core/server/handlers/completion.go

package handlers

import (
	"database/sql"
	"encoding/json"
	"fmt"
	"io"
	"log"
	"net/http"

	"github.com/gin-gonic/gin"

	"git.gobha.me/xcaliber/chat-switchboard/database"
	"git.gobha.me/xcaliber/chat-switchboard/providers"
)

// ── Request Types ───────────────────────────

type completionRequest struct {
	ChannelID   string   `json:"channel_id"` // preferred; validated manually below
	ChatID      string   `json:"chat_id"`    // deprecated alias — maps to channel_id
	Content     string   `json:"content" binding:"required"`
	Model       string   `json:"model,omitempty"`
	PresetID    string   `json:"preset_id,omitempty"` // if set, unwraps preset → base model + config
	APIConfigID string   `json:"api_config_id,omitempty"`
	MaxTokens   int      `json:"max_tokens,omitempty"`
	Temperature *float64 `json:"temperature,omitempty"`
	TopP        *float64 `json:"top_p,omitempty"`
	Stream      *bool    `json:"stream,omitempty"`
}

// CompletionHandler proxies LLM requests through the backend.
type CompletionHandler struct{}

// NewCompletionHandler creates a new handler.
func NewCompletionHandler() *CompletionHandler {
	return &CompletionHandler{}
}

// ── Chat Completion ─────────────────────────
// POST /api/v1/chat/completions
//
// Flow:
//   1. Validate request, verify chat ownership
//   2. Resolve api_config (from request, chat, or user default)
//   3. Load conversation history from messages
//   4. Persist user message
//   5. Call provider (stream or non-stream)
//   6. Stream SSE to client / return JSON
//   7. Persist assistant message with token counts

func (h *CompletionHandler) Complete(c *gin.Context) {
	var req completionRequest
	if err := c.ShouldBindJSON(&req); err != nil {
		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
		return
	}

	// Support chat_id as alias during frontend transition
	channelID := req.ChannelID
	if channelID == "" {
		channelID = req.ChatID
	}
	if channelID == "" {
		c.JSON(http.StatusBadRequest, gin.H{"error": "channel_id is required"})
		return
	}

	userID := getUserID(c)

	// Verify channel ownership
	if !userOwnsChannel(c, channelID, userID) {
		return
	}

	// ── Preset unwrap: preset overrides defaults, explicit request fields win ──
	var presetSystemPrompt string
	if req.PresetID != "" {
		preset := ResolvePreset(req.PresetID, userID)
		if preset == nil {
			c.JSON(http.StatusBadRequest, gin.H{"error": "preset not found or not accessible"})
			return
		}
		// Preset provides defaults; explicit request fields take priority
		if req.Model == "" {
			req.Model = preset.BaseModelID
		}
		if req.APIConfigID == "" && preset.APIConfigID != nil {
			req.APIConfigID = *preset.APIConfigID
		}
		if req.Temperature == nil && preset.Temperature != nil {
			req.Temperature = preset.Temperature
		}
		if req.MaxTokens == 0 && preset.MaxTokens != nil {
			req.MaxTokens = *preset.MaxTokens
		}
		if preset.SystemPrompt != "" {
			presetSystemPrompt = preset.SystemPrompt
		}
	}

	// Resolve provider config
	providerCfg, providerID, model, configID, err := h.resolveConfig(userID, channelID, req)
	if err != nil {
		c.JSON(http.StatusBadRequest, gin.H{"error": err.Error()})
		return
	}

	provider, err := providers.Get(providerID)
	if err != nil {
		c.JSON(http.StatusInternalServerError, gin.H{"error": err.Error()})
		return
	}

	// Load conversation history
	messages, err := h.loadConversation(channelID, presetSystemPrompt)
	if err != nil {
		c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to load conversation"})
		return
	}

	// Add the new user message
	messages = append(messages, providers.Message{
		Role:    "user",
		Content: req.Content,
	})

	// Persist user message
	if err := h.persistMessage(channelID, "user", req.Content, "", 0, 0); err != nil {
		log.Printf("Failed to persist user message: %v", err)
	}

	// Build provider request
	provReq := providers.CompletionRequest{
		Model:    model,
		Messages: messages,
	}

	// Resolve capabilities for this model — auto-set defaults
	caps := h.getModelCapabilities(model, configID)

	if req.MaxTokens > 0 {
		provReq.MaxTokens = req.MaxTokens
	} else {
		// ResolveMaxOutput checks: caps → known models → context/8 → 4096
		provReq.MaxTokens = providers.ResolveMaxOutput(model, caps)
	}

	if req.Temperature != nil {
		provReq.Temperature = req.Temperature
	}
	if req.TopP != nil {
		provReq.TopP = req.TopP
	}

	// Determine streaming
	stream := true // default
	if req.Stream != nil {
		stream = *req.Stream
	}

	if stream {
		h.streamCompletion(c, provider, providerCfg, provReq, channelID, model)
	} else {
		h.syncCompletion(c, provider, providerCfg, provReq, channelID, model)
	}
}

// ── Streaming Completion (SSE) ──────────────

func (h *CompletionHandler) streamCompletion(
	c *gin.Context,
	provider providers.Provider,
	cfg providers.ProviderConfig,
	req providers.CompletionRequest,
	channelID, model string,
) {
	ch, err := provider.StreamCompletion(c.Request.Context(), cfg, req)
	if err != nil {
		c.JSON(http.StatusBadGateway, gin.H{"error": "provider error: " + err.Error()})
		return
	}

	// Set SSE headers
	c.Header("Content-Type", "text/event-stream")
	c.Header("Cache-Control", "no-cache")
	c.Header("Connection", "keep-alive")
	c.Header("X-Accel-Buffering", "no") // Disable nginx buffering

	c.Status(http.StatusOK)

	var fullContent string
	flusher, _ := c.Writer.(http.Flusher)

	for event := range ch {
		if event.Error != nil {
			fmt.Fprintf(c.Writer, "data: {\"error\":\"%s\"}\n\n", event.Error.Error())
			if flusher != nil {
				flusher.Flush()
			}
			break
		}

		if event.Delta != "" {
			fullContent += event.Delta
			// OpenAI-compatible SSE format
			fmt.Fprintf(c.Writer, "data: {\"choices\":[{\"delta\":{\"content\":%q},\"finish_reason\":null}],\"model\":%q}\n\n",
				event.Delta, model)
			if flusher != nil {
				flusher.Flush()
			}
		}

		if event.Done {
			finishReason := event.FinishReason
			if finishReason == "" {
				finishReason = "stop"
			}
			fmt.Fprintf(c.Writer, "data: {\"choices\":[{\"delta\":{},\"finish_reason\":%q}],\"model\":%q}\n\n",
				finishReason, model)
			io.WriteString(c.Writer, "data: [DONE]\n\n")
			if flusher != nil {
				flusher.Flush()
			}
			break
		}
	}

	// Persist assistant response
	if fullContent != "" {
		if err := h.persistMessage(channelID, "assistant", fullContent, model, 0, 0); err != nil {
			log.Printf("Failed to persist assistant message: %v", err)
		}
	}
}

// ── Non-Streaming Completion ────────────────

func (h *CompletionHandler) syncCompletion(
	c *gin.Context,
	provider providers.Provider,
	cfg providers.ProviderConfig,
	req providers.CompletionRequest,
	channelID, model string,
) {
	resp, err := provider.ChatCompletion(c.Request.Context(), cfg, req)
	if err != nil {
		c.JSON(http.StatusBadGateway, gin.H{"error": "provider error: " + err.Error()})
		return
	}

	// Persist assistant response
	if err := h.persistMessage(channelID, "assistant", resp.Content, model, resp.InputTokens, resp.OutputTokens); err != nil {
		log.Printf("Failed to persist assistant message: %v", err)
	}

	// Return OpenAI-compatible response
	c.JSON(http.StatusOK, gin.H{
		"model": resp.Model,
		"choices": []gin.H{
			{
				"message": gin.H{
					"role":    "assistant",
					"content": resp.Content,
				},
				"finish_reason": resp.FinishReason,
			},
		},
		"usage": gin.H{
			"prompt_tokens":     resp.InputTokens,
			"completion_tokens": resp.OutputTokens,
			"total_tokens":      resp.InputTokens + resp.OutputTokens,
		},
	})
}

// ── Model Capabilities ──────────────────────

// getModelCapabilities looks up capabilities from model_configs DB,
// then overlays with known model defaults and heuristic detection.
func (h *CompletionHandler) getModelCapabilities(model, apiConfigID string) providers.ModelCapabilities {
	// Start with known table or heuristic
	// Start with known model table or heuristics
	caps, found := providers.LookupKnownModel(model)
	if !found {
		caps = providers.InferCapabilities(model)
	}

	// Overlay with DB-stored capabilities (from provider sync or admin edit — authoritative)
	var capsJSON []byte
	err := database.DB.QueryRow(`
		SELECT capabilities FROM model_configs
		WHERE model_id = $1 AND api_config_id = $2
	`, model, apiConfigID).Scan(&capsJSON)

	if err == nil && capsJSON != nil {
		var dbCaps providers.ModelCapabilities
		if err := json.Unmarshal(capsJSON, &dbCaps); err == nil {
			if dbCaps.HasProviderData() {
				// DB has real provider data — use as authoritative, fill gaps
				caps = providers.MergeCapabilities(dbCaps, model)
			}
		}
	}

	return caps
}

// ── Config Resolution ───────────────────────
// Priority: request.api_config_id → chat.api_config_id → user's first active config

func (h *CompletionHandler) resolveConfig(userID string, channelID string, req completionRequest) (providers.ProviderConfig, string, string, string, error) {
	var configID string

	// 1. Explicit config from request
	if req.APIConfigID != "" {
		configID = req.APIConfigID
	}

	// 2. Config from channel
	if configID == "" {
		var channelConfigID *string
		err := database.DB.QueryRow(
			`SELECT api_config_id FROM channels WHERE id = $1`, channelID,
		).Scan(&channelConfigID)
		if err == nil && channelConfigID != nil {
			configID = *channelConfigID
		}
	}

	// 3. User's first active config (personal first, then global)
	if configID == "" {
		err := database.DB.QueryRow(`
			SELECT id FROM api_configs
			WHERE (user_id = $1 OR is_global = true OR user_id IS NULL) AND is_active = true
			ORDER BY user_id NULLS LAST, created_at ASC
			LIMIT 1
		`, userID).Scan(&configID)
		if err != nil {
			return providers.ProviderConfig{}, "", "", "", fmt.Errorf("no API config found — add one at /api-configs")
		}
	}

	// Load the config including custom headers and provider settings
	var providerID, endpoint string
	var apiKey, modelDefault *string
	var customHeadersJSON, providerSettingsJSON []byte
	err := database.DB.QueryRow(`
		SELECT provider, endpoint, api_key_encrypted, model_default, custom_headers, provider_settings
		FROM api_configs
		WHERE id = $1 AND (user_id = $2 OR is_global = true OR user_id IS NULL) AND is_active = true
	`, configID, userID).Scan(&providerID, &endpoint, &apiKey, &modelDefault, &customHeadersJSON, &providerSettingsJSON)

	if err == sql.ErrNoRows {
		return providers.ProviderConfig{}, "", "", "", fmt.Errorf("API config not found or not accessible")
	}
	if err != nil {
		return providers.ProviderConfig{}, "", "", "", fmt.Errorf("failed to load API config: %w", err)
	}

	// Resolve model: request > config default
	model := req.Model
	if model == "" && modelDefault != nil {
		model = *modelDefault
	}
	if model == "" {
		return providers.ProviderConfig{}, "", "", "", fmt.Errorf("no model specified and no default model in config")
	}

	key := ""
	if apiKey != nil {
		key = *apiKey
	}

	// Parse custom headers
	customHeaders := make(map[string]string)
	if customHeadersJSON != nil {
		_ = json.Unmarshal(customHeadersJSON, &customHeaders)
	}

	// Parse provider-specific settings
	providerSettings := make(map[string]interface{})
	if providerSettingsJSON != nil {
		_ = json.Unmarshal(providerSettingsJSON, &providerSettings)
	}

	return providers.ProviderConfig{
		Endpoint:      endpoint,
		APIKey:        key,
		CustomHeaders: customHeaders,
		Settings:      providerSettings,
	}, providerID, model, configID, nil
}

// ── Conversation Loader ─────────────────────

func (h *CompletionHandler) loadConversation(channelID string, presetSystemPrompt string) ([]providers.Message, error) {
	// Load system prompt from channel
	var systemPrompt *string
	_ = database.DB.QueryRow(
		`SELECT system_prompt FROM channels WHERE id = $1`, channelID,
	).Scan(&systemPrompt)

	messages := make([]providers.Message, 0)

	// Preset system prompt takes priority; channel system prompt is fallback
	if presetSystemPrompt != "" {
		messages = append(messages, providers.Message{
			Role:    "system",
			Content: presetSystemPrompt,
		})
	} else if systemPrompt != nil && *systemPrompt != "" {
		messages = append(messages, providers.Message{
			Role:    "system",
			Content: *systemPrompt,
		})
	}

	// Load message history (oldest first)
	rows, err := database.DB.Query(`
		SELECT role, content FROM messages
		WHERE channel_id = $1
		ORDER BY created_at ASC
	`, channelID)
	if err != nil {
		return nil, err
	}
	defer rows.Close()

	for rows.Next() {
		var msg providers.Message
		if err := rows.Scan(&msg.Role, &msg.Content); err != nil {
			return nil, err
		}
		messages = append(messages, msg)
	}

	return messages, nil
}

// ── Message Persistence ─────────────────────

func (h *CompletionHandler) persistMessage(channelID, role, content, model string, inputTokens, outputTokens int) error {
	var tokensUsed *int
	if inputTokens > 0 || outputTokens > 0 {
		total := inputTokens + outputTokens
		tokensUsed = &total
	}

	// Find current leaf for parent_id
	var parentID *string
	_ = database.DB.QueryRow(
		`SELECT id FROM messages WHERE channel_id = $1 ORDER BY created_at DESC LIMIT 1`,
		channelID,
	).Scan(&parentID)

	participantType := "user"
	participantID := channelID // will be overridden below
	if role == "assistant" {
		participantType = "model"
		participantID = model
	} else {
		// Get channel owner as participant_id for user messages
		var ownerID string
		if err := database.DB.QueryRow(`SELECT user_id FROM channels WHERE id = $1`, channelID).Scan(&ownerID); err == nil {
			participantID = ownerID
		}
	}

	_, err := database.DB.Exec(`
		INSERT INTO messages (channel_id, role, content, model, tokens_used, parent_id, participant_type, participant_id)
		VALUES ($1, $2, $3, NULLIF($4, ''), $5, $6, $7, $8)
	`, channelID, role, content, model, tokensUsed, parentID, participantType, participantID)
	if err != nil {
		return err
	}

	// Touch channel updated_at
	_, _ = database.DB.Exec(`UPDATE channels SET updated_at = NOW() WHERE id = $1`, channelID)
	return nil
}