feat: route chatd provider traffic through aibridge (#25629)

## Summary

Routes chatd model calls backed by concrete AI Provider rows through the
in-process aibridge transport by default, with deployment options to use
direct provider routing when AI Gateway is disabled or chat AI Gateway
routing is disabled.

- Splits model routing into common, direct provider, and AI Gateway
paths behind a single deployment-mode entry point.
- Builds chatd models through explicit request, route, and options data.
Active API key attribution is passed explicitly instead of being hidden
inside generic model construction.
- For AI Gateway BYOK routes, resolves the user's provider key in chatd,
forwards it through provider-specific auth headers, and sets
`X-Coder-AI-Governance-Token` to the `delegated` marker so aibridge
preserves those headers while still stripping Coder-specific metadata.
- Keeps central provider credentials and deployment fallback credentials
out of forwarded provider auth headers, so AI Gateway central policy
remains authoritative.
- Redacts delegated provider auth from default string formatting to
avoid accidental plaintext logging of user BYOK credentials.
- Covers selected chat models, advisor overrides, title and quickgen
paths, subagent overrides, computer use model selection, and an
integration-style chat turn through the aibridge transport path.
- Persists initiating API key IDs on chat and queued user messages,
including subagent child messages, and fails closed for AI
Gateway-routed model builds without an active key.
- Removes unused `api_key_id` indexes while keeping the persistence
columns and foreign keys.
- Keeps the deployment option available through config and env parsing,
but hides it from CLI help and generated docs.
- Stabilizes the subagent poll fallback test so background CreateChat
processing cannot win the state transition under slower CI environments.

## Tests

- `go test ./coderd/x/chatd -run
'TestAIGatewayProviderAuthForUser|TestAIGatewayProviderAuthRedactsFormatting|TestResolveModelRouteForConfigAIGatewayProviderAuth|TestAIGatewayModelForwardsProviderAuth|TestProcessChat_AIGatewayRoutingUsesDelegatedAPIKey|TestAwaitSubagentCompletion'
-count=1`
- `go test ./coderd/aibridged -run
'TestServeHTTP_DelegatedAPIKey|TestServeHTTP_StripCoderToken' -count=1`
- `git diff --check HEAD~1..HEAD`
- `make lint`

> Mux working on behalf of Mike.

coderd/x/chatd/quickgen.go

@@ -4,7 +4,6 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"net/http"
 	"slices"
 	"strings"
 	"time"
@@ -69,9 +68,39 @@ var preferredTitleModels = []struct {
 type shortTextCandidate struct {
 	provider string
 	model    string
+	route    resolvedModelRoute
 	lm       fantasy.LanguageModel
 }
 
+func (p *Server) preferredShortTextCandidates(
+	chat database.Chat,
+	keys chatprovider.ProviderAPIKeys,
+) []shortTextCandidate {
+	if p.shouldUseAIGatewayRouting() {
+		return nil
+	}
+
+	candidates := make([]shortTextCandidate, 0, len(preferredTitleModels)+1)
+	userAgent := chatprovider.UserAgent()
+	extraHeaders := chatprovider.CoderHeaders(chat)
+	for _, candidate := range preferredTitleModels {
+		model, err := chatprovider.ModelFromConfig(
+			candidate.provider, candidate.model, keys, userAgent,
+			extraHeaders,
+			nil,
+		)
+		if err == nil {
+			candidates = append(candidates, shortTextCandidate{
+				provider: candidate.provider,
+				model:    candidate.model,
+				route:    newDirectModelRoute(candidate.provider, keys),
+				lm:       model,
+			})
+		}
+	}
+	return candidates
+}
+
 func selectPreferredConfiguredShortTextModelConfig(
 	configs []database.ChatModelConfig,
 ) (database.ChatModelConfig, bool) {
@@ -121,7 +150,9 @@ func (p *Server) maybeGenerateChatTitle(
 	fallbackProvider string,
 	fallbackModelName string,
 	fallbackModel fantasy.LanguageModel,
+	fallbackRoute resolvedModelRoute,
 	keys chatprovider.ProviderAPIKeys,
+	modelOpts modelBuildOptions,
 	generatedTitle *generatedChatTitle,
 	logger slog.Logger,
 	debugSvc *chatdebug.Service,
@@ -135,10 +166,11 @@ func (p *Server) maybeGenerateChatTitle(
 	titleCtx, cancel := context.WithTimeout(ctx, 30*time.Second)
 	defer cancel()
 
-	overrideConfig, overrideModel, overrideKeys, overrideSet, overrideErr := p.resolveTitleGenerationModelOverride(
+	overrideConfig, overrideModel, _, overrideRoute, overrideSet, overrideErr := p.resolveTitleGenerationModelOverride(
 		titleCtx,
 		chat,
 		keys,
+		modelOpts,
 	)
 	if overrideErr != nil {
 		if overrideSet {
@@ -161,29 +193,15 @@ func (p *Server) maybeGenerateChatTitle(
 		candidates = []shortTextCandidate{{
 			provider: overrideConfig.Provider,
 			model:    overrideConfig.Model,
+			route:    overrideRoute,
 			lm:       overrideModel,
 		}}
 	} else {
-		// Build candidate list: preferred lightweight models first,
-		// then the user's chat model as last resort.
-		candidates = make([]shortTextCandidate, 0, len(preferredTitleModels)+1)
-		for _, c := range preferredTitleModels {
-			m, err := chatprovider.ModelFromConfig(
-				c.provider, c.model, keys, chatprovider.UserAgent(),
-				chatprovider.CoderHeaders(chat),
-				nil,
-			)
-			if err == nil {
-				candidates = append(candidates, shortTextCandidate{
-					provider: c.provider,
-					model:    c.model,
-					lm:       m,
-				})
-			}
-		}
+		candidates = p.preferredShortTextCandidates(chat, keys)
 		candidates = append(candidates, shortTextCandidate{
 			provider: fallbackProvider,
 			model:    fallbackModelName,
+			route:    fallbackRoute,
 			lm:       fallbackModel,
 		})
 	}
@@ -213,17 +231,13 @@ func (p *Server) maybeGenerateChatTitle(
 		candidateCtx := titleCtx
 		candidateModel := candidate.lm
 		finishDebugRun := func(error) {}
-		candidateKeys := keys
-		if overrideSet {
-			candidateKeys = overrideKeys
-		}
 		if debugEnabled {
-			candidateCtx, candidateModel, finishDebugRun = prepareQuickgenDebugCandidate(
+			candidateCtx, candidateModel, finishDebugRun = p.prepareQuickgenDebugCandidate(
 				titleCtx,
 				chat,
-				candidateKeys,
 				debugSvc,
 				candidate,
+				modelOpts,
 				chatdebug.KindTitleGeneration,
 				triggerMessageID,
 				historyTipMessageID,
@@ -293,32 +307,26 @@ func (p *Server) maybeGenerateChatTitle(
 	}
 }
 
-func newQuickgenDebugModel(
+func (p *Server) newQuickgenDebugModel(
+	ctx context.Context,
 	chat database.Chat,
-	keys chatprovider.ProviderAPIKeys,
 	debugSvc *chatdebug.Service,
 	provider string,
 	model string,
+	route resolvedModelRoute,
+	modelOpts modelBuildOptions,
 ) (fantasy.LanguageModel, error) {
-	httpClient := &http.Client{Transport: &chatdebug.RecordingTransport{}}
-	debugModel, err := chatprovider.ModelFromConfig(
-		provider,
-		model,
-		keys,
-		chatprovider.UserAgent(),
-		chatprovider.CoderHeaders(chat),
-		httpClient,
-	)
+	debugOpts := modelOpts
+	debugOpts.RecordHTTP = true
+	debugModel, err := p.newModel(ctx, modelClientRequest{
+		Chat:         chat,
+		ModelName:    model,
+		UserAgent:    chatprovider.UserAgent(),
+		ExtraHeaders: chatprovider.CoderHeaders(chat),
+	}, route, debugOpts)
 	if err != nil {
 		return nil, err
 	}
-	if debugModel == nil {
-		return nil, xerrors.Errorf(
-			"create model for %s/%s returned nil",
-			provider,
-			model,
-		)
-	}
 
 	return chatdebug.WrapModel(debugModel, debugSvc, chatdebug.RecorderOptions{
 		ChatID:   chat.ID,
@@ -328,12 +336,12 @@ func newQuickgenDebugModel(
 	}), nil
 }
 
-func prepareQuickgenDebugCandidate(
+func (p *Server) prepareQuickgenDebugCandidate(
 	ctx context.Context,
 	chat database.Chat,
-	keys chatprovider.ProviderAPIKeys,
 	debugSvc *chatdebug.Service,
 	candidate shortTextCandidate,
+	modelOpts modelBuildOptions,
 	kind chatdebug.RunKind,
 	triggerMessageID int64,
 	historyTipMessageID int64,
@@ -345,12 +353,14 @@ func prepareQuickgenDebugCandidate(
 		return ctx, candidate.lm, finishDebugRun
 	}
 
-	debugModel, err := newQuickgenDebugModel(
+	debugModel, err := p.newQuickgenDebugModel(
+		ctx,
 		chat,
-		keys,
 		debugSvc,
 		candidate.provider,
 		candidate.model,
+		candidate.route,
+		modelOpts,
 	)
 	if err != nil {
 		logger.Warn(ctx, "failed to build short-text debug model",
@@ -393,18 +403,8 @@ func prepareQuickgenDebugCandidate(
 		return ctx, candidate.lm, finishDebugRun
 	}
 
-	runCtx := chatdebug.ContextWithRun(
-		ctx,
-		&chatdebug.RunContext{
-			RunID:               run.ID,
-			ChatID:              chat.ID,
-			TriggerMessageID:    triggerMessageID,
-			HistoryTipMessageID: historyTipMessageID,
-			Kind:                kind,
-			Provider:            candidate.provider,
-			Model:               candidate.model,
-		},
-	)
+	runContext := chatdebugRunContext(run)
+	runCtx := chatdebug.ContextWithRun(ctx, &runContext)
 	finishDebugRun = func(runErr error) {
 		if finalizeErr := debugSvc.FinalizeRun(ctx, chatdebug.FinalizeRunParams{
 			RunID:       run.ID,
@@ -824,7 +824,7 @@ const turnStatusLabelPrompt = "You write compact chat status labels for a sideba
 // message text. It follows the same candidate-selection strategy
 // as title generation: try preferred lightweight models first, then
 // fall back to the provided model. Returns "" on any failure.
-func generateTurnStatusLabel(
+func (p *Server) generateTurnStatusLabel(
 	ctx context.Context,
 	chat database.Chat,
 	status database.ChatStatus,
@@ -832,7 +832,9 @@ func generateTurnStatusLabel(
 	fallbackProvider string,
 	fallbackModelName string,
 	fallbackModel fantasy.LanguageModel,
+	fallbackRoute resolvedModelRoute,
 	keys chatprovider.ProviderAPIKeys,
+	modelOpts modelBuildOptions,
 	logger slog.Logger,
 	debugSvc *chatdebug.Service,
 	triggerMessageID int64,
@@ -848,24 +850,11 @@ func generateTurnStatusLabel(
 		"\nChat title: " + chat.Title +
 		"\n\nAgent's latest message:\n" + assistantText
 
-	candidates := make([]shortTextCandidate, 0, len(preferredTitleModels)+1)
-	for _, c := range preferredTitleModels {
-		m, err := chatprovider.ModelFromConfig(
-			c.provider, c.model, keys, chatprovider.UserAgent(),
-			chatprovider.CoderHeaders(chat),
-			nil,
-		)
-		if err == nil {
-			candidates = append(candidates, shortTextCandidate{
-				provider: c.provider,
-				model:    c.model,
-				lm:       m,
-			})
-		}
-	}
+	candidates := p.preferredShortTextCandidates(chat, keys)
 	candidates = append(candidates, shortTextCandidate{
 		provider: fallbackProvider,
 		model:    fallbackModelName,
+		route:    fallbackRoute,
 		lm:       fallbackModel,
 	})
 
@@ -876,12 +865,12 @@ func generateTurnStatusLabel(
 		candidateModel := candidate.lm
 		finishDebugRun := func(error) {}
 		if debugEnabled {
-			candidateCtx, candidateModel, finishDebugRun = prepareQuickgenDebugCandidate(
+			candidateCtx, candidateModel, finishDebugRun = p.prepareQuickgenDebugCandidate(
 				labelCtx,
 				chat,
-				keys,
 				debugSvc,
 				candidate,
+				modelOpts,
 				chatdebug.KindQuickgen,
 				triggerMessageID,
 				historyTipMessageID,