How abhinavalwala/enterprise-ai-search works
This is an evaluated retrieval and RAG platform. The serving workflow runs from an HTTP request through the API/service boundary, permission-aware hybrid retrieval, reranking, and either search results or evidence-grounded generation with citation checks; request tracing and bounded metrics observe requests. A separate privileged CLI supports corpus preparation and scientific comparisons, including claim verification. The graph preserves the documented evaluation and safety branches. Some service/API and evaluation internals are unsampled; those connections rely on the README or SOURCE INDEX references, not inspection of the full modules.
Main entry points: service.py Search service · cli.py Scientific CLI · rag.py RAG and citations · bm25.py BM25 index
Components 21
Serving
- FastAPI
src/enterprise_ai_search/api.py - Search service
src/enterprise_ai_search/service.py
Retrieval
- Chunking and text
src/enterprise_ai_search/text.py - Document models
src/enterprise_ai_search/models.py - BM25 index
src/enterprise_ai_search/bm25.py - Dense index
src/enterprise_ai_search/dense.py - Hybrid fusion
src/enterprise_ai_search/hybrid.py - Cross-encoder reranker
src/enterprise_ai_search/reranker.py
RAG and verification
- RAG and citations
src/enterprise_ai_search/rag.py - HTTP generator
src/enterprise_ai_search/generation.py - Claim verification
src/enterprise_ai_search/claim_verification.py
Corpus and evaluation
- Corpus loader
src/enterprise_ai_search/dataset.py - Scientific CLI
src/enterprise_ai_search/cli.py - Retrieval evaluation
src/enterprise_ai_search/evaluation.py - Method comparison
src/enterprise_ai_search/comparison.py - Hybrid evaluation
src/enterprise_ai_search/hybrid_evaluation.py - Claim evaluation
src/enterprise_ai_search/claim_evaluation.py
Policy and observability
- Authorization policy
src/enterprise_ai_search/authorization.py - Request observability
src/enterprise_ai_search/observability.py
Other components
- HTTP client
- Generation endpoint
Connections 35
- HTTP client → FastAPI: sends requests
README.md - FastAPI → Search service: dispatches to
src/enterprise_ai_search/api.py - FastAPI → Request observability: records requests
src/enterprise_ai_search/api.py - Search service → Corpus loader: loads corpus
src/enterprise_ai_search/service.py - Search service → Chunking and text: chunks documents
src/enterprise_ai_search/service.py - Search service → BM25 index: builds index
src/enterprise_ai_search/service.py - Search service → Dense index: prepares embeddings
src/enterprise_ai_search/service.py - Search service → Hybrid fusion: builds index
src/enterprise_ai_search/service.py - Search service → Cross-encoder reranker: loads model
src/enterprise_ai_search/service.py - Search service → Authorization policy: checks access
src/enterprise_ai_search/service.py - Search service → RAG and citations: answers questions
src/enterprise_ai_search/service.py - Search service → HTTP generator: configures generator
src/enterprise_ai_search/service.py - Search service → Request observability: logs events
src/enterprise_ai_search/service.py - Chunking and text → Document models: creates chunks
src/enterprise_ai_search/text.py - BM25 index → Chunking and text: tokenizes text
src/enterprise_ai_search/bm25.py - BM25 index → Document models: returns results
src/enterprise_ai_search/bm25.py - Dense index → Chunking and text: normalizes text
src/enterprise_ai_search/dense.py - Dense index → Document models: returns results
src/enterprise_ai_search/dense.py - Hybrid fusion → BM25 index: retrieves candidates
src/enterprise_ai_search/hybrid.py - Hybrid fusion → Dense index: retrieves candidates
src/enterprise_ai_search/hybrid.py - RAG and citations → Hybrid fusion: retrieves evidence
src/enterprise_ai_search/rag.py - RAG and citations → Cross-encoder reranker: reranks candidates
src/enterprise_ai_search/rag.py - RAG and citations → HTTP generator: sends evidence
src/enterprise_ai_search/rag.py - HTTP generator → Generation endpoint: sends prompt
src/enterprise_ai_search/generation.py - Claim evaluation → Claim verification: evaluates claims
src/enterprise_ai_search/claim_evaluation.py - Claim verification → RAG and citations: uses evidence flow
src/enterprise_ai_search/claim_verification.py - Claim verification → HTTP generator: requests verdict
src/enterprise_ai_search/claim_verification.py - Scientific CLI → Corpus loader: loads datasets
src/enterprise_ai_search/cli.py - Scientific CLI → Retrieval evaluation: runs evaluation
src/enterprise_ai_search/cli.py - Scientific CLI → Method comparison: compares methods
src/enterprise_ai_search/cli.py - Scientific CLI → Hybrid evaluation: runs evaluation
src/enterprise_ai_search/cli.py - Scientific CLI → Claim evaluation: runs evaluation
src/enterprise_ai_search/cli.py - Method comparison → Retrieval evaluation: uses metrics
src/enterprise_ai_search/comparison.py - Hybrid evaluation → Retrieval evaluation: uses metrics
src/enterprise_ai_search/hybrid_evaluation.py - Claim evaluation → Corpus loader: loads claims
src/enterprise_ai_search/claim_evaluation.py