Using manifest.json for Lineage Retrieval

This is the second fallback method when MCP lineage tools are not available. Read and parse the manifest.json file directly to extract lineage information.

How to use

1. Locate the manifest.json file:

   • Usually in target/manifest.json in the dbt project root
   • May also be in project root as manifest.json

2. Read the manifest.json file:

   • First check the file size - if it's very large (>10MB), you may need to use streaming or partial reads
   • Look for the target model in the nodes section

3. Extract lineage from the manifest structure:

   {
     "nodes": {
       "model.project.model_name": {
         "unique_id": "model.project.model_name",
         "resource_type": "model",
         "depends_on": {
           "nodes": ["model.project.upstream_model", "source.project.source_name"]
         },
         "original_file_path": "models/path/to/model.sql"
       }
     }
   }

4. Build the lineage graph:

   • Parents: Found in the depends_on.nodes array
   • Children: Search all nodes for ones that have this model in their depends_on.nodes

5. For each node in the lineage, extract:

   • unique_id
   • resource_type (model, source, seed, snapshot, exposure, test)
   • original_file_path or path
   • Any other relevant metadata

Benefits

• ✅ Works offline
• ✅ No MCP server required
• ✅ Contains complete lineage information
• ✅ Includes all metadata

Limitations

• ❌ Manifest can be very large (100MB+)
• ❌ May be slow to parse
• ❌ May not exist if dbt parse hasn't been run
• ❌ Only reflects last parse, not current uncommitted changes

When to use

Use this method when:

• Both get_lineage_dev and get_lineage MCP tools are NOT available
• The manifest.json file exists and is reasonably sized (<10MB)
• You need complete lineage information

Tips

• If the manifest is very large, consider reading it in chunks or using grep/search instead
• If you only need a specific model's lineage, you can use Grep to find just that section
• Check the manifest file size before attempting to read the entire file