make_source.yaml

---
module: make_source
description: ''
tags:
- builder
- core
- loader
- storage
category: []
used_in: []
entrypoint: true
related_nodes: []
objects:
- name: EMBEDDING_MODEL
  kind: const
  description: ''
  references: []
- name: tokenize
  kind: function
  decorators: []
  description: ''
  references: []
- name: detect_language
  kind: function
  decorators: []
  description: ''
  references:
  - langdetect.LangDetectException
  - langdetect.detect
- name: load_companion_yaml
  kind: function
  decorators: []
  description: Load a companion .yaml file for an .md file if it exists.
  references:
  - yaml.YAMLError
  - yaml.safe_load
- name: _normalize_list
  kind: function
  decorators: []
  description: Normalize a YAML value to a list of strings.
  references: []
- name: process_md_file
  kind: function
  decorators: []
  description: ''
  references: []
- name: process_go_yaml
  kind: function
  decorators: []
  description: Process a .go.yaml companion file as structured knowledge chunks. The
    .go source file is not indexed (code excluded by design); this YAML is the sole
    knowledge source for Go files in the KB. Produces one chunk per object defined
    in the file. File-level metadata (tags, category, used_in) is applied to all chunks.
  references:
  - yaml.YAMLError
  - yaml.safe_load
- name: process_yaml_file
  kind: function
  decorators: []
  description: Processes a YAML file as a single, large chunk to preserve context.
  references:
  - yaml.YAMLError
  - yaml.dump
  - yaml.safe_load
- name: create_embeddings
  kind: function
  decorators: []
  description: Encode texts using a multilingual sentence transformer model.
  references:
  - numpy.array
  - sentence_transformers.SentenceTransformer
- name: build_faiss_index
  kind: function
  decorators: []
  description: Build FAISS inner-product index (cosine sim with normalized vectors).
  references:
  - faiss.IndexFlatIP
- name: build_bm25_index
  kind: function
  decorators: []
  description: Build BM25 index for lexical search.
  references:
  - rank_bm25.BM25Okapi
- name: create_bm25_inverted_index
  kind: function
  decorators: []
  description: Lightweight inverted index from BM25 scores (for SQLite compat).
  references: []
- name: build_metadata_index
  kind: function
  decorators: []
  description: Build O(1) lookup indexes for tags, category, used_in, entrypoints.
  references: []
- name: _resolve_related_node
  kind: function
  decorators: []
  description: Resolve a relative related_node path to chunk_ids.
  references: []
- name: create_knowledge_graph_with_content
  kind: function
  decorators: []
  description: Build knowledge graph using embedding cosine similarity + YAML related_nodes.
  references: []
- name: _content_hash
  kind: function
  decorators: []
  description: Return SHA256 hex digest of normalized text content.
  references: []
- name: _dedup_chunks
  kind: function
  decorators: []
  description: 'Deduplicate chunks with identical content. Chunks with the same text
    are merged into a single chunk: - file_path  : canonical (first seen) path — kept
    for backward compat - file_paths : list of ALL paths where this content appears
    - tags / category / used_in / related_nodes : union of all occurrences'
  references: []
- name: _is_go_meta_yaml
  kind: function
  decorators: []
  description: Return True if the YAML file is a Go meta companion (has 'package'
    + 'objects'). Used when the sibling .go file is absent (e.g. in docs-only submodules).
  references:
  - yaml.safe_load
- name: process_directory
  kind: function
  decorators: []
  description: ''
  references: []
- name: build_knowledge_base
  kind: function
  decorators: []
  description: ''
  references: []
- name: save_knowledge_base_legacy
  kind: function
  decorators: []
  description: ''
  references:
  - faiss.write_index
- name: save_kb_to_sqlite
  kind: function
  decorators: []
  description: ''
  references: []
- name: generate_edge_types_doc
  kind: function
  decorators: []
  description: Generates a markdown file documenting all unique edge types.
  references: []