention alternatives only briefly.