Add Chaibot client for ship-help MCP integration

cmd/slack-bot/main.go

@@ -34,6 +34,7 @@ import (
 
 	userv1 "github.com/openshift/api/user/v1"
 
+	"github.com/openshift/ci-tools/pkg/chaibot"
 	"github.com/openshift/ci-tools/pkg/jira"
 	"github.com/openshift/ci-tools/pkg/pagerdutyutil"
 	eventhandler "github.com/openshift/ci-tools/pkg/slack/events"
@@ -66,6 +67,9 @@ type options struct {
 	requireWorkflowsInForum bool
 	supportRequestChannelID string
 	supportRequestThreshold int
+
+	enableTriage     bool
+	triageConfigPath string
 }
 
 func (o *options) Validate() error {
@@ -117,6 +121,8 @@ func gatherOptions(fs *flag.FlagSet, args ...string) options {
 	fs.BoolVar(&o.requireWorkflowsInForum, "require-workflows-in-forum", true, "Require the use of workflows in the designated forum channel")
 	fs.StringVar(&o.supportRequestChannelID, "support-request-channel-id", "CBN38N3MW", "Channel ID where support request mode watches long threads (defaults to #forum-ocp-testplatform)")
 	fs.IntVar(&o.supportRequestThreshold, "support-request-threshold", 12, "Create a support-request Jira when a thread has more than this many messages (total count includes the root message)")
+	fs.BoolVar(&o.enableTriage, "enable-triage", false, "Enable Chaibot automatic test failure triage")
+	fs.StringVar(&o.triageConfigPath, "triage-config-path", "", "Path to triage configuration file")
 
 	if err := fs.Parse(args); err != nil {
 		logrus.WithError(err).Fatal("Could not parse args.")
@@ -198,6 +204,57 @@ func main() {
 		}
 	}
 
+	// Initialize Chaibot if enabled
+	var chaibotAnalyzer *chaibot.Analyzer
+	var chaibotChannels []string
+	if o.enableTriage {
+		// Fail fast if required config is missing
+		if o.triageConfigPath == "" {
+			logrus.Fatal("--enable-triage requires --triage-config-path to be set")
+		}
+
+		mcpURL := os.Getenv("SHIP_HELP_MCP_URL")
+		mcpToken := os.Getenv("SHIP_HELP_MCP_TOKEN")
+
+		// Fail fast if required env vars are missing
+		if mcpURL == "" || mcpToken == "" {
+			logrus.Fatal("--enable-triage requires both SHIP_HELP_MCP_URL and SHIP_HELP_MCP_TOKEN environment variables")
+		}
+
+		type triageConfig struct {
+			Enabled           bool `yaml:"enabled"`
+			MonitoredChannels []struct {
+				Name      string `yaml:"name"`
+				ChannelID string `yaml:"channel_id"`
+			} `yaml:"monitored_channels"`
+			Analysis struct {
+				AIProvider     string `yaml:"ai_provider"`
+				PromptTemplate string `yaml:"prompt_template"`
+			} `yaml:"analysis"`
+		}
+
+		configData, err := os.ReadFile(o.triageConfigPath)
+		if err != nil {
+			logrus.WithError(err).Fatal("Failed to read triage config")
+		}
+
+		var cfg triageConfig
+		if err := yaml.Unmarshal(configData, &cfg); err != nil {
+			logrus.WithError(err).Fatal("Failed to parse triage config")
+		}
+
+		chaibotAnalyzer = chaibot.NewAnalyzer(mcpURL, mcpToken, cfg.Analysis.PromptTemplate)
+
+		for _, ch := range cfg.MonitoredChannels {
+			chaibotChannels = append(chaibotChannels, ch.ChannelID)
+		}
+
+		logrus.WithFields(logrus.Fields{
+			"channels": len(chaibotChannels),
+			"provider": cfg.Analysis.AIProvider,
+		}).Info("Chaibot triage enabled")
+	}
+
 	metrics.ExposeMetrics("slack-bot", config.PushGateway{}, o.instrumentationOptions.MetricsPort)
 	simplifier := simplifypath.NewSimplifier(l("", // shadow element mimicing the root
 		l(""), // for black-box health checks
@@ -215,7 +272,7 @@ func main() {
 	// handle the root to allow for a simple uptime probe
 	mux.Handle("/", handler(http.HandlerFunc(func(writer http.ResponseWriter, request *http.Request) { writer.WriteHeader(http.StatusOK) })))
 	mux.Handle("/slack/interactive-endpoint", handler(handleInteraction(secret.GetTokenGenerator(o.slackSigningSecretPath), interactionrouter.ForModals(issueFiler, slackClient))))
-	mux.Handle("/slack/events-endpoint", handler(handleEvent(secret.GetTokenGenerator(o.slackSigningSecretPath), eventrouter.ForEvents(slackClient, issueFiler, kubeClient, configAgent.Config, gcsClient, keywordsConfig, o.helpdeskAlias, o.forumChannelId, o.reviewRequestWorkflowID, o.namespace, o.supportRequestChannelID, o.supportRequestThreshold, o.requireWorkflowsInForum))))
+	mux.Handle("/slack/events-endpoint", handler(handleEvent(secret.GetTokenGenerator(o.slackSigningSecretPath), eventrouter.ForEvents(slackClient, issueFiler, kubeClient, configAgent.Config, gcsClient, keywordsConfig, o.helpdeskAlias, o.forumChannelId, o.reviewRequestWorkflowID, o.namespace, o.supportRequestChannelID, o.supportRequestThreshold, o.requireWorkflowsInForum, chaibotAnalyzer, chaibotChannels))))
 	server := &http.Server{Addr: ":" + strconv.Itoa(o.port), Handler: mux}
 
 	health.ServeReady()

pkg/chaibot/README.md

@@ -0,0 +1,182 @@
+# Chaibot - Ship-Help MCP Integration
+
+This package provides test failure analysis using Chai Bot (ship-help MCP) for automatic Slack triage.
+
+## Overview
+
+Chaibot monitors Slack channels for Prow CI job failure URLs and automatically posts analysis using the Chai Bot service via ship-help MCP.
+
+## Files in This PR
+
+- `pkg/chaibot/analyzer.go` - Ship-help MCP client implementation
+- `pkg/chaibot/analyzer_test.go` - Unit tests
+- `pkg/slack/events/chaibot/handler.go` - Slack event handler (monitors for Prow URLs)
+- `pkg/slack/events/chaibot/handler_test.go` - Event handler tests
+- `pkg/slack/events/router/router.go` - Updated to register Chaibot handler
+- `cmd/slack-bot/main.go` - Updated with Chaibot initialization
+
+**This PR provides the complete implementation.** The integration is ready to use once deployed.
+
+## How It Works
+
+### 1. Event Handler Pattern (Already Implemented)
+
+Chaibot uses the existing event handler pattern in `openshift/ci-tools`:
+
+**Implementation files:**
+- `pkg/slack/events/chaibot/handler.go` - Monitors Slack messages for Prow URLs
+- Registered in `pkg/slack/events/router/router.go`
+- Initialized in `cmd/slack-bot/main.go`
+
+**What the handler does:**
+1. Monitors configured Slack channels (e.g., `#opp-discussion`)
+2. Detects Prow CI job URLs in messages
+3. Calls `analyzer.AnalyzeFailure()` asynchronously
+4. Posts analysis results in a thread
+
+### 2. Initialization in cmd/slack-bot/main.go
+
+**Already implemented in this PR:**
+
+```go
+// Command-line flags (added)
+--enable-triage          // Enable Chaibot
+--triage-config-path     // Path to triage-config.yaml
+
+// Initialization (added to main())
+if o.enableTriage && o.triageConfigPath != "" {
+    mcpURL := os.Getenv("SHIP_HELP_MCP_URL")
+    mcpToken := os.Getenv("SHIP_HELP_MCP_TOKEN")
+
+    // Create analyzer
+    chaibotAnalyzer = chaibot.NewAnalyzer(mcpURL, mcpToken, promptTemplate)
+
+    // Handler is registered in router.ForEvents()
+}
+```
+
+### 3. Event Router Registration
+
+**Already implemented in pkg/slack/events/router/router.go:**
+
+```go
+func ForEvents(client *slack.Client, chaibotAnalyzer *chaibot.Analyzer, chaibotChannels []string, ...) {
+    // ... existing handlers ...
+
+    if chaibotAnalyzer != nil && len(chaibotChannels) > 0 {
+        handlers = append(handlers, chaibothandler.Handler(client, chaibotAnalyzer, chaibotChannels))
+    }
+}
+```
+
+## NOT in This PR (Requires openshift/release Configuration)
+
+The following configuration files are in **openshift/release#80559**, not this PR:
+
+- `core-services/ci-chat-bot/triage-config.yaml` - Chaibot configuration
+- `clusters/app.ci/ci-chat-bot/chaibot-configmap.yaml` - Kubernetes ConfigMap
+- `clusters/app.ci/ci-chat-bot/ci-chat-bot.yaml` - Deployment with environment variables
+- `core-services/ci-secret-bootstrap/chaibot-secret-config.yaml` - Ship-help token secret
+
+## Usage
+
+Once both PRs are merged and deployed:
+
+1. **Post a Prow URL in a monitored channel:**
+   ```
+   Job failed: https://prow.ci.openshift.org/view/gs/test-platform-results/logs/periodic-ci-stolostron-policy-collection-main-ocp4.22-interop-opp-aws/2066255424226594816
+   ```
+
+2. **Chaibot responds in a thread within 30-60 seconds** with:
+   - Which step(s) failed
+   - Root cause analysis (product bug, test issue, or infrastructure)
+   - Related Jira tickets
+   - Pass rate history
+   - Recommended fixes
+
+## Configuration
+
+**Deployment configuration is in openshift/release#80559:**
+
+- `core-services/ci-chat-bot/triage-config.yaml` - Main config:
+  - Monitored channels (e.g., `#opp-discussion`)
+  - Ship-help MCP endpoint
+  - Analysis prompt template
+  - Rate limiting settings
+
+- `clusters/app.ci/ci-chat-bot/ci-chat-bot.yaml` - Deployment:
+  - Environment variables: `SHIP_HELP_MCP_URL`, `SHIP_HELP_MCP_TOKEN`
+  - ConfigMap mount: `/etc/triage-config/triage-config.yaml`
+
+## How to Enable Chaibot
+
+Chaibot is enabled via **command-line flags** (not environment variables):
+
+**Command-line flags (required):**
+- `--enable-triage` - Enable Chaibot functionality
+- `--triage-config-path=/etc/triage-config/triage-config.yaml` - Path to config file
+
+**Environment variables (required):**
+- `SHIP_HELP_MCP_URL` - Ship-help MCP endpoint
+- `SHIP_HELP_MCP_TOKEN` - Authentication token (from Kubernetes secret)
+
+**Example deployment command:**
+```yaml
+# In clusters/app.ci/ci-chat-bot/ci-chat-bot.yaml
+args:
+  - --enable-triage
+  - --triage-config-path=/etc/triage-config/triage-config.yaml
+env:
+  - name: SHIP_HELP_MCP_URL
+    value: "https://ship-help-mcp-continuous-release-tooling--ship-help-bot.apps.gpc.ocp-hub.prod.psi.redhat.com/personas/ocp_ai_helpdesk/mcp"
+  - name: SHIP_HELP_MCP_TOKEN
+    valueFrom:
+      secretKeyRef:
+        name: cluster-secrets-chaibot-ship-help
+        key: ship-help-token
+```
+
+**Without these flags, Chaibot will NOT activate** - even if environment variables are set.
+
+## Related PRs
+
+- **This PR (openshift/ci-tools#5251)** - Chaibot implementation (analyzer, handler, router, main.go)
+- **openshift/release#80559** - Configuration and deployment (config files, secrets, ConfigMaps)
+- Based on `/analyze-failure` skill by MPEX Integrity team
+- Alternative to PR openshift/release#80476 (OpenAI approach)
+
+## Architecture
+
+```
+User posts Prow URL in Slack
+         ↓
+Slack Event API → ci-chat-bot deployment
+         ↓
+pkg/slack/events/chaibot/handler.go
+    - Detects Prow URL
+    - Extracts job URL
+         ↓
+pkg/chaibot/analyzer.go
+    - Calls ship-help MCP (ask_persona tool)
+    - Sends prompt with job URL
+         ↓
+Ship-Help MCP (ocp_ai_helpdesk persona)
+    - Searches Jira, Sippy, Prow logs
+    - Analyzes failure
+    - Returns comprehensive analysis
+         ↓
+pkg/chaibot/analyzer.go
+    - Formats response as Slack Block Kit
+         ↓
+Slack API
+    - Posts analysis in thread
+```
+
+## Cost Comparison
+
+| Solution | Cost | Data Sources |
+|----------|------|--------------|
+| **Chaibot (ship-help MCP)** | $0/month | 9+ sources (Jira, Sippy, Prow, GitHub, etc.) |
+| OpenAI GPT-4o (PR #80476) | ~$1,080/year | 3 sources (limited context) |
+
+**Chaibot uses internal Red Hat infrastructure** - no external API costs.