v1.1.2

Summary

Patch release: Groq model migration and LLM resilience for the public Streamlit demo — moves the cloud default from deprecated Llama models to openai/gpt-oss-20b, auto-migrates legacy Groq IDs (8b → 20b, 70b → 120b), and adds retrieval-only fallback so users still get MITRE/KEV sources when the LLM is unavailable or misconfigured.

Live demo: https://threat-intelligence-rag-assistant.streamlit.app/

---

What's new

Groq migration (proactive)

• Cloud default LLM: openai/gpt-oss-20b (replaces llama-3.1-8b-instant; Groq shutdown 2026-08-16)
• Settings auto-migrate deprecated IDs:
  • llama-3.1-8b-instant → openai/gpt-oss-20b
  • llama-3.3-70b-versatile → openai/gpt-oss-120b
• Updated .env.example, docker-compose.yml, and README tech stack / validation baseline

Retrieval-only fallback (reactive)

• When LLM generation fails after retrieval, return formatted MITRE/KEV sources without synthesis (RETRIEVAL_ONLY_FALLBACK=true by default)
• Streamlit sidebar warning and toast for degraded retrieval-only mode
• Clearer Groq 404 / model unavailable error mapping

Testing and docs

• RAG integration tests — full chain.invoke path verifies answer, sidebar-equivalent sources, and confidence (CI: mocked LLM; optional live Groq when GROQ_API_KEY is set)
• README Resilience bullet under Safety; docs/architecture.md updated for degraded mode and v1.1.2

---

Quick start

Unchanged from v1.1.1 — see README Quick Start.

git clone https://github.com/rvong65/threat-intelligence-assistant.git
cd threat-intelligence-assistant
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env        # Windows: copy .env.example .env
streamlit run app.py

Docker:

git clone https://github.com/rvong65/threat-intelligence-assistant.git
cd threat-intelligence-assistant
docker compose up --build
# open http://localhost:8501

---

Full changelog

See CHANGELOG.md#1.1.2