How zebir-sir/agentic-rag-paper-assistant works
PaperWeave is a personal research-paper knowledge space. Researchers import PDFs into a durable queue, then read and annotate papers, explore a relationship graph, or ask evidence-grounded questions through the React workspace. The README documents ingestion, multi-source retrieval, answer review, memory, and reader capabilities; source excerpts and the SOURCE INDEX support selected implementation links. Several subsystems were not sampled, so their documented relationships are cited to the README rather than presented as inspected implementation details.
Main entry points: agent_langgraph.py Research workflow · ingest.py Ingest and embed · api.py FastAPI service · ingestion_jobs.py Ingestion tasks
Components 24
Research workspace
- React workspace
web/src/App.jsx - PDF reader
web/src/Reader.jsx - Knowledge star map
web/src/PaperGraph.jsx - API client
web/src/api.js
Paper ingestion
- Ingestion tasks
agent/ingestion_jobs.py - Task queue state
agent/ingestion_tasks_db.py - RabbitMQ queue
agent/rabbitmq_producer.py - Ingestion worker
agent/ingestion_worker.py - PDF extraction
ingestion/extract_files.py - Section chunking
ingestion/chunker.py - Figure analysis
ingestion/vision_client.py - Ingest and embed
ingestion/ingest.py
Research assistant
- FastAPI service
agent/api.py - Intent and source planning
agent/intent_planner.py - Research workflow
agent/agent_langgraph.py - Retrieval tools
agent/tools.py - Evidence review
agent/answer_review_runtime.py
Knowledge services
- Paper evidence store
agent/db_utils.py - Relationship runtime
agent/graph_runtime.py
Conversation memory
- Memory runtime
agent/memory_utils.py - Conversation records
agent/db_utils.py
Other components
- Researcher
- OpenAlex
- Web search
Connections 31
- Researcher → React workspace: uses
README.md - React workspace → API client: calls
web/src/App.jsx - PDF reader → API client: calls
web/src/Reader.jsx - Knowledge star map → API client: uses
README.md - API client → FastAPI service: requests
README.md - FastAPI service → Ingestion tasks: submits PDFs
README.md - Ingestion tasks → Task queue state: creates tasks
agent/ingestion_jobs.py - Ingestion tasks → RabbitMQ queue: enqueues
agent/ingestion_jobs.py - RabbitMQ queue → Ingestion worker: dispatches
README.md - Ingestion worker → Ingestion tasks: executes tasks
agent/ingestion_worker.py - Ingestion worker → Ingest and embed: runs ingestion
agent/ingestion_worker.py - Ingest and embed → PDF extraction: extracts PDF
ingestion/ingest.py - Ingest and embed → Section chunking: builds chunks
ingestion/ingest.py - Ingest and embed → Figure analysis: analyzes figures
ingestion/ingest.py - Ingest and embed → Paper evidence store: persists evidence
ingestion/ingest.py - Ingest and embed → Relationship runtime: refreshes graph
ingestion/ingest.py - FastAPI service → Intent and source planning: plans questions
README.md - Intent and source planning → Research workflow: selects sources
README.md - Research workflow → Retrieval tools: retrieves evidence
agent/agent_langgraph.py - Retrieval tools → Paper evidence store: searches papers
agent/tools.py - Research workflow → Relationship runtime: expands candidates
README.md - Relationship runtime → Paper evidence store: reads relationships
README.md - Research workflow → OpenAlex: searches metadata
README.md - Research workflow → Web search: searches web
README.md - Research workflow → Evidence review: reviews evidence
README.md - Evidence review → React workspace: returns answer
README.md - PDF reader → FastAPI service: requests reading
README.md - FastAPI service → Paper evidence store: loads paper data
README.md - FastAPI service → Memory runtime: uses memory
README.md - Memory runtime → Conversation records: reads and writes
README.md - Research workflow → Conversation records: stores messages
README.md