How do I switch from the default Phi-2 model to SecBERT Instructional? #2
Answered
by
netdevs-net
netdevs-net
asked this question in
Q&A
|
README mentions LLM integration defaults to Phi-2 quantized GGUF, with SecBERT Instructional available via LLM_MODEL env var. What's the exact env var value/config to switch, and does it need a different model file mounted or does it pull automatically? |
Answered by
netdevs-net
Aug 5, 2026
Replies: 1 comment
|
Set the env var LLM_MODEL=secbert (case-insensitive) before starting the stack — it defaults to phi2 otherwise. See metAIsploit_assistant/actions/chat.py:158. No manual download needed: SecBERT-Instructional pulls automatically from HuggingFace (jackaduma/SecBERT-Instructional) since its file_location is None, unlike Phi-2 which needs the local GGUF file at /app/models/phi-2.Q4_K_M.gguf (fetched via scripts/download_phi2_gguf.py). |
0 replies
Answer selected by
netdevs-net
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Set the env var LLM_MODEL=secbert (case-insensitive) before starting the stack — it defaults to phi2 otherwise. See metAIsploit_assistant/actions/chat.py:158. No manual download needed: SecBERT-Instructional pulls automatically from HuggingFace (jackaduma/SecBERT-Instructional) since its file_location is None, unlike Phi-2 which needs the local GGUF file at /app/models/phi-2.Q4_K_M.gguf (fetched via scripts/download_phi2_gguf.py).