tools/search.go

package tools

import (
	"context"
	"fmt"
	"strings"

	"github.com/korap/korap-mcp/service"
	"github.com/korap/korap-mcp/validation"
	"github.com/mark3labs/mcp-go/mcp"
	"github.com/rs/zerolog/log"
)

// SearchTool implements the Tool interface for KorAP corpus search
type SearchTool struct {
	client    *service.Client
	validator *validation.Validator
}

// NewSearchTool creates a new search tool instance
func NewSearchTool(client *service.Client) *SearchTool {
	return &SearchTool{
		client:    client,
		validator: validation.New(log.Logger),
	}
}

// Name returns the tool name
func (s *SearchTool) Name() string {
	return "korap_search"
}

// Description returns the tool description
func (s *SearchTool) Description() string {
	return "Search for words or phrases in KorAP corpora using various query languages"
}

// InputSchema returns the JSON schema for tool parameters
func (s *SearchTool) InputSchema() map[string]any {
	return map[string]any{
		"type": "object",
		"properties": map[string]any{
			"query": map[string]any{
				"type":        "string",
				"description": "The search query. Supports different query languages like Poliqarp, CosmasII, or Annis depending on the selected query_language parameter.",
				"minLength":   1,
				"maxLength":   1000,
				"examples":    []string{"Haus", "[pos=NN]", "der /w1:5 Mann"},
			},
			"query_language": map[string]any{
				"type":        "string",
				"description": "Query language to use for parsing the search query. Supported languages: 'poliqarp' (default; extended Poliqarp QL), 'cosmas2' (corpus query syntax of COSMAS II), 'annis' (multi-layer annotation queries), 'cql' (corpus query language), 'cqp' (Corpus Query Processor syntax), 'fcsql' (Federated Content Search queries).",
				"enum":        []string{"poliqarp", "cosmas2", "annis", "cql", "cqp", "fcsql"},
				"default":     "poliqarp",
				"examples":    []string{"poliqarp", "cosmas2", "annis", "cql", "cqp", "fcsql"},
			},
			"corpus": map[string]any{
				"type":        "string",
				"description": "Virtual corpus query to filter search results based on metadata fields. Supports boolean operations (& | !), comparison operators (= != < > in), and regular expressions (/pattern/). Use metadata fields like corpusSigle, textClass, pubDate, textType, availability, etc. When not provided, searches all available data accessible to the user.",
				"pattern":     "^[a-zA-Z0-9._\\-\\s&|!=<>()/*\"']+$",
				"examples":    []string{"corpusSigle = \"GOE\"", "textClass = \"politics\" & pubDate in 2020", "textType = \"news\" | textType = \"blog\"", "availability = /CC.*/ & textClass != \"fiction\""},
			},
			"count": map[string]any{
				"type":        "integer",
				"description": "Maximum number of search results to return. Higher values may increase response time. Use smaller values for faster responses when doing exploratory searches.",
				"minimum":     0,
				"maximum":     10000,
				"default":     25,
				"examples":    []any{10, 25, 50, 100},
			},
		},
		"required":             []string{"query"},
		"additionalProperties": false,
		"title":                "KorAP Search Parameters",
		"description":          "Parameters for searching text corpora using KorAP's powerful query languages and filtering capabilities.",
	}
}

// Execute performs the search operation
func (s *SearchTool) Execute(ctx context.Context, request mcp.CallToolRequest) (*mcp.CallToolResult, error) {
	log.Debug().
		Str("tool", s.Name()).
		Msg("Executing search tool")

	// Extract required query parameter
	query, err := request.RequireString("query")
	if err != nil {
		return nil, fmt.Errorf("query parameter is required: %w", err)
	}

	// Extract optional parameters with defaults
	queryLang := request.GetString("query_language", "poliqarp")
	corpus := request.GetString("corpus", "")
	count := request.GetInt("count", 25)

	// Validate the search request using the validation package
	searchReq := validation.SearchRequest{
		Query:         query,
		QueryLanguage: queryLang,
		Corpus:        corpus,
		Count:         count,
	}

	if err := s.validator.ValidateSearchRequest(searchReq); err != nil {
		log.Warn().
			Err(err).
			Interface("request", searchReq).
			Msg("Search request validation failed")
		return nil, fmt.Errorf("invalid search request: %w", err)
	}

	// Sanitize inputs
	query = s.validator.SanitizeQuery(query)
	if corpus != "" {
		corpus = s.validator.SanitizeCorpusID(corpus)
	}

	log.Debug().
		Str("query", query).
		Str("query_language", queryLang).
		Str("corpus", corpus).
		Int("count", count).
		Msg("Parsed and validated search parameters")

	// Check if client is available and authenticated
	if s.client == nil {
		return nil, fmt.Errorf("KorAP client not configured")
	}

	if !s.client.IsAuthenticated() {
		log.Warn().Msg("Client not authenticated, attempting authentication")
		if err := s.client.AuthenticateWithClientCredentials(ctx); err != nil {
			return nil, fmt.Errorf("authentication failed: %w", err)
		}
	}

	// Prepare search request
	korapSearchReq := service.SearchRequest{
		Query:      query,
		QueryLang:  queryLang,
		Collection: corpus,
		Count:      count,
	}

	// Perform the search
	var searchResp service.SearchResponse
	err = s.client.PostJSON(ctx, "search", korapSearchReq, &searchResp)
	if err != nil {
		log.Error().
			Err(err).
			Str("query", query).
			Msg("Search request failed")
		return nil, fmt.Errorf("search failed: %w", err)
	}

	// Validate the response
	if err := s.validator.ValidateSearchResponse(&searchResp); err != nil {
		log.Warn().
			Err(err).
			Msg("Search response validation failed, but continuing with potentially invalid data")
		// Continue processing despite validation errors to be resilient
	}

	log.Info().
		Str("query", query).
		Int("total_results", searchResp.Meta.TotalResults).
		Int("returned_matches", len(searchResp.Matches)).
		Float64("search_time", searchResp.Meta.SearchTime).
		Msg("Search completed successfully")

	// Format the response
	result := s.formatSearchResults(&searchResp)

	return mcp.NewToolResultText(result), nil
}

// formatSearchResults formats the search response into a readable text format
func (s *SearchTool) formatSearchResults(response *service.SearchResponse) string {
	var result strings.Builder

	result.WriteString("KorAP Search Results\n")
	result.WriteString("====================\n\n")

	// Query information
	result.WriteString(fmt.Sprintf("Query: %s\n", response.Query.Query))
	if response.Query.QueryLang != "" {
		result.WriteString(fmt.Sprintf("Query Language: %s\n", response.Query.QueryLang))
	}
	if response.Query.Collection != "" {
		result.WriteString(fmt.Sprintf("Corpus: %s\n", response.Query.Collection))
	}
	result.WriteString("\n")

	// Result statistics
	result.WriteString("Results Summary:\n")
	result.WriteString(fmt.Sprintf("  Total Results: %d\n", response.Meta.TotalResults))
	result.WriteString(fmt.Sprintf("  Shown: %d-%d\n",
		response.Meta.StartIndex+1,
		response.Meta.StartIndex+len(response.Matches)))
	if response.Meta.SearchTime > 0 {
		result.WriteString(fmt.Sprintf("  Search Time: %.3f seconds\n", response.Meta.SearchTime))
	}
	result.WriteString("\n")

	// Individual matches
	if len(response.Matches) > 0 {
		result.WriteString("Matches:\n")
		result.WriteString("--------\n")

		for i, match := range response.Matches {
			result.WriteString(fmt.Sprintf("\n%d. Text: %s\n", i+1, match.TextSigle))
			if match.Snippet != "" {
				result.WriteString(fmt.Sprintf("   Snippet: %s\n", match.Snippet))
			}
			if match.PubPlace != "" {
				result.WriteString(fmt.Sprintf("   Publication: %s\n", match.PubPlace))
			}
			if match.MatchID != "" {
				result.WriteString(fmt.Sprintf("   Match ID: %s\n", match.MatchID))
			}
			result.WriteString(fmt.Sprintf("   Position: %d\n", match.Position))
		}
	} else {
		result.WriteString("No matches found.\n")
	}

	// Additional information
	if response.Query.CutOff {
		result.WriteString("\nNote: Results were cut off due to limits.\n")
	}
	if response.Query.TimeExceeded {
		result.WriteString("\nNote: Search time limit was exceeded.\n")
	}

	return result.String()
}