PrismGuard - Self-hosted LLM guard — auditable blocks, not scores
by•
PrismGuard is an open-source firewall for LLM apps. Block prompt injection before it hits your model — every allow/block returns resolution_gate (which rule decided), not a probability score your team argues about in Slack.
Rules-first: pip install prismguard. Optional local ONNX. Apache-2.0. Cold holdout: 14/14 vs LLM Guard 9/14.
Wire at agent entry, RAG chunks, or sidecar. v0.1.6 on PyPI — alpha, feedback welcome.
Replies
Hey Product Hunt 👋 — I'm Amin Parva, designer and builder of ChorusGraph.
THE PROBLEM
Most teams building AI agents start with LangGraph or CrewAI — then spend months
bolting on cache, RAG, memory, and audit logs. Same questions get re-answered.
Same tokens get billed twice. And when something goes wrong, nobody can trace
which step caused it.
WHAT CHORUSGRAPH IS
An agentic AI builder for Python — one pip install, full runtime:
→ Agent workflows (nodes, routes, multi-agent graphs)
→ Semantic cache built in — stop paying for repeat questions
→ Route Ledger — audit every hop your agent took
→ Search & memory ports — swappable, not a second project
→ Native engine — not a LangGraph wrapper
TRY IT IN 30 SECONDS
pip install chorusgraph
chorusgraph-demo
NEW IN V1.0.3 — FREE SAVINGS CHECK (no API key):
chorusgraph-audit --log your_queries.jsonl
Upload your query log → estimated cache hit rate before you touch production.
BENCHMARK VS LANGGRAPH (fair: same model, same tools, only framework changed)
| Scenario | LangGraph | ChorusGraph |
|-------------------------|-----------|-------------|
| Finance single-agent | 87.5% | 100% |
| Finance multi-agent | 75% | 87.5% |
| Healthcare single-agent | 72.5% | 72.5% |
| Healthcare multi-agent | 57.5% | 87.5% |
Overall: ~44% fewer LLM calls · ~35% lower modeled cost
Full methodology on GitHub (open, reproducible).
HONEST CAVEATS
• LangGraph appears in our repo only as a benchmark baseline — not in the product path
• Healthcare cache re-runs clinical judgment on hits (safety by design)
• Vector RAG is optional; keyword retrieval works out of the box
LINKS
🔗 Product page: https://www.insightits.com/produ...
📦 PyPI: https://pypi.org/project/chorusg...
💻 GitHub: https://github.com/insightitsGit...
📖 Install guide: https://github.com/insightitsGit...
Open source · Apache 2.0 · Python 3.11+ · 357 tests, no live API key required
I'd love your feedback:
→ What agent framework are you on today?
→ What's the one thing you wish was built in?
Happy to answer every question today. Thanks for checking us out 🙏