Config reference: enrichment

• Enterprise tuning surface

  ---

  Defaults + constraints are rendered directly from Pydantic.

• Env keys when available

  ---

  Many fields have an env-style alias (from TriBridConfig.to_flat_dict()).

• Tooltip-level guidance

  ---

  If a matching glossary entry exists, you’ll see deeper tuning notes.

Config reference Config API & workflow Glossary

Total parameters: 6

Group index

• (root)

(root)

JSON key	Env key(s)	Type	Default	Constraints	Summary
enrichment.chunk_summaries_enrich_default	CHUNK_SUMMARIES_ENRICH_DEFAULT	int	1	≥ 0, ≤ 1	Enable chunk_summary enrichment by default
enrichment.chunk_summaries_max	CHUNK_SUMMARIES_MAX	int	100	≥ 10, ≤ 1000	Max chunk_summaries to generate
enrichment.enrich_code_chunks	ENRICH_CODE_CHUNKS	int	1	≥ 0, ≤ 1	Enable chunk enrichment
enrichment.enrich_max_chars	ENRICH_MAX_CHARS	int	1000	≥ 100, ≤ 5000	Max chars for enrichment prompt
enrichment.enrich_min_chars	ENRICH_MIN_CHARS	int	50	≥ 10, ≤ 500	Min chars for enrichment
enrichment.enrich_timeout	ENRICH_TIMEOUT	int	30	≥ 5, ≤ 120	Enrichment timeout (seconds)

Details (glossary)

enrichment.chunk_summaries_enrich_default (CHUNK_SUMMARIES_ENRICH_DEFAULT) — Chunk Summaries Enrich Default

Category: general

Controls whether chunk summaries are generated with richer, model-assisted metadata by default. Enriched summaries can add intent, entities, API surface hints, and semantic cues that improve retrieval and reranking beyond raw embeddings alone. The trade-off is higher indexing cost and longer build times, especially on large repositories. Enable enrichment when search quality and explainability matter more than ingestion speed, and disable it for rapid iteration pipelines where you need frequent low-cost reindexing.

Badges: - Metadata quality

Links: - Code-Craft Summarization (2025) - OpenAI Summarization Cookbook - LlamaIndex Vector Store Index - LangChain Retrieval Concepts

enrichment.chunk_summaries_max (CHUNK_SUMMARIES_MAX) — Max Chunk Summaries

Category: general

Caps how many chunk summaries are produced for a corpus. This is a budget control over indexing cost, storage footprint, and retrieval metadata coverage. A low cap is fast but can miss important modules, while a high cap improves coverage and long-tail recall at the cost of longer ingestion and larger indexes. Choose this value based on corpus size and criticality, then validate with retrieval benchmarks so the limit reflects actual answer quality rather than arbitrary round numbers.

Badges: - Coverage budget

Links: - MIRAGE Benchmark (2025) - T2-RAGBench (2025) - OpenAI Summarization Cookbook - LlamaIndex Vector Store Index

enrichment.enrich_code_chunks (ENRICH_CODE_CHUNKS) — Enrich Code Chunks

Category: chunking

When enabled, each code chunk is augmented with model-generated summaries or semantic descriptors during indexing. This often improves conceptual retrieval because rerankers can match intent signals beyond literal token overlap. The tradeoff is extra indexing time, compute cost, and the risk of noisy metadata if prompts or models are weak. Chunk size and model selection both matter: oversized chunks produce vague summaries, while tiny chunks lose architectural context. Evaluate this feature with task-based retrieval metrics to confirm the added metadata improves real query outcomes.

Badges: - Slower indexing

Links: - EyeLayer: Human Attention for Code Summarization (arXiv 2026) - Meta-RAG on Large Codebases Using Code Summarization (arXiv 2025) - LlamaIndex Repository - LangChain Repository