How saicharan67778/queriom-ai works
Queriom is a Streamlit document-intelligence app: users upload PDFs or TXT files, ask grounded questions, or request executive reports. Ingestion extracts page-aware text and chunks it; indexing enables dense and lexical retrieval, while report and Q&A synthesis use Groq and provide citations. Feedback is logged and approved or corrected answers can enter semantic memory; the database also supports DPO export. The README documents additional capabilities, but some UI wiring and implementation details fall outside the excerpts, so uncertain links are omitted.
Main entry points: app.py Streamlit App · feedback_db.py Feedback Database · rag_pipeline.py Q&A Synthesis · document_processor.py Document Processor
Components 16
User Interface
- Streamlit App
app.py
Document Ingestion
- Document Processor
core/document_processor.py
Retrieval and Synthesis
- Corpus Index
core/rag_pipeline.py - Dense Retrieval
core/rag_pipeline.py - BM25 Retrieval
core/rag_pipeline.py - Hybrid Retrieval
core/rag_pipeline.py - Distance Gate
core/rag_pipeline.py - Q&A Synthesis
core/rag_pipeline.py - Report Synthesis
core/rag_pipeline.py - Corpus Digest
core/rag_pipeline.py
Feedback and Memory
- Feedback Database
core/feedback_db.py - Query Logs
core/feedback_db.py - Semantic Memory
core/feedback_db.py - DPO Export
core/feedback_db.py
Other components
- User
- Groq Inference
Connections 27
- User → Streamlit App: uploads and asks
README.md - Streamlit App → Document Processor: processes uploads
app.py - Streamlit App → Corpus Index: initializes index
app.py - Document Processor → Corpus Index: supplies chunks
app.py - Streamlit App → Corpus Digest: requests digest
app.py - Streamlit App → Q&A Synthesis: submits questions
app.py - Streamlit App → Report Synthesis: requests reports
app.py - Corpus Index → Dense Retrieval: adds documents
core/rag_pipeline.py - Corpus Index → BM25 Retrieval: builds retriever
core/rag_pipeline.py - Q&A Synthesis → Dense Retrieval: searches similarity
core/rag_pipeline.py - Q&A Synthesis → BM25 Retrieval: retrieves terms
core/rag_pipeline.py - Report Synthesis → Dense Retrieval: searches context
README.md - Report Synthesis → BM25 Retrieval: uses hybrid search
core/rag_pipeline.py - Dense Retrieval → Hybrid Retrieval: supplies matches
core/rag_pipeline.py - BM25 Retrieval → Hybrid Retrieval: supplies matches
core/rag_pipeline.py - Hybrid Retrieval → Distance Gate: checks relevance
core/rag_pipeline.py - Distance Gate → Q&A Synthesis: passes context
core/rag_pipeline.py - Q&A Synthesis → Groq Inference: generates answer
core/rag_pipeline.py - Report Synthesis → Groq Inference: generates report
core/rag_pipeline.py - Corpus Digest → Groq Inference: summarizes corpus
core/rag_pipeline.py - Streamlit App → Feedback Database: records feedback
app.py - Feedback Database → Query Logs: logs queries and edits
core/feedback_db.py - Feedback Database → Semantic Memory: indexes approved answers
core/feedback_db.py - Semantic Memory → Feedback Database: returns similar queries
core/feedback_db.py - Feedback Database → DPO Export: exports preference pairs
core/feedback_db.py - Streamlit App → Feedback Database: looks up memory
app.py - Feedback Database → Q&A Synthesis: provides exemplars