kothapavan1998/deeprecall

Implements recursive multi-turn search where the LLM executes `search_db()` calls in a sandboxed Python REPL, analyzes results, refines queries iteratively, and synthesizes final answers—rather than static one-shot retrieval. Supports pluggable vector stores (ChromaDB, Milvus, Qdrant, Pinecone, FAISS) and LLM backends with built-in cost tracking, execution budgets (timeouts/tokens/errors), context compaction, and callbacks for observability. Includes adapters for LangChain and LlamaIndex ecosystems.

Used by 1 other package. Available on PyPI.