ChromaDB + RAG

Security References

CodeCritic uses a ChromaDB RAG pipeline to ground AI suggestions in real security standards. Our in-depth reviews retrieve relevant CWE and OWASP entries from a vector store, so every vulnerability finding is backed by authoritative references.

How RAG Works

  1. Classify risks — The LLM reads your code diff and identifies security risk types (e.g., SQL injection, weak auth).
  2. Query ChromaDB — That risk summary is used to retrieve the top 10 semantically relevant CWE/OWASP entries from the vector store.
  3. Generate grounded suggestions — The LLM produces fixes using only the retrieved references, citing real CWE IDs and OWASP categories.
  4. Validate citations — Every cited CWE ID is checked against ChromaDB. Invalid citations trigger a retry to prevent hallucination.

CWE — Common Weakness Enumeration

A catalog of specific software vulnerabilities maintained by MITRE. We store 0 entries.

OWASP Top 10

The most critical web application security risks. Each category groups related CWEs.

Go to Dashboard