Reliai Demo

Explore a realistic AI reliability workflow in under two minutes.

Simulated failure

Hallucination spike detected

A prompt update introduced hallucinated responses. Reliai detected the regression, opened an incident, and recommended a guardrail before users noticed.

Demo scenario

AI Support Copilot serves production support traffic.

Hallucination spike detected appeared after a prompt rollout and surfaced before end-user complaints arrived.

This walkthrough shows the full operator loop: detect, explain, mitigate, and decide on rollout safety.

01

System health

Start at the control panel.

Reliai system status page

AI reliability control panel

AI Support Copilot

Default status page for this AI system. It answers what is happening, whether it is safe, and where an operator should click next.

Is this system safe right now?

Answer: MAYBE

This AI system needs review before the next change.

The system is stable enough to operate, but current signals show elevated reliability risk.

System Health

Reliability score

92

Active incidents

1

Guardrails protecting

17

Traffic

Traces analyzed (24h)

2.3M

Throughput

27

traces/sec · 1m avg

Active services

6

System status

What needs attention next

Latest deployment

Today

Risk score 0.24

Incident pressure

1 incidents / 24h

Latest: Hallucination spike after retriever prompt rollout

Guardrail pressure

17 triggers / 24h

Top policy: structured_output

Deployment risk

Safety before the next rollout

Risk levellow
Risk score0.24
Simulation riskmedium

Guardrail activity

Runtime protection coverage

structured output11
latency retry4
cost budget2

Policy compliance

Organization guardrail coverage

structured output

Mode: enforce

98.0%

Violations last 24h3

cost budget

Mode: warn

96.0%

Violations last 24h5

latency retry

Mode: enforce

94.0%

Violations last 24h2

Recommended next step

Operator guidance

Add hallucination guardrail to retriever prompt

structured_output -> Add hallucination guardrail to retriever prompt. for gpt-4.1

Add retry policy for retrieval failures

latency_retry -> Retry retrieval failures once before fallback

02

Incident

Responses referencing non-existent documentation

Reliai incident command center

Incident command center

Retrieval latency regression after prompt rollout

AI Support Copilot · retrieval_latency_ms · opened Today

highopen

Likely root cause

What probably broke

Prompt rollout changed retrieval behavior

Confidence 71%

Graph related patterns

Prompt rollout expanded retrieval context from 4 to 6 chunks

Retrieval timeout under long-context prompts

Global platform failures

Prompt rollout changed retrieval behavior · 71%

Retrieval stage is exceeding expected latency · 62%

Similar long-context retrieval failures seen elsewhere · 48%

Deployment changes

gpt-4.1-mini

support/refund-v42

82 min before incident

Guardrail triggers

structured_output · 11 triggers

latency_retry · 4 triggers

Recommended mitigation

What the operator should do next

Enable latency retry on retrieval and compare v42 against the previous prompt before expanding rollout.

Recommended guardrails

Enable structured output validation on the full response path

Add retry policy for retrieval failures before fallback

Pause rollout of prompt version support/refund-v42

Rollback suggestion

A linked deployment was active near incident start. Verify the rollout and consider rollback if the trace compare confirms regression.

Model change risk

gpt-4.1-mini is part of the deployment context and should be treated as a candidate change vector.

Roll back prompt v42 after enabling retrieval retry

The likely change vector is the prompt rollout. Restore the previous prompt, enable latency retry, then compare baseline vs failing traces before redeploying.

Incident status

Scope · environment:production

Window · 30 min

Owner · owner@acme.test

Ack · Today

Deployment context

Environment · production

Deployed · Today

Time to incident · 82 min

Prompt · support/refund-v42

Model · gpt-4.1-mini

Guardrail activity

structured_output

11 triggers

Last trigger · Today

latency_retry

4 triggers

Last trigger · Today

03

Trace graph

Slowest span: retrieval · Token heavy span: llm_call

Reliai trace debugger

Execution graph

trace_demo_agent_94f3

AI system debugging view for one request across retrieval, prompt construction, model execution, tool calls, guardrails, and post-processing.

Spans

6

Edges

5

Environment

production

Trace analysis

What stands out in this request

Slowest step

retrieve_context

980 ms

Largest token consumer

answer_customer

3840 tokens

Guardrail retries

structured_output

1 retries

Estimated cost surface

6940 tokens

Across all recorded spans in this trace graph

Span legend

retrieval
prompt build
llm call
tool call
postprocess
guardrail

Execution breakdown

Span tree

ai_request

request

gpt-4.1-mini

span span_root · root span

Success2310 ms

retrieve_context

retrieval

pgvector

span span_retrieval · parent span_root

Success980 ms
latency_retry · retrySlowest span

build_prompt

prompt build

prompt-assembler

span span_prompt · parent span_root

Success180 ms3100 tokens

answer_customer

llm call

gpt-4.1-mini

span span_llm · parent span_root

Success760 ms3840 tokens
structured_output · retryLargest token spanGuardrail retry span

postprocess_answer

postprocess

response-normalizer

span span_post · parent span_root

Success150 ms

lookup_order_status

tool call

order-service

span span_tool · parent span_llm

Success240 ms

04

Root cause explanation

Prompt update deployed earlier today

Likely cause

Prompt update deployed earlier today

The failing trace shows retrieval pressure first, then a model response that required guardrail retry. The deployment window lines up with the incident start.

Failure surface

Responses referencing non-existent documentation

Linked signal

Prompt update deployed 82 minutes before incident start.

05

Guardrail recommendation

Recommended runtime protections before blast radius expands.

enforce

Structured output policy

Schema validation catches malformed tool responses before they reach users.

warn

Latency retry policy

Retrieval retry cushions transient upstream failures during traffic spikes.

06

Deployment safety

Finish at the rollout gate.

Reliai deployment gate

Deployment detail

support/refund-v42 · gpt-4.1-mini

production · current release window

Deployment Safety Check: WARNING
release-bot@acme.test

Deployment metadata

Change record

Prompt version

support/refund-v42

Model version

gpt-4.1-mini

{
  "rollout": "progressive 20%",
  "release_reason": "refund handling improvement",
  "linked_simulation": "sim_demo_v42"
}

Deployment safety check

Gate decision

Safety state

WARNING

Risk score: 72/100

Deployment risk factors

  • High regression probability in retrieval latency
  • Similar failures seen across organizations
  • Insufficient guardrail coverage for retrieval retries
  • Recent incident correlation detected during rollout

AI reliability insights

Known failure patterns

Known reliability patterns detected

  • Similar retrieval latency failures have been seen across other deployments with expanded context windows.
  • Current rollout has insufficient retry coverage for retrieval failures.

Retrieval timeout under long-context prompts

high risk · 142 traces

Recommended guardrails

enable latency_retryenable structured_output

Deployment risk factors

Why this change is safe or risky

High regression probability

Current deployment signals indicate elevated regression risk.

Cross-organization failures

Similar failure patterns have been seen in the reliability graph.

Guardrail coverage

Additional guardrail coverage is recommended before rollout.

Recent incident correlation

1 linked incident already reference this deployment.