Skip to content

v0.10.0 — Multi-Agent Infrastructure

Choose a tag to compare

View all tags
@VforVitorio VforVitorio released this 22 Mar 17:55
· 819 commits to main since this release
v0.10.0
ef501c2

v0.10.0 — Multi-Agent Infrastructure

First milestone of the LangGraph multi-agent phase. Two of seven sub-agents are complete, along with the full RAG indexing pipeline and the importable src/rag/ module.

What's new

N25 — Pace Agent

  • Wraps the N06 XGBoost model as a LangGraph ReAct agent
  • Returns PaceOutput: lap time prediction + delta vs session median + bootstrap CI (P10/P90, N=200)
  • Tools: predict_pace_tool, get_session_median_tool
  • Export: data/models/agents/pace_agent_config_v1.json

N30 — RAG Agent

  • Retrieval-augmented generation over FIA Sporting Regulations 2023–2025
  • Embedding: BAAI/bge-m3 (1024-dim) · Vector store: Qdrant local · 2,279 chunks indexed
  • Returns RegulationContext: LLM answer + source chunks + article references from metadata
  • Tool: query_rag_tool (LangChain @tool, importable by N31)
  • Export: data/models/agents/rag_agent_config_v1.json

src/rag/ module — first active src/ module outside telemetry

  • RagRetriever class: initialises Qdrant client + BGE-M3 encoder once; singleton via get_retriever()
  • query_rag_tool: LangChain tool ready to pass to any LangGraph agent
  • N31 import: from src.rag.retriever import get_retriever, query_rag_tool

Scripts

  • scripts/download_fia_pdfs.py: scrapes FIA PDF URLs with DownloadConfig dataclass
  • scripts/build_rag_index.py: PDF → chunk → embed → upsert pipeline with hash-based deduplication

Documentation

  • src/rag/README.md — public API reference + usage examples
  • src/agents/README.md, src/nlp/README.md, src/strategy/README.md, src/data_extraction/README.md — status + legacy notices per subpackage

Next

N26 (Tire) → N27 (Race Situation) → N28 (Pit Strategy) → N29 (Radio) → N31 (Strategy Orchestrator)

Assets 2
Loading