doc_index

doc_index parses a document file — Markdown, CSV, YAML, JSON, plain text — and breaks it into sections stored as graph nodes. After indexing, use doc_toc, doc_get, and doc_query to read it efficiently without loading the entire file.

Usage

doc_index(path="docs/ARCHITECTURE.md")
doc_index(path="/abs/path/to/spec.yaml")
doc_index(path="data.csv", project_root="/home/user/myproject")

Parameters

Parameter	Type	Default	Description
path	string	required	Path to the file to index (absolute or relative)
project_root	string	—	Override the project root for relative path resolution

Supported formats

.md, .markdown, .csv, .txt, .rst, .toml, .yaml, .yml, .json

Code files (.ts, .rs, .py, etc.) are handled by the code graph — use code_get for those.

doc_id naming

After indexing, the file is referenced by a doc_id derived from its path with non-alphanumeric characters replaced by _:

Path	doc_id
docs/ARCHITECTURE.md	docs_ARCHITECTURE_md
/abs/path/to/spec.yaml	abs_path_to_spec_yaml
data.csv	data_csv

Standard workflow

# 1. Index once
doc_index(path="docs/ARCHITECTURE.md")

# 2. See structure
doc_toc(doc_id="docs_ARCHITECTURE_md")

# 3. Fetch a section (~200 tokens)
doc_get(doc_id="docs_ARCHITECTURE_md", section="Authentication")

# 4. Search across all chunks
doc_query(doc_id="docs_ARCHITECTURE_md", keyword="rate limit")

Re-run doc_index to refresh after the source file changes.