vinsongqj/42_rag_against_the_machine

How vinsongqj/42_rag_against_the_machine works

RAG against the machine ingests a corpus, chunks and indexes it, retrieves ranked source locations, and can generate grounded answers or evaluate retrieval recall. Users initiate work through the CLI or optional HTTP API; search supports BM25, semantic, and hybrid retrieval, with query caching. The graph preserves separate ingestion, chunking, indexing, ranking, generation, and evaluation stages. Edges reflect README workflows and shown calls/import references; several CLI and batch details are documented but their implementations were not fully sampled.

Main entry points: retrieve.py Source retrieval · __main__.py CLI commands · api.py HTTP API · index.py Index builder

Components 18

User entry points

CLI commands src/__main__.py
HTTP API src/api.py

Corpus indexing

File ingestion src/ingest.py
Chunking src/chunk.py
Index builder src/index.py
BM25 index src/index.py
Semantic index src/semantic.py

Retrieval and ranking

Source retrieval src/retrieve.py
Search ranking src/rank.py
Query cache src/cache.py
Index versioning src/version.py

Answer and evaluation

Answer generation src/generate.py
Recall evaluation src/evaluate.py
Data models src/models.py

Other components

CLI user
HTTP client
Corpus files
Ollama model

Connections 24

CLI user → CLI commands: runs commands README.md
HTTP client → HTTP API: sends requests README.md
Corpus files → File ingestion: provides files src/ingest.py
CLI commands → Index builder: builds indexes src/__main__.py
Index builder → File ingestion: ingests corpus src/index.py
File ingestion → Chunking: chunks files src/ingest.py
Index builder → BM25 index: writes index src/index.py
Index builder → Semantic index: optionally builds src/index.py
Semantic index → Semantic index: persists vectors src/semantic.py
CLI commands → Source retrieval: runs search README.md
Source retrieval → BM25 index: loads index src/retrieve.py
Source retrieval → Search ranking: dispatches search src/retrieve.py
Search ranking → Semantic index: searches vectors src/rank.py
Source retrieval → Query cache: reads and writes src/retrieve.py
Source retrieval → Index versioning: checks versions src/retrieve.py
CLI commands → Answer generation: generates answers README.md
Answer generation → Ollama model: requests generation src/generate.py
CLI commands → Recall evaluation: evaluates results README.md
HTTP API → Source retrieval: retrieves sources src/api.py
HTTP API → Answer generation: generates answers src/api.py
CLI commands → HTTP API: starts server src/__main__.py
File ingestion → Data models: creates chunks src/ingest.py
Source retrieval → Data models: returns sources src/retrieve.py
HTTP API → Data models: uses schemas src/api.py

More Python diagrams: elimorals/skills_mx · elithril/blender-kiln · epfl-vilab/adversarial-prompts · firtoz/trellis · fmetivier/openpattern

Drawn by GitDiagram on October 9, 2026 from the repository's file tree, README and source files. Read as Markdown ·