By Sagar Shankaran, Founder of CallSphere
Every PR that touches a prompt, tool, or model should run the eval suite and block on regression. Here is the GitHub Actions setup we ship across 37 agents.
Key takeaways
TL;DR — If your agent eval doesn't run in CI, it doesn't run. Wire your golden set into GitHub Actions, post the diff as a PR comment, block merge on regression. By April 2026, six platforms (LangSmith, Langfuse, Arize, Helicone, Datadog, Honeycomb) all support this — pick one and ship.
Without a CI gate, three things happen:
LangSmith's regression suite pattern: 100–500 test cases per candidate prompt or model, aggregate scores per PR, gate merges on threshold. Fast (parallel), automatic, visible — that's the bar.
flowchart LR
A[PR Opened] --> B[GitHub Actions]
B --> C[Run Eval Suite]
C --> D[Scores]
D --> E[Compare to Main]
E -->|regression| F[Block Merge]
E -->|pass| G[Allow Merge]
D --> H[PR Comment]
H --> I[Reviewer Sees Diff]
The CI eval gate has three parts: (1) the eval runs on every PR (or every push to main if you do trunk-based), (2) results post as a PR comment with deltas vs main, (3) required status check blocks merge below threshold.
Gates we use: pass-rate must be within -2 points of main; no P0 case may flip from pass to fail; latency p95 must be within +20%; cost-per-request must be within +15%.
CallSphere ships 37 agents · 90+ tools · 115+ DB tables · 6 verticals. Every PR runs a vertical-aware eval suite: touch the Healthcare prompt, the 312-case healthcare set runs (~9 minutes parallel). Touch a shared library, all 6 vertical sets run (~22 minutes). Promptfoo as the harness, GitHub Actions as the runner, results posted as a sticky PR comment.
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
Required checks: pass-rate, P0 flip count, latency p95, cost-per-request, hallucination rate. Pricing $149 / $499 / $1499 · 14-day trial · 22% affiliate.
.github/workflows/agent-evals.yml triggers on pull_request.main post-merge, run the full set including slow audio probes.How do I keep CI fast? Parallelize, cache, run only the changed-domain suite per PR.
What's a P0 case? Compliance-critical (HIPAA leak, refund-policy violation), or known historical incident.
Should I run evals on every commit? Smoke set yes; full set on PR open and main.
What about LLM-judge cost in CI? Use a smaller judge for CI (Haiku, GPT-5-mini), bigger judge for nightly.
Does the trial include the eval gate? Trial tenants run on our shared eval infra; Pro+ pricing gets a dedicated suite. See the demo.
CI/CD Eval Gates in 2026: Failing PRs on Regression for Voice and Chat Agents forces a tension most teams underestimate: agent handoff state. A single LLM call is easy. A booking agent that hands a confirmed slot to a billing agent that hands a follow-up to an escalation agent — that's where context loss, hallucinated IDs, and double-bookings live. Solving it well means treating the conversation as a stateful workflow, not a chat.
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
Production AI agents live or die on three loops: evals, retries, and handoff state. CallSphere runs 37 agents across 6 verticals, each with its own eval suite — synthetic call transcripts replayed nightly with assertion checks on extracted entities (date, time, party size, insurance, address). Without that loop, prompt regressions ship silently and you only find out when bookings drop.
Structured tools beat free-form text every time. Our 90+ function tools all enforce JSON schemas validated server-side; if the model hallucinates an integer where a string is required, we retry with a corrective system message before falling back to a deterministic path. For long-running flows, we treat agent handoffs as a state machine — booking → confirmation → SMS — so context survives turn boundaries.
The Realtime API vs. async decision usually comes down to "is the user holding the phone right now?" If yes, Realtime; if no (callback queue, after-hours voicemail), async wins on cost-per-conversation, which we track per agent in 115+ database tables spanning all 6 verticals.
How does this apply to a CallSphere pilot specifically?
Real Estate runs as a 6-container pod (frontend, gateway, ai-worker, voice-server, NATS event bus, Redis) backed by Postgres realestate_voice with row-level security so multi-tenant data never crosses tenants. For a topic like "CI/CD Eval Gates in 2026: Failing PRs on Regression for Voice and Chat Agents", that means you're not starting from scratch — you're configuring an agent template that's already been hardened across thousands of conversations.
What does the typical first-week implementation look like? Day one is integration mapping (scheduler, CRM, messaging) and prompt tuning against your top 20 real call transcripts. Day two through five is shadow-mode running, where the agent transcribes and recommends but a human still answers, so you can compare side-by-side. Go-live is the moment your eval pass-rate clears your internal bar.
Where does this break down at scale? The honest answer: it scales until your tool catalog gets stale. The agent is only as good as the integrations it can actually call, so the operational discipline is keeping schemas, webhooks, and fallback paths green. The platform handles the rest — observability, retries, multi-region routing — without your team owning the GPU layer.
Want to see how this maps to your stack? Book a live walkthrough at calendly.com/sagar-callsphere/new-meeting, or try the vertical-specific demo at salon.callsphere.tech. 14-day trial, no credit card, pilot live in 3–5 business days.
Written by
Sagar Shankaran· Founder, CallSphere
Sagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
How the modern agent eval stack actually flows: instrument, trace, dataset, evaluator, score, CI gate. The full pipeline that keeps agents from regressing.
Version your prompts in git, run a 50-case eval suite on every PR, block merges below threshold, and ship a new agent prompt with confidence — full GitHub Actions tutorial.
Standard benchmarks miss agent regressions because they grade only final outputs. Trajectory-aware evals in CI catch the 20–40% of regressions that single-turn scoring hides.
WebArena 2.0 brings real-browser tasks and harder evaluation conditions for browsing agents. The benchmark numbers and what they mean for real production browsing builds.
Braintrust positioned itself as the eval platform for serious AI teams in 2026. Datasets, scorers, and CI integrations in practice with concrete pricing and trade-offs.
Langfuse's April 2026 release ships online evals, prompt versioning, and dataset workflows. Why self-hosted observability is worth the operational lift in 2026 builds.
© 2026 CallSphere LLC. All rights reserved.