Skip to content

Go SDK

158 methods for Go backends. Covers the core eval / security / gateway / trace surface; the TypeScript SDK has the broadest method coverage.

go get github.com/EvalGuardAi/evalguard-go

Security — the published module is affected (SEC-051)

Every published version up to and including v1.6.0 — what a bare go get resolves to today — follows HTTP redirects on the guardrail-verdict transport, because net/http follows whenever CheckRedirect is nil and the client never set it. A 302 anywhere on the API path (an nginx trailing-slash rule or a load balancer is enough — no attacker required) rewrites the check to a bodyless GET: the screened text is never transmitted, and the reply comes back as an authoritative allow. On 307/308 the body survives and the screened text is re-POSTed to the redirect target; a same-hostname/different-port hop also forwards your Authorization header. This is scoped to the redirect path: v1.6.0's verdict guards are correct on malformed or truncated bodies, but a redirected reply is well formed — an answer about nothing — so shape checks cannot see it. No fixed tag is on proxy.golang.org yet. Until one is, do not rely on the firewall methods to block anything: configure a base URL that is the exact path the API serves (no trailing slash, no apex-to-www hop) and terminate TLS at an origin you control. Questions: security@evalguard.ai.

Installation

terminal
go get github.com/EvalGuardAi/evalguard-go

Quick Start

main.go
package main

import (
  "context"
  "fmt"
  "log"

  "github.com/EvalGuardAi/evalguard-go"
)

func main() {
  client, err := evalguard.NewClient("eg_live_your_key",
    evalguard.WithBaseURL("https://evalguard.ai/api/v1"),
  )
  if err != nil {
    log.Fatal(err)
  }

  // Run an evaluation (async — returns a run handle)
  result, err := client.RunEval(context.Background(), &evalguard.RunEvalRequest{
    Name:      "regression-suite",
    ProjectID: "proj_abc123",
    Model:     "gpt-4o",
    Prompt:    "Answer concisely: {{input}}",
    Cases:     []evalguard.EvalCase{{Input: "2+2?", ExpectedOutput: "4"}},
    Scorers:   []string{"exact-match"},
  })
  if err != nil {
    log.Fatal(err)
  }
  // RunEval is async; poll client.GetEval(ctx, result.ID) for results
  fmt.Printf("Run %s started: %s\n", result.ID, result.Status)
}

Configuration

main.go
// Custom timeout
client, _ := evalguard.NewClient("eg_...",
  evalguard.WithTimeout(60 * time.Second),
)

// Custom HTTP client (for proxies, mTLS, etc.)
client, _ := evalguard.NewClient("eg_...",
  evalguard.WithHTTPClient(&http.Client{
    Transport: customTransport,
  }),
)

Common Methods (68 of 158)

Evaluations

4
  • RunEval()
  • GetEval()
  • ListEvals()
  • ListEvalRuns()

Security

5
  • RunSecurityScan()
  • GetSecurityGraders()
  • GetSecurityEffectiveness()
  • GetSecurityReport()
  • CodeScan()

Shadow AI

1
  • AnalyzeShadowAI()

AI-SPM

1
  • GetAIPosture()

Smart Copilot

1
  • AnalyzeCopilot()

Gateway

2
  • GetGatewayHealth()
  • GetGatewayStats()

Traces

5
  • GetTraces()
  • GetTrace()
  • SearchTraces()
  • CreateTrace()
  • IngestOTLP()

Cost / FinOps

6
  • GetCost()
  • GetCostForecast()
  • GetCostSavings()
  • GetCostBudget()
  • GetCostAnomalies()
  • GetCostRecommendations()

Monitoring

4
  • GetMonitoringAlerts()
  • GetMonitoringDrift()
  • GetMonitoringAnalytics()
  • GetMonitoringSLA()

Compliance

5
  • GetCompliance()
  • CheckCompliance()
  • GetComplianceGaps()
  • GetEUAIAct()
  • GetModelCards()

Prompts

2
  • CreatePrompt()
  • ListPrompts()

Datasets

2
  • CreateDataset()
  • ListDatasets()

Firewall & Guardrails

3
  • ListFirewallRules()
  • ListGuardrails()
  • GenerateGuardrails()

Team & Org

4
  • ListTeam()
  • ListApiKeys()
  • ListWebhooks()
  • GetAuditLogs()

Other

23
  • Ask()
  • FormalVerify()
  • GetLeaderboard()
  • SubmitTicket()
  • GetThreatIntelligence()
  • GetAISBOM()
  • GenerateAISBOM()
  • GetSIEMConnectors()
  • GetSettings()
  • ListNotifications()
  • ListTemplates()
  • GetMarketplace()
  • ListEvalSchedules()
  • ListIncidents()
  • GetDashboardStats()
  • DetectDrift()
  • SmartRoute()
  • GetAutopilotConfig()
  • ListPipelines()
  • ListAnnotations()
  • CreateAnnotation()
  • Search()
  • ListTickets()

Error Handling

errors.go
// Typed errors for precise handling
result, err := client.RunEval(ctx, req)
if err != nil {
  var authErr *evalguard.AuthError
  var rlErr *evalguard.RateLimitError

  switch {
  case errors.As(err, &authErr):
    log.Fatal("Invalid API key:", authErr.Message)
  case errors.As(err, &rlErr):
    log.Printf("Rate limited, retry after %v", rlErr.RetryAfter)
    time.Sleep(rlErr.RetryAfter)
  default:
    log.Fatal("API error:", err)
  }
}

Shadow AI Detection

shadow.go
result, err := client.AnalyzeShadowAI(ctx, &evalguard.ShadowAIRequest{
  Input:    "My SSN is 123-45-6789 and CC 4111-1111-1111-1111",
  Provider: "openai",
  Model:    "gpt-4o",
})
// result.PIIDetails["types"] = ["ssn", "creditCard"]
// result.Event["riskScore"] = 25

AI-SPM (Security Posture)

spm.go
posture, err := client.GetAIPosture(ctx, "project-id")
// posture.Posture.OverallScore = 85
// posture.Models[0]["name"] = "gpt-4o"
// posture.Models[0]["misconfigurations"] = [...]
// posture.DataFlows[0]["crossBorder"] = true

Smart Copilot

copilot.go
analysis, err := client.AnalyzeCopilot(ctx, &evalguard.CopilotAnalyzeRequest{
  Type:     "security",
  Model:    "gpt-4o",
  PassRate: 0.6,
  Findings: []map[string]any{
    {"type": "prompt-injection", "severity": "critical", "title": "Injection", "description": "...", "passed": false},
  },
})
// analysis.Analysis["overallRisk"] = "critical"
// analysis.Analysis["immediateActions"] = [...]
// analysis.Analysis["complianceImpact"] = [{framework: "GDPR", status: "at-risk"}]