jizhi/backend/MiaoJiZhang.Api/Controllers/ParseController.cs

using System.Security.Claims;
using MiaoJiZhang.Api.Contracts;
using MiaoJiZhang.Api.Services;
using MiaoJiZhang.Domain.Entities;
using MiaoJiZhang.Domain.Enums;
using MiaoJiZhang.Infrastructure.Persistence;
using Microsoft.AspNetCore.Authorization;
using Microsoft.AspNetCore.Mvc;
using Microsoft.AspNetCore.RateLimiting;
using Microsoft.EntityFrameworkCore;

namespace MiaoJiZhang.Api.Controllers;

/// <summary>
/// OCR/语音/图片辅助解析。客户端确认后再调用交易接口正式入账。
/// 图片模式由后端多模态模型提取一笔或多笔账单草稿。
/// </summary>
[ApiController]
[Authorize]
[Route("api/parse")]
[RequireAiPermission]
public class ParseController(AppDbContext db, ILlmClient llm, AgentService agent) : ControllerBase
{
    private long Uid => long.Parse(
        User.FindFirstValue(ClaimTypes.NameIdentifier) ??
        User.FindFirstValue("sub")!);

    [HttpPost]
    [EnableRateLimiting("ai")]
    public async Task<ActionResult<OcrParseResponse>> Parse(OcrParseRequest req)
    {
        return await ParseWithAgent(req, HttpContext.RequestAborted);
    }

    private async Task<ActionResult<OcrParseResponse>> ParseWithAgent(
        OcrParseRequest req,
        CancellationToken ct)
    {
        if (string.IsNullOrWhiteSpace(req.Text))
            return BadRequest(new ApiError("TEXT_EMPTY", "内容不能为空"));
        var featureKey = req.Source == "voice"
            ? "feature.voice_enabled"
            : "feature.ocr_enabled";
        if (!await FeatureEnabled(featureKey))
            return StatusCode(403, new ApiError("FEATURE_DISABLED", "该解析功能已由后台关闭"));

        IReadOnlyList<ParsedBill> drafts;
        try
        {
            drafts = await agent.ParseDraftsAsync(Uid, req.Text.Trim(), ct);
        }
        catch (InvalidOperationException ex)
        {
            return StatusCode(
                503,
                new ApiError("AGENT_PARSE_UNAVAILABLE", ex.Message));
        }
        catch (Exception)
        {
            return StatusCode(
                502,
                new ApiError(
                    "AGENT_PARSE_FAILED",
                    "AI 暂时无法解析，没有生成账单草稿"));
        }

        if (drafts.Count == 0)
        {
            var fallback = await db.Categories.FirstAsync(
                c => c.Name == "其他" &&
                     c.Type == TransactionType.Expense &&
                     c.UserId == null,
                ct);
            return Ok(new OcrParseResponse(
                false,
                fallback.Id,
                fallback.Name,
                fallback.IconKey,
                0,
                null,
                req.Text,
                "expense"));
        }

        var bill = drafts[0];
        var category = await db.Categories.FirstAsync(
            c => c.Id == bill.CategoryId && c.Type == bill.Type,
            ct);
        return Ok(new OcrParseResponse(
            true,
            category.Id,
            category.Name,
            category.IconKey,
            bill.Amount,
            bill.PaymentMethod,
            bill.Note,
            bill.Type == TransactionType.Income ? "income" : "expense"));
    }

    /// <summary>上传截屏或小票，提取其中全部独立交易。</summary>
    [HttpPost("image")]
    [EnableRateLimiting("upload")]
    [RequestSizeLimit(10 * 1024 * 1024)]
    [RequestFormLimits(MultipartBodyLengthLimit = 10 * 1024 * 1024)]
    public async Task<ActionResult<ImageParseResponse>> ParseImage(
        IFormFile file,
        [FromQuery] string source = "image",
        CancellationToken ct = default)
    {
        var featureKey = source == "screenshot"
            ? "feature.screenshot_bookkeeping_enabled"
            : "feature.ocr_enabled";
        if (!await FeatureEnabled(featureKey))
            return StatusCode(403, new ApiError("FEATURE_DISABLED", "该图片识别功能已由后台关闭"));
        if (file is null || file.Length == 0)
            return BadRequest(new ApiError("FILE_EMPTY", "请上传图片"));
        if (!file.ContentType.StartsWith("image/"))
            return BadRequest(new ApiError("NOT_IMAGE", "仅支持图片文件"));
        if (file.Length > 10 * 1024 * 1024)
            return BadRequest(new ApiError("FILE_TOO_LARGE", "图片不能超过 10MB"));

        if (!llm.IsEnabled)
        {
            return StatusCode(
                503,
                new ApiError(
                    "LLM_NOT_CONFIGURED",
                    "AI 图片解析未配置，请在管理后台设置 LLM API Key"));
        }

        using var stream = new MemoryStream();
        await file.CopyToAsync(stream, ct);

        IReadOnlyList<ImageParseResult>? results;
        try
        {
            results = await llm.AnalyzeImageAsync(
                stream.ToArray(),
                file.ContentType,
                ct);
        }
        catch (InvalidOperationException ex)
        {
            return StatusCode(502, new ApiError("LLM_IMAGE_ERROR", ex.Message));
        }

        if (results is null)
            return StatusCode(500, new ApiError("LLM_ERROR", "AI 解析失败，请稍后重试"));

        var categories = await db.Categories
            .Where(c => !c.IsDeleted && (c.UserId == null || c.UserId == Uid))
            .ToListAsync(ct);

        if (results.Count == 0)
        {
            return Ok(new ImageParseResponse(
                false,
                0,
                "待确认",
                "tag",
                0,
                null,
                "",
                "unknown",
                [],
                null));
        }

        var items = new List<ImageParseItemResponse>();
        foreach (var result in results.Take(20))
        {
            var normalizedType = result.Type.Trim().ToLowerInvariant();
            if (normalizedType is not ("income" or "expense"))
            {
                items.Add(new ImageParseItemResponse(
                    false, 0, "待确认", "tag", "unknown",
                    result.Amount, result.PaymentMethod, result.Note,
                    result.OccurredAt));
                continue;
            }
            var type = normalizedType == "income"
                ? TransactionType.Income
                : TransactionType.Expense;
            var category = FindCategory(categories, type, result.CategoryName);
            items.Add(new ImageParseItemResponse(
                true,
                category.Id,
                category.Name,
                category.IconKey,
                normalizedType,
                result.Amount,
                result.PaymentMethod,
                result.Note,
                result.OccurredAt));
        }

        var first = items[0];
        return Ok(new ImageParseResponse(
            true,
            first.CategoryId,
            first.CategoryName,
            first.CategoryIcon,
            first.Amount,
            first.PaymentMethod,
            first.Note,
            first.Type,
            items,
            first.OccurredAt));
    }

    private async Task<bool> FeatureEnabled(string key)
    {
        var value = await db.AppConfigs
            .Where(config => config.Key == key)
            .Select(config => config.Value)
            .FirstOrDefaultAsync();
        return value is null || !value.Equals("false", StringComparison.OrdinalIgnoreCase);
    }

    private Category FindCategory(
        List<Category> categories,
        TransactionType type,
        string requestedName)
    {
        if (type == TransactionType.Income &&
            requestedName is "退款" or "返现" or "退回")
        {
            requestedName = "报销";
        }
        return categories
                   .Where(c => c.Type == type && c.Name == requestedName)
                   .OrderByDescending(c => c.UserId == Uid)
                   .FirstOrDefault()
               ?? categories.First(
                   c => c.Type == type &&
                        c.Name == "其他" &&
                        c.UserId == null);
    }
}