Tier 0: Ingestion Only¶

Goal¶

Collect a stable, low-cost record for every processed document.

Tier 0 does not perform semantic reasoning. It only captures ingestion metadata that later tiers can build on.

Required Fields Per Document¶

id
source
timestamp
author/agent
doc_length
basic metadata

Current Record Shape¶

Tier0Record includes: - id: stable document identifier (currently relative path) - source: source document path (relative to scan root) - timestamp: last modified time (unix seconds string, when available) - author_agent: extracted from frontmatter fields such as author, agent, author_agent, created_by - doc_length: document size in bytes - metadata: lightweight document metadata map

Basic Metadata (Current)¶

concept
raw_concept
file_ext
heading_count
outbound_link_count
path
file_size_bytes
frontmatter presence flags (when present)

Linking Metadata to Documents¶

Tier 0 links metadata to documents through id and source. Both map back to the processed document path.

Tier 0 Index File¶

Use the CLI to write a persistent index:

./target/debug/lint-ai /path/to/repo --tier0-index-out

Default output file: - tier0-index.json

Custom output path:

./target/debug/lint-ai /path/to/repo --tier0-index-out out/my-index

If the provided output path has no extension, .json is appended automatically.

Index Structure¶

The index file includes: - tier - generated_at_unix - path - document_count - documents_by_id (id -> Tier0Record)

This provides a stable lookup table for Tier 1+ pipelines.