Example proposal: deterministic physics/bounds guardrail (no-LLM, recomputable verdicts) alongside existing guardrails
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Newbie friendliness
- 66/100
- Issue type
- Feature
- Clarity
- Mostly clear
- Activity status
- Active
- Tech stack
- python
- Domain
- ai, documentation
Research direction
Read the existing examples layout and inspect crewai/utilities/guardrail.py and tasks/llm_guardrail.py to see how task guardrails are used. Add an example that applies the proposed deterministic physics checker to a task and shows both a caught violation and a clean pass. Done means the example follows the repository's conventions and demonstrates both outcomes.
Written by the indexing model from the issue text.
Description
Hi team — crewAI's per-task guardrails (crewai/utilities/guardrail.py, tasks/llm_guardrail.py) are a clean hook for output checking. Proposal: an example guardrail that validates outputs against deterministic physics/bounds rules — no LLM in the loop, verdicts recomputable from the public, versioned rule table (phm-lite-check).
What it catches: probability 1.8, negative mass, speed above c, below-absolute-zero temperatures, NaN/Inf in numeric fields. Content-safety guardrails catch semantics; this catches physics-invalid values — complementary, and it composes with your existing guardrail API (guardrail: str | Callable).
Several rules were corrected following review comments in microsoft/PyRIT#3507 (negative binding energies are legal; growth percentages are not shares; signed velocities are legal; malformed input must not crash the gate) - the full thread is public. The checker also ships as a Dify Marketplace plugin and an MCP-registry-listed package, and is smoke-tested on nemo-relay==0.11.1 (3-case pipeline test, public repo).
Ask: would a PR adding an examples entry (physics guardrail on a task, showing a caught violation and a clean pass) be welcome? Happy to follow your example layout.
— PHM-Audit
- Dominant language
- Python
- Stars
- 60.4k
- Forks
- 9k
- Avg merge
- 1d 4h
- Merged PRs (30d)
- 210
Getting set up
- No Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from crewAIInc/crewAI
-
feature-request
Difficulty 2/5 1-3 hours Newbie friendliness 66/100
crewAIInc/crewAI#9200 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
crewAIInc/crewAI#9132 · 5 comments ·
Maintainers usually reply within 1 day
-
[BUG] Scrape tools reject the documented cookies mapping and drop cookies when the URL is passed at runtimePossibly taken @Luckyou-d claimed this 3 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
Maintainers usually reply within 1 day
-
[BUG] OpenAI-compatible embedding model alias breaks embedder constructionPossibly taken @gokay-ai claimed this 10 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 65/100
crewAIInc/crewAI#8761 · 1 comment ·
Maintainers usually reply within 1 day
-
[FEATURE] Add api_base_url to TavilySearchTool and TavilyExtractorToolPossibly taken @jayfeng7 claimed this 11 days ago. Open
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
crewAIInc/crewAI#8635 · 1 comment ·
Maintainers usually reply within 1 day
All issues in crewAIInc/crewAI
Similar issues
-
[Bug] vllm-sr config migrate drops providers.models[].deploymentPossibly taken @Bevisy claimed this today. Openneeds-acceptance wg/developer-experience-ecosystem
Difficulty 2/5 1-3 hours Newbie friendliness 86/100
vllm-project/semantic-router#5996 ·
Maintainers usually reply within 2 days
-
bugfix low-priority plugin:abstract plugin:leyline small-effort source:discussion
Difficulty 2/5 1-3 hours Newbie friendliness 66/100
athola/claude-night-market#1132 ·
-
Difficulty 2/5 1-3 hours Newbie friendliness 77/100
WeblateOrg/language-data#2773 ·
Maintainers usually reply within 1 day
-
kind/bug priority/medium qa_automation
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
yugabyte/yugabyte-db#37324 ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
Niko1221/Strata#5553 · 2 comments ·
Maintainers usually reply within 4 days