Changes

rag-api/Services/CachedAiClient.cs

@@ -1,52 +0,0 @@
-using Microsoft.Extensions.Options;
-using Api.Services.Contracts;
-using Api.Settings;
-
-namespace Api.Services;
-
-public sealed class CachedAiClient : IAiClient
-{
-    private readonly RawAiClient _raw;
-    private readonly IRagRepository _repository;
-    private readonly AiSettings _settings;
-
-    public CachedAiClient(RawAiClient raw, IRagRepository repository, IOptions<AiSettings> options)
-    {
-        _raw = raw;
-        _repository = repository;
-        _settings = options.Value;
-    }
-
-    public async Task<float[]> CreateEmbeddingAsync(string input, CancellationToken ct)
-    {
-        var model = GetEmbeddingModel();
-        var textHash = HashHelper.Compute(input);
-        var cacheKey = HashHelper.Compute($"embedding:{_settings.Provider}:{model}:{textHash}");
-        var cached = await _repository.GetEmbeddingAsync(cacheKey, ct);
-        if (cached is not null) return cached;
-
-        var vector = await _raw.CreateEmbeddingAsync(input, ct);
-        await _repository.SaveEmbeddingAsync(cacheKey, model, textHash, vector, ct);
-        return vector;
-    }
-
-    public async Task<string> CreateChatCompletionAsync(string systemPrompt, string userPrompt, decimal temperature, CancellationToken ct)
-    {
-        var model = GetChatModel();
-        var cacheKey = HashHelper.Compute($"chat:{_settings.Provider}:{model}:{temperature:0.00}:{systemPrompt}:{userPrompt}");
-        var cached = await _repository.GetChatCompletionAsync(cacheKey, ct);
-        if (cached is not null) return cached;
-
-        var response = await _raw.CreateChatCompletionAsync(systemPrompt, userPrompt, temperature, ct);
-        await _repository.SaveChatCompletionAsync(cacheKey, model, temperature, response, ct);
-        return response;
-    }
-
-    private string GetEmbeddingModel() => string.Equals(_settings.Provider, "Ollama", StringComparison.OrdinalIgnoreCase)
-        ? _settings.Ollama.EmbeddingModel
-        : _settings.OpenAI.EmbeddingModel;
-
-    private string GetChatModel() => string.Equals(_settings.Provider, "Ollama", StringComparison.OrdinalIgnoreCase)
-        ? _settings.Ollama.ChatModel
-        : _settings.OpenAI.ChatModel;
-}

rag-api/Services/Contracts/IAiClient.cs

@@ -1,7 +0,0 @@
-namespace Api.Services.Contracts;
-
-public interface IAiClient
-{
-    Task<float[]> CreateEmbeddingAsync(string input, CancellationToken ct);
-    Task<string> CreateChatCompletionAsync(string systemPrompt, string userPrompt, decimal temperature, CancellationToken ct);
-}

rag-api/Services/Contracts/IDocumentClassifier.cs

@@ -1,4 +1,4 @@
-using Api.Services.Contracts.Models;
+using Api.Models;
 
 namespace Api.Services.Contracts;
 

rag-api/Services/Contracts/IRagRepository.cs

@@ -1,16 +0,0 @@
-using Api.Services.Contracts.Models;
-
-namespace Api.Services.Contracts;
-
-public interface IRagRepository
-{
-    Task InitializeAsync(CancellationToken ct);
-    Task<RagDocumentRecord?> GetDocumentByTextHashAsync(string textHash, string? sourceUrl, CancellationToken ct);
-    Task<RagDocumentRecord?> GetDocumentByIdAsync(string id, CancellationToken ct);
-    Task SaveDocumentAsync(RagDocumentRecord document, IReadOnlyList<RagChunkRecord> chunks, CancellationToken ct);
-    Task<IReadOnlyList<SearchCandidateChunk>> SearchChunksAsync(float[] queryEmbedding, IReadOnlyList<string>? targetTypes, int topK, CancellationToken ct);
-    Task<float[]?> GetEmbeddingAsync(string cacheKey, CancellationToken ct);
-    Task SaveEmbeddingAsync(string cacheKey, string model, string textHash, float[] vector, CancellationToken ct);
-    Task<string?> GetChatCompletionAsync(string cacheKey, CancellationToken ct);
-    Task SaveChatCompletionAsync(string cacheKey, string model, decimal temperature, string responseText, CancellationToken ct);
-}

rag-api/Services/Contracts/IRagService.cs

@@ -1,6 +1,6 @@
-using Api.Requests;
-using Api.Responses;
-using Api.Services.Contracts.Models;
+using Api.Models;
+using Api.Models.Requests;
+using Api.Models.Responses;
 
 namespace Api.Services.Contracts;
 

rag-api/Services/Contracts/Models/DocumentClassification.cs

@@ -1,10 +0,0 @@
-namespace Api.Services.Contracts.Models
-{
-    public sealed class DocumentClassification
-    {
-        public required string DocumentType { get; init; }
-        public double Confidence { get; init; }
-        public required string Title { get; init; }
-        public Dictionary<string, string> Metadata { get; init; } = [];
-    }
-}

rag-api/Services/Contracts/Models/RagChunkRecord.cs

@@ -1,11 +0,0 @@
-namespace Api.Services.Contracts.Models
-{
-    public sealed class RagChunkRecord
-    {
-        public required string Id { get; init; }
-        public required string DocumentId { get; init; }
-        public int ChunkIndex { get; init; }
-        public required string Text { get; init; }
-        public required float[] Embedding { get; init; }
-    }
-}

rag-api/Services/Contracts/Models/RagDocumentDetails.cs

@@ -1,13 +0,0 @@
-namespace Api.Services.Contracts.Models
-{
-    public sealed class RagDocumentDetails
-    {
-        public required string Id { get; init; }
-        public required string DocumentType { get; init; }
-        public required string Title { get; init; }
-        public string? SourceUrl { get; init; }
-        public required string Text { get; init; }
-        public required string TextHash { get; init; }
-        public DateTimeOffset CreatedAt { get; init; }
-    }
-}

rag-api/Services/Contracts/Models/RagDocumentRecord.cs

@@ -1,15 +0,0 @@
-namespace Api.Services.Contracts.Models
-{
-    public sealed class RagDocumentRecord
-    {
-        public required string Id { get; init; }
-        public required string DocumentType { get; init; }
-        public required string Title { get; init; }
-        public string? SourceUrl { get; init; }
-        public required string Text { get; init; }
-        public required string TextHash { get; init; }
-        public double TypeConfidence { get; init; }
-        public string MetadataJson { get; init; } = "{}";
-        public DateTimeOffset CreatedAt { get; init; }
-    }
-}

rag-api/Services/Contracts/Models/SearchCandidateChunk.cs

@@ -1,9 +0,0 @@
-namespace Api.Services.Contracts.Models
-{
-    public sealed class SearchCandidateChunk
-    {
-        public required RagDocumentRecord Document { get; init; }
-        public required RagChunkRecord Chunk { get; init; }
-        public double Score { get; init; }
-    }
-}

rag-api/Services/DocumentClassifier.cs

@@ -1,6 +1,6 @@
 using System.Text.RegularExpressions;
 using Api.Services.Contracts;
-using Api.Services.Contracts.Models;
+using Api.Models;
 
 namespace Api.Services;
 

rag-api/Services/EfRagRepository.cs

@@ -1,195 +0,0 @@
-using Api.Data;
-using Api.Data.Entities;
-using Api.Services.Contracts;
-using Api.Services.Contracts.Models;
-using Microsoft.EntityFrameworkCore;
-
-namespace Api.Services;
-
-public sealed class EfRagRepository : IRagRepository
-{
-    private readonly RagDbContext _db;
-    private readonly ILogger<EfRagRepository> _logger;
-
-    public EfRagRepository(RagDbContext db, ILogger<EfRagRepository> logger)
-    {
-        _db = db;
-        _logger = logger;
-    }
-
-    public async Task InitializeAsync(CancellationToken ct)
-    {
-        _logger.LogInformation("Ensuring RAG database schema exists using EF Core");
-        await _db.Database.EnsureCreatedAsync(ct);
-    }
-
-    public async Task<RagDocumentRecord?> GetDocumentByTextHashAsync(string textHash, string? sourceUrl, CancellationToken ct)
-    {
-        var query = _db.RagDocuments
-            .AsNoTracking()
-            .Where(x => x.TextHash == textHash);
-
-        if (!string.IsNullOrWhiteSpace(sourceUrl))
-        {
-            query = query.Where(x => x.SourceUrl == sourceUrl);
-        }
-
-        var entity = await query
-            .OrderByDescending(x => x.CreatedAt)
-            .FirstOrDefaultAsync(ct);
-
-        return entity is null ? null : ToRecord(entity);
-    }
-
-    public async Task<RagDocumentRecord?> GetDocumentByIdAsync(string id, CancellationToken ct)
-    {
-        var entity = await _db.RagDocuments
-            .AsNoTracking()
-            .FirstOrDefaultAsync(x => x.Id == id, ct);
-
-        return entity is null ? null : ToRecord(entity);
-    }
-
-    public async Task SaveDocumentAsync(RagDocumentRecord document, IReadOnlyList<RagChunkRecord> chunks, CancellationToken ct)
-    {
-        var exists = await _db.RagDocuments.AnyAsync(x => x.Id == document.Id, ct);
-        if (exists)
-        {
-            _logger.LogInformation("RAG document already exists. DocumentId={DocumentId}", document.Id);
-            return;
-        }
-
-        var entity = new RagDocumentEntity
-        {
-            Id = document.Id,
-            DocumentType = document.DocumentType,
-            Title = document.Title,
-            SourceUrl = document.SourceUrl,
-            RawText = document.Text,
-            TextHash = document.TextHash,
-            TypeConfidence = document.TypeConfidence,
-            MetadataJson = document.MetadataJson,
-            CreatedAt = document.CreatedAt.UtcDateTime,
-            Chunks = chunks.Select(chunk => new RagChunkEntity
-            {
-                Id = chunk.Id,
-                DocumentId = chunk.DocumentId,
-                ChunkIndex = chunk.ChunkIndex,
-                Text = chunk.Text,
-                Embedding = VectorSerializer.ToBytes(chunk.Embedding)
-            }).ToList()
-        };
-
-        _db.RagDocuments.Add(entity);
-        await _db.SaveChangesAsync(ct);
-    }
-
-    public async Task<IReadOnlyList<SearchCandidateChunk>> SearchChunksAsync(
-        float[] queryEmbedding,
-        IReadOnlyList<string>? targetTypes,
-        int topK,
-        CancellationToken ct)
-    {
-        var types = targetTypes?
-            .Where(x => !string.IsNullOrWhiteSpace(x))
-            .Select(x => x.Trim().ToLowerInvariant())
-            .Distinct()
-            .ToArray() ?? [];
-
-        var query = _db.RagChunks
-            .AsNoTracking()
-            .Include(x => x.Document)
-            .AsQueryable();
-
-        if (types.Length > 0)
-        {
-            query = query.Where(x => x.Document != null && types.Contains(x.Document.DocumentType.ToLower()));
-        }
-
-        var rows = await query.ToListAsync(ct);
-
-        return rows
-            .Where(x => x.Document is not null)
-            .Select(x => new SearchCandidateChunk
-            {
-                Document = ToRecord(x.Document!),
-                Chunk = new RagChunkRecord
-                {
-                    Id = x.Id,
-                    DocumentId = x.DocumentId,
-                    ChunkIndex = x.ChunkIndex,
-                    Text = x.Text,
-                    Embedding = VectorSerializer.FromBytes(x.Embedding)
-                },
-                Score = VectorSerializer.CosineSimilarity(queryEmbedding, VectorSerializer.FromBytes(x.Embedding))
-            })
-            .OrderByDescending(x => x.Score)
-            .Take(Math.Max(topK * 4, topK))
-            .ToList();
-    }
-
-    public async Task<float[]?> GetEmbeddingAsync(string cacheKey, CancellationToken ct)
-    {
-        var entry = await _db.RagEmbeddingCache
-            .AsNoTracking()
-            .FirstOrDefaultAsync(x => x.CacheKey == cacheKey, ct);
-
-        return entry is null ? null : VectorSerializer.FromBytes(entry.Vector);
-    }
-
-    public async Task SaveEmbeddingAsync(string cacheKey, string model, string textHash, float[] vector, CancellationToken ct)
-    {
-        var exists = await _db.RagEmbeddingCache.AnyAsync(x => x.CacheKey == cacheKey, ct);
-        if (exists) return;
-
-        _db.RagEmbeddingCache.Add(new RagEmbeddingCacheEntity
-        {
-            CacheKey = cacheKey,
-            Model = model,
-            TextHash = textHash,
-            Vector = VectorSerializer.ToBytes(vector),
-            CreatedAt = DateTime.UtcNow
-        });
-
-        await _db.SaveChangesAsync(ct);
-    }
-
-    public async Task<string?> GetChatCompletionAsync(string cacheKey, CancellationToken ct)
-    {
-        return await _db.RagChatCompletionCache
-            .AsNoTracking()
-            .Where(x => x.CacheKey == cacheKey)
-            .Select(x => x.ResponseText)
-            .FirstOrDefaultAsync(ct);
-    }
-
-    public async Task SaveChatCompletionAsync(string cacheKey, string model, decimal temperature, string responseText, CancellationToken ct)
-    {
-        var exists = await _db.RagChatCompletionCache.AnyAsync(x => x.CacheKey == cacheKey, ct);
-        if (exists) return;
-
-        _db.RagChatCompletionCache.Add(new RagChatCompletionCacheEntity
-        {
-            CacheKey = cacheKey,
-            Model = model,
-            Temperature = temperature,
-            ResponseText = responseText,
-            CreatedAt = DateTime.UtcNow
-        });
-
-        await _db.SaveChangesAsync(ct);
-    }
-
-    private static RagDocumentRecord ToRecord(RagDocumentEntity entity) => new()
-    {
-        Id = entity.Id,
-        DocumentType = entity.DocumentType,
-        Title = entity.Title,
-        SourceUrl = entity.SourceUrl,
-        Text = entity.RawText,
-        TextHash = entity.TextHash,
-        TypeConfidence = entity.TypeConfidence,
-        MetadataJson = entity.MetadataJson,
-        CreatedAt = new DateTimeOffset(DateTime.SpecifyKind(entity.CreatedAt, DateTimeKind.Utc))
-    };
-}

rag-api/Services/HashHelper.cs

@@ -1,14 +0,0 @@
-using System.Security.Cryptography;
-using System.Text;
-
-namespace Api.Services;
-
-public static class HashHelper
-{
-    public static string Compute(string value)
-    {
-        using var sha = SHA256.Create();
-        var bytes = sha.ComputeHash(Encoding.UTF8.GetBytes(value ?? string.Empty));
-        return Convert.ToHexString(bytes);
-    }
-}

rag-api/Services/RagService.cs

@@ -1,10 +1,13 @@
 using System.Text.Json;
 using Microsoft.Extensions.Options;
 using Api.Services.Contracts;
-using Api.Settings;
-using Api.Responses;
-using Api.Requests;
-using Api.Services.Contracts.Models;
+using Api.Models.Requests;
+using Api.Models.Responses;
+using Api.Models.Settings;
+using Api.Data.Repositories.Contracts;
+using Api.Clients.Ai.Contracts;
+using Api.Clients.Ai;
+using Api.Models;
 
 namespace Api.Services;
 

rag-api/Services/RawAiClient.cs

@@ -1,116 +0,0 @@
-using System.Net.Http.Headers;
-using System.Text;
-using System.Text.Json;
-using System.Text.Json.Serialization;
-using Microsoft.Extensions.Options;
-using Api.Services.Contracts;
-using Api.Settings;
-
-namespace Api.Services;
-
-public sealed class RawAiClient : IAiClient
-{
-    private readonly HttpClient _http;
-    private readonly AiSettings _settings;
-    private static readonly JsonSerializerOptions JsonOptions = new(JsonSerializerDefaults.Web)
-    {
-        DefaultIgnoreCondition = JsonIgnoreCondition.WhenWritingNull
-    };
-
-    public RawAiClient(HttpClient http, IOptions<AiSettings> options)
-    {
-        _http = http;
-        _settings = options.Value;
-    }
-
-    public async Task<float[]> CreateEmbeddingAsync(string input, CancellationToken ct)
-    {
-        return IsOllama() ? await CreateOllamaEmbeddingAsync(input, ct) : await CreateOpenAiEmbeddingAsync(input, ct);
-    }
-
-    public async Task<string> CreateChatCompletionAsync(string systemPrompt, string userPrompt, decimal temperature, CancellationToken ct)
-    {
-        return IsOllama()
-            ? await CreateOllamaChatCompletionAsync(systemPrompt, userPrompt, temperature, ct)
-            : await CreateOpenAiChatCompletionAsync(systemPrompt, userPrompt, temperature, ct);
-    }
-
-    private bool IsOllama() => string.Equals(_settings.Provider, "Ollama", StringComparison.OrdinalIgnoreCase);
-
-    private async Task<float[]> CreateOpenAiEmbeddingAsync(string input, CancellationToken ct)
-    {
-        if (string.IsNullOrWhiteSpace(_settings.OpenAI.ApiKey)) throw new InvalidOperationException("OpenAI API key is missing.");
-        using var request = new HttpRequestMessage(HttpMethod.Post, "https://api.openai.com/v1/embeddings");
-        request.Headers.Authorization = new AuthenticationHeaderValue("Bearer", _settings.OpenAI.ApiKey);
-        request.Content = ToJson(new { model = _settings.OpenAI.EmbeddingModel, input });
-        using var cts = CancellationTokenSource.CreateLinkedTokenSource(ct);
-        cts.CancelAfter(TimeSpan.FromSeconds(Math.Max(15, _settings.OpenAI.TimeoutSeconds)));
-        using var response = await _http.SendAsync(request, cts.Token);
-        var json = await response.Content.ReadAsStringAsync(cts.Token);
-        if (!response.IsSuccessStatusCode) throw new InvalidOperationException($"OpenAI embeddings failed: {(int)response.StatusCode} {json}");
-        using var doc = JsonDocument.Parse(json);
-        return doc.RootElement.GetProperty("data")[0].GetProperty("embedding").EnumerateArray().Select(x => x.GetSingle()).ToArray();
-    }
-
-    private async Task<string> CreateOpenAiChatCompletionAsync(string systemPrompt, string userPrompt, decimal temperature, CancellationToken ct)
-    {
-        if (string.IsNullOrWhiteSpace(_settings.OpenAI.ApiKey)) throw new InvalidOperationException("OpenAI API key is missing.");
-        using var request = new HttpRequestMessage(HttpMethod.Post, "https://api.openai.com/v1/chat/completions");
-        request.Headers.Authorization = new AuthenticationHeaderValue("Bearer", _settings.OpenAI.ApiKey);
-        request.Content = ToJson(new
-        {
-            model = _settings.OpenAI.ChatModel,
-            temperature,
-            response_format = new { type = "json_object" },
-            messages = new[]
-            {
-                new { role = "system", content = systemPrompt },
-                new { role = "user", content = userPrompt }
-            }
-        });
-        using var cts = CancellationTokenSource.CreateLinkedTokenSource(ct);
-        cts.CancelAfter(TimeSpan.FromSeconds(Math.Max(15, _settings.OpenAI.TimeoutSeconds)));
-        using var response = await _http.SendAsync(request, cts.Token);
-        var json = await response.Content.ReadAsStringAsync(cts.Token);
-        if (!response.IsSuccessStatusCode) throw new InvalidOperationException($"OpenAI chat failed: {(int)response.StatusCode} {json}");
-        using var doc = JsonDocument.Parse(json);
-        return doc.RootElement.GetProperty("choices")[0].GetProperty("message").GetProperty("content").GetString() ?? "{}";
-    }
-
-    private async Task<float[]> CreateOllamaEmbeddingAsync(string input, CancellationToken ct)
-    {
-        var baseUrl = _settings.Ollama.BaseUrl.TrimEnd('/');
-        using var cts = CancellationTokenSource.CreateLinkedTokenSource(ct);
-        cts.CancelAfter(TimeSpan.FromSeconds(Math.Max(30, _settings.Ollama.TimeoutSeconds)));
-        using var response = await _http.PostAsync($"{baseUrl}/api/embeddings", ToJson(new { model = _settings.Ollama.EmbeddingModel, prompt = input }), cts.Token);
-        var json = await response.Content.ReadAsStringAsync(cts.Token);
-        if (!response.IsSuccessStatusCode) throw new InvalidOperationException($"Ollama embeddings failed: {(int)response.StatusCode} {json}");
-        using var doc = JsonDocument.Parse(json);
-        return doc.RootElement.GetProperty("embedding").EnumerateArray().Select(x => x.GetSingle()).ToArray();
-    }
-
-    private async Task<string> CreateOllamaChatCompletionAsync(string systemPrompt, string userPrompt, decimal temperature, CancellationToken ct)
-    {
-        var baseUrl = _settings.Ollama.BaseUrl.TrimEnd('/');
-        using var cts = CancellationTokenSource.CreateLinkedTokenSource(ct);
-        cts.CancelAfter(TimeSpan.FromSeconds(Math.Max(30, _settings.Ollama.TimeoutSeconds)));
-        using var response = await _http.PostAsync($"{baseUrl}/api/chat", ToJson(new
-        {
-            model = _settings.Ollama.ChatModel,
-            stream = false,
-            format = "json",
-            messages = new[]
-            {
-                new { role = "system", content = systemPrompt },
-                new { role = "user", content = userPrompt }
-            },
-            options = new { temperature = (float)temperature }
-        }), cts.Token);
-        var json = await response.Content.ReadAsStringAsync(cts.Token);
-        if (!response.IsSuccessStatusCode) throw new InvalidOperationException($"Ollama chat failed: {(int)response.StatusCode} {json}");
-        using var doc = JsonDocument.Parse(json);
-        return doc.RootElement.GetProperty("message").GetProperty("content").GetString() ?? "{}";
-    }
-
-    private static StringContent ToJson<T>(T payload) => new(JsonSerializer.Serialize(payload, JsonOptions), Encoding.UTF8, "application/json");
-}

rag-api/Services/VectorSerializer.cs

@@ -1,31 +0,0 @@
-namespace Api.Services;
-
-public static class VectorSerializer
-{
-    public static byte[] ToBytes(float[] vector)
-    {
-        var bytes = new byte[vector.Length * sizeof(float)];
-        Buffer.BlockCopy(vector, 0, bytes, 0, bytes.Length);
-        return bytes;
-    }
-
-    public static float[] FromBytes(byte[] bytes)
-    {
-        var vector = new float[bytes.Length / sizeof(float)];
-        Buffer.BlockCopy(bytes, 0, vector, 0, bytes.Length);
-        return vector;
-    }
-
-    public static double CosineSimilarity(float[] a, float[] b)
-    {
-        if (a.Length == 0 || a.Length != b.Length) return 0;
-        double dot = 0, magA = 0, magB = 0;
-        for (var i = 0; i < a.Length; i++)
-        {
-            dot += a[i] * b[i];
-            magA += a[i] * a[i];
-            magB += b[i] * b[i];
-        }
-        return magA == 0 || magB == 0 ? 0 : dot / (Math.Sqrt(magA) * Math.Sqrt(magB));
-    }
-}