Case Studies: Drop-In Replacements

qortex augments existing agent frameworks by implementing their memory/knowledge/vector interfaces. Same API, zero code changes, plus graph-enhanced retrieval and feedback-driven learning layered on top of what these excellent frameworks already provide.

Each case study is a self-contained E2E demonstration backed by real tests in the repo.

The Lineup

Framework What We Replace Protocol Test File
LangChain VectorStore Chroma, FAISS, Pinecone, any VectorStore Python ABC tests/test_langchain_e2e_dogfood.py
Mastra Any MastraVector impl (PgVector, Chroma, Pinecone...) MCP (cross-language) tests/test_mastra_mcp_dogfood.py
Agno Any KnowledgeProtocol impl Python protocol (duck-typed) tests/test_e2e_agno.py
CrewAI KnowledgeStorage (ChromaDB default) Python ABC tests/test_framework_compat.py
LangChain Retriever Any BaseRetriever Python ABC (Pydantic) tests/test_framework_compat.py

The Pitch

Before (great, but vector-only): 

Your Agent → Framework's Vector Store → Cosine Similarity → Results

After (augmented with qortex): 

Your Agent → qortex Adapter → Vec + Graph + Feedback → Results That Improve

Swap one import and the agent gets graph structure and continuous learning without any other code changes.

What Makes This Different

Every framework above provides excellent cosine similarity search. qortex builds on that foundation and adds:

  1. Graph-enhanced retrieval (Level 1): typed edges between concepts, PPR over knowledge graph
  2. Rules auto-surfaced in results: linked rules appear in query responses, zero consumer effort
  3. Graph exploration: explore(node_id) traverses typed edges, surfaces neighbors and rules
  4. Feedback-driven learning (Level 2): consumer outcomes adjust retrieval weights over time
  5. Cross-framework consistency: same backend, same results, regardless of which adapter you use
  6. Cross-language via MCP: TypeScript consumers (Mastra) talk to qortex over JSON-RPC

The consumer starts at Level 0 (vec parity) and gets Levels 1-2 for free as the system matures.