Proposal: Governance Component for Haystack Pipelines

Problem

Haystack excels at building production RAG and agent pipelines, but lacks built-in governance primitives for:

• Policy enforcement — Cap token usage, restrict tool calls, block dangerous content patterns
• Trust-gated routing — Verify agent identity before allowing pipeline transitions
• Audit trails — Tamper-evident logging of all pipeline actions with Merkle chain hashing

What we'd contribute

A haystack-agentmesh integration package providing Haystack components:

from haystack import Pipeline
from haystack_agentmesh import GovernancePolicyChecker, TrustGate, AuditLogger

pipeline = Pipeline()
pipeline.add_component("policy_check", GovernancePolicyChecker(policy="policy.yaml"))
pipeline.add_component("trust_gate", TrustGate(min_score=0.7))
pipeline.add_component("audit", AuditLogger(merkle_chain=True))

• GovernancePolicyChecker — @component that validates actions against governance policies before execution
• TrustGate — @component that routes based on agent trust scores (pass/fail/review)
• AuditLogger — @component with tamper-evident Merkle chain logging

Context

This is backed by AgentMesh (1,600+ tests). We've shipped similar integrations merged into:

• Dify (65K stars) — Merged
• LlamaIndex (47K stars) — Merged
• Microsoft Agent-Lightning (15K stars) — Merged
• LangGraph — Published langgraph-trust on PyPI
• OpenClaw — Published on ClawHub

Happy to build this as a standalone haystack-agentmesh package following Haystack's integration patterns. Would love maintainer feedback on the approach.