How farahchannir/ai-codebase-intelligence-engine works
The checked-in runtime is an early code-ingestion and parsing pipeline: a caller supplies a repository source, and the pipeline returns code chunks. The README describes a broader planned codebase-intelligence product, including RAG, Q&A, bug detection, documentation, dependency analysis, APIs, and a UI; those capabilities are not present in the sampled runtime, so they are not represented as implemented components. Source excerpts are partial, so internal helper wiring is shown only where supported.
Main entry points: ingestion.py Ingest sources · chunking.py Chunk dispatcher · main.py Pipeline runner
Components 12
Code ingestion
- Pipeline runner
main.py - Ingest sources
core/ingestion.py - Git clone
core/ingestion.py - ZIP extraction
core/ingestion.py - File discovery and filtering
core/ingestion.py - Language detection
core/ingestion.py - File records
core/ingestion.py
Parsing and chunking
- Chunk dispatcher
core/chunking.py - Structure-aware parsers
core/chunking.py - Window fallback
core/chunking.py - Code chunks
core/chunking.py
Other components
- Pipeline caller
Connections 13
- Pipeline caller → Pipeline runner: runs
- Pipeline runner → Ingest sources: calls
- Ingest sources → Git clone: clones
- Ingest sources → ZIP extraction: extracts
- Ingest sources → File discovery and filtering: filters files
- Ingest sources → Language detection: detects language
- Ingest sources → File records: returns records
- Pipeline runner → Chunk dispatcher: calls per record
- File records → Chunk dispatcher: provides record
- Chunk dispatcher → Structure-aware parsers: dispatches supported languages
- Chunk dispatcher → Window fallback: falls back
- Chunk dispatcher → Code chunks: returns chunks
- Pipeline runner → Code chunks: returns result