tests and config

Author: KSemenenko

AGENTS.md

@@ -18,6 +18,7 @@
 - Do not add trivial POCO/unit tests (e.g., pure record equality); focus tests on meaningful flows and integration scenarios.
 - When registering graph stores (especially Postgres) use `AddKeyedSingleton<T>(key)` without bespoke lambdas unless strictly required; rely on DI to construct services.
 - All integration tests must run against real dependencies via Testcontainers modules (Neo4j, Postgres/AGE, Cosmos, Janus, etc.); do not fall back to Aspire seeding or mock containers.
+- Keep integration tests provider-agnostic; avoid adding Postgres-only scenarios when the same flow should apply to every backend.
 
 # Conversations
 any resulting updates to agents.md should go under the section "## Rules to follow"

README.md

@@ -164,6 +164,38 @@ See [`docs/indexing-and-query.md`](docs/indexing-and-query.md) for the full list
 
 ---
 
+## Configuration Parity
+
+The .NET configuration surface now mirrors the original Python CLI. `GraphRagConfig` exposes the same sections as `graphrag.config`, including cache providers, NLP-driven graph extraction, claim extraction, graph pruning, and every search mode (local/global/DRIFT/basic). This makes it straightforward to move existing Python configs across without rethinking every knob:
+
+```json
+{
+  "GraphRag": {
+    "Cache": { "Type": "File", "BaseDir": "cache" },
+    "ExtractGraphNlp": {
+      "ConcurrentRequests": 25,
+      "TextAnalyzer": { "ModelName": "en_core_web_md", "IncludeNamedEntities": true }
+    },
+    "ExtractClaims": {
+      "Enabled": true,
+      "ModelId": "chat_model",
+      "Prompt": "prompts/claims.txt",
+      "MaxGleanings": 2
+    },
+    "PruneGraph": { "MinNodeFrequency": 2, "MinEdgeWeightPercentile": 40 },
+    "EmbedGraph": { "Enabled": false, "Dimensions": 1536 },
+    "LocalSearch": { "ChatModelId": "chat_model", "EmbeddingModelId": "embedding_model" },
+    "GlobalSearch": { "MapPrompt": "prompts/global_map.txt", "ReducePrompt": "prompts/global_reduce.txt" },
+    "DriftSearch": { "Prompt": "prompts/drift.txt", "Concurrency": 32 },
+    "BasicSearch": { "K": 10 }
+  }
+}
+```
+
+`ClaimExtractionConfig.GetResolvedStrategy` mirrors the Python behaviour by loading prompt files from the configured root (and throwing if the file is missing) while still letting you override the entire `Strategy` block when needed.
+
+---
+
 ## Community Detection & Graph Analytics
 
 Community creation defaults to the fast label propagation algorithm. Tweak clustering directly through configuration:

src/ManagedCode.GraphRag.Postgres/AssemblyInfo.cs

@@ -0,0 +1,3 @@
+using System.Runtime.CompilerServices;
+
+[assembly: InternalsVisibleTo("ManagedCode.GraphRag.Tests")]

src/ManagedCode.GraphRag.Postgres/PostgresGraphStore.cs

@@ -574,7 +574,7 @@ private static (IReadOnlyDictionary<string, object?> Writes, IReadOnlyCollection
             writes[key] = value;
         }
 
-        IReadOnlyDictionary<string, object?> writeResult = writes.Count == 0 ? EmptyProperties : writes;
+        var writeResult = writes.Count == 0 ? EmptyProperties : writes;
         IReadOnlyCollection<string> removeResult = removes.Count == 0 ? Array.Empty<string>() : removes;
         return (writeResult, removeResult);
     }

src/ManagedCode.GraphRag/Config/BasicSearchConfig.cs

@@ -0,0 +1,14 @@
+namespace GraphRag.Config;
+
+public sealed class BasicSearchConfig
+{
+    public string? Prompt { get; set; }
+
+    public string ChatModelId { get; set; } = "default_chat_model";
+
+    public string EmbeddingModelId { get; set; } = "default_embedding_model";
+
+    public int K { get; set; } = 10;
+
+    public int MaxContextTokens { get; set; } = 12_000;
+}

src/ManagedCode.GraphRag/Config/CacheConfig.cs

@@ -0,0 +1,16 @@
+namespace GraphRag.Config;
+
+public sealed class CacheConfig
+{
+    public CacheType Type { get; set; } = CacheType.File;
+
+    public string BaseDir { get; set; } = "cache";
+
+    public string? ConnectionString { get; set; }
+
+    public string? ContainerName { get; set; }
+
+    public string? StorageAccountBlobUrl { get; set; }
+
+    public string? CosmosDbAccountUrl { get; set; }
+}

src/ManagedCode.GraphRag/Config/ClaimExtractionConfig.cs

@@ -0,0 +1,53 @@
+namespace GraphRag.Config;
+
+public sealed class ClaimExtractionConfig
+{
+    private const string DefaultDescription = "Any claims or facts that could be relevant to information discovery.";
+
+    public bool Enabled { get; set; }
+
+    public string ModelId { get; set; } = "default_chat_model";
+
+    public string? Prompt { get; set; }
+
+    public string Description { get; set; } = DefaultDescription;
+
+    public int MaxGleanings { get; set; } = 1;
+
+    public Dictionary<string, object?>? Strategy { get; set; }
+        = new(StringComparer.OrdinalIgnoreCase);
+
+    public Dictionary<string, object?> GetResolvedStrategy(string? rootDirectory = null)
+    {
+        if (Strategy is { Count: > 0 })
+        {
+            return new Dictionary<string, object?>(Strategy, StringComparer.OrdinalIgnoreCase);
+        }
+
+        string? promptPayload = null;
+        if (!string.IsNullOrWhiteSpace(Prompt))
+        {
+            var baseDir = string.IsNullOrWhiteSpace(rootDirectory)
+                ? Directory.GetCurrentDirectory()
+                : rootDirectory!;
+            var fullPath = Path.IsPathRooted(Prompt!)
+                ? Prompt!
+                : Path.Combine(baseDir, Prompt!);
+
+            if (!File.Exists(fullPath))
+            {
+                throw new FileNotFoundException($"Claim extraction prompt '{fullPath}' could not be found.", fullPath);
+            }
+
+            promptPayload = File.ReadAllText(fullPath);
+        }
+
+        return new Dictionary<string, object?>(StringComparer.OrdinalIgnoreCase)
+        {
+            ["model_id"] = ModelId,
+            ["extraction_prompt"] = promptPayload,
+            ["claim_description"] = Description,
+            ["max_gleanings"] = MaxGleanings
+        };
+    }
+}

src/ManagedCode.GraphRag/Config/DriftSearchConfig.cs

@@ -0,0 +1,50 @@
+namespace GraphRag.Config;
+
+public sealed class DriftSearchConfig
+{
+    public string? Prompt { get; set; }
+
+    public string? ReducePrompt { get; set; }
+
+    public string ChatModelId { get; set; } = "default_chat_model";
+
+    public string EmbeddingModelId { get; set; } = "default_embedding_model";
+
+    public int DataMaxTokens { get; set; } = 12_000;
+
+    public int? ReduceMaxTokens { get; set; }
+
+    public double ReduceTemperature { get; set; }
+
+    public int? ReduceMaxCompletionTokens { get; set; }
+
+    public int Concurrency { get; set; } = 32;
+
+    public int DriftFollowupCount { get; set; } = 20;
+
+    public int PrimerFolds { get; set; } = 5;
+
+    public int PrimerMaxTokens { get; set; } = 12_000;
+
+    public int Depth { get; set; } = 3;
+
+    public double LocalSearchTextUnitProportion { get; set; } = 0.9;
+
+    public double LocalSearchCommunityProportion { get; set; } = 0.1;
+
+    public int LocalSearchTopKEntities { get; set; } = 10;
+
+    public int LocalSearchTopKRelationships { get; set; } = 10;
+
+    public int LocalSearchMaxDataTokens { get; set; } = 12_000;
+
+    public double LocalSearchTemperature { get; set; }
+
+    public double LocalSearchTopP { get; set; } = 1.0;
+
+    public int LocalSearchSampleCount { get; set; } = 1;
+
+    public int? LocalSearchMaxGeneratedTokens { get; set; }
+
+    public int? LocalSearchMaxCompletionTokens { get; set; }
+}

src/ManagedCode.GraphRag/Config/EmbedGraphConfig.cs

@@ -0,0 +1,20 @@
+namespace GraphRag.Config;
+
+public sealed class EmbedGraphConfig
+{
+    public bool Enabled { get; set; }
+
+    public int Dimensions { get; set; } = 1536;
+
+    public int WalkLength { get; set; } = 40;
+
+    public int NumWalks { get; set; } = 10;
+
+    public int WindowSize { get; set; } = 2;
+
+    public int Iterations { get; set; } = 3;
+
+    public int RandomSeed { get; set; } = 597_832;
+
+    public bool UseLargestConnectedComponent { get; set; } = true;
+}

src/ManagedCode.GraphRag/Config/Enums.cs

@@ -7,6 +7,15 @@ public enum InputFileType
     Json
 }
 
+public enum CacheType
+{
+    File,
+    Memory,
+    None,
+    Blob,
+    CosmosDb
+}
+
 public enum StorageType
 {
     File,
@@ -34,6 +43,12 @@ public enum ChunkStrategyType
     Sentence
 }
 
+public enum AsyncType
+{
+    Threaded,
+    AsyncIO
+}
+
 public enum SearchMethod
 {
     Local,