mirror of
https://github.com/MindWorkAI/AI-Studio.git
synced 2026-10-13 20:53:48 +00:00
Merge branch 'main' into chunk-data
Resolved 29 conflicting files. The notable decisions: Confidence: main's tool-calling gate (RequiredProviderConfidence) and this branch's local-RAG gate (DataConfidenceLevel) turned out to be the same rule on the same axis, so they are now one field. Both tool results and data sources raise it through RequireProviderConfidence(). The gate checks the level strictly and no longer exempts providers trusted by configuration: TrustedProviderIds is documented as applying to data-source security checks only, and organizations set confidence through DataConfidence .CustomConfidenceScheme instead. The security axis (DataSecurity, ERI, IsTrustedForDataSourceSecurityChecks) is unchanged. Provider creation: main's CreateProvider signature won (hfEndpointKind, capabilityOverrides, no model parameter); tokenizerPath was added to it and is set for every provider, including the new Hetzner, IONOS and LiteLLM. Provider and EmbeddingProvider combine the record parameters, Lua parsing and Lua serialization of both sides. File types: main's hierarchy (ODT leaf, WORD parent, PowerPoint without the legacy .ppt, TABULAR instead of DELIMITED_TABLE) plus this branch's SPREADSHEET parent with ODS and the xlsm/xlsb/xla/xlam extensions, which the runtime already reads. Both sides had added a conflicting HTML filter; the reading family keeps the name, and the export path uses a narrow HTML_DOCUMENT, following the existing LATEX/TEX split. Runtime: main's file_data.rs is the base, including the prompt-injection sanitizer and the extraction routes. Token counting and chunk segmentation moved into take_released, so they act on the text the filter has released rather than on text it is still holding. A failed count is logged and left out instead of ending the extraction, because the app counts such a segment itself. Data sources: the participating-provider checks of this branch are kept, and main's GetAllowedDataSources overload now builds on them. DirectChatService resolves the launched chat's data source options before the check, so filter and chat see the same options. .NET and Rust both build clean; I18N regenerated to 4060 keys.
This commit is contained in:
640 files changed
+45059
-4952
No files matched your search
@@ -43,7 +43,7 @@ public sealed class AugmentationOne : IAugmentationProcess
|
||||
{
|
||||
// Let's get the validation agent & set up its provider:
|
||||
var validationAgent = Program.SERVICE_PROVIDER.GetService<AgentRetrievalContextValidation>()!;
|
||||
if (validationAgent.SetLLMProvider(provider, chatThread.DataSecurity, chatThread.DataConfidenceLevel))
|
||||
if (validationAgent.SetLLMProvider(provider, chatThread.DataSecurity, chatThread.RequiredProviderConfidence))
|
||||
{
|
||||
try
|
||||
{
|
||||
|
||||
@@ -1,6 +1,7 @@
|
||||
using System.Text;
|
||||
|
||||
using AIStudio.Chat;
|
||||
using AIStudio.Tools.Security;
|
||||
|
||||
namespace AIStudio.Tools.RAG;
|
||||
|
||||
@@ -13,86 +14,117 @@ public static class IRetrievalContextExtensions
|
||||
sb ??= new StringBuilder();
|
||||
var index = 0;
|
||||
|
||||
//
|
||||
// One report for the whole retrieval run: a query may pull in dozens of contexts, and
|
||||
// the user wants to know that something was filtered, not to acknowledge it per context.
|
||||
//
|
||||
var guardService = Program.SERVICE_PROVIDER.GetRequiredService<PromptInjectionGuardService>();
|
||||
await using var reportingScope = guardService.BeginAction();
|
||||
|
||||
foreach(var retrievalContext in retrievalContexts)
|
||||
{
|
||||
index++;
|
||||
await retrievalContext.AsMarkdown(sb, index, retrievalContexts.Count, token);
|
||||
}
|
||||
|
||||
|
||||
return sb.ToString();
|
||||
}
|
||||
|
||||
public static async Task<string> AsMarkdown(this IRetrievalContext retrievalContext, StringBuilder? sb = null, int index = -1, int numTotalRetrievalContexts = -1, CancellationToken token = default)
|
||||
{
|
||||
sb ??= new StringBuilder();
|
||||
var contextBuilder = new StringBuilder();
|
||||
switch (index)
|
||||
{
|
||||
case > 0 when numTotalRetrievalContexts is -1:
|
||||
sb.AppendLine($"# Retrieval context {index}");
|
||||
contextBuilder.AppendLine($"# Retrieval context {index}");
|
||||
break;
|
||||
|
||||
case > 0 when numTotalRetrievalContexts > 0:
|
||||
sb.AppendLine($"# Retrieval context {index} of {numTotalRetrievalContexts}");
|
||||
contextBuilder.AppendLine($"# Retrieval context {index} of {numTotalRetrievalContexts}");
|
||||
break;
|
||||
|
||||
default:
|
||||
sb.AppendLine("# Retrieval context");
|
||||
contextBuilder.AppendLine("# Retrieval context");
|
||||
break;
|
||||
}
|
||||
|
||||
sb.AppendLine($"Data source name: {retrievalContext.DataSourceName}");
|
||||
sb.AppendLine($"Content category: {retrievalContext.Category}");
|
||||
sb.AppendLine($"Content type: {retrievalContext.Type}");
|
||||
sb.AppendLine($"Content path: {retrievalContext.Path}");
|
||||
contextBuilder.AppendLine($"Data source name: {retrievalContext.DataSourceName}");
|
||||
contextBuilder.AppendLine($"Content category: {retrievalContext.Category}");
|
||||
contextBuilder.AppendLine($"Content type: {retrievalContext.Type}");
|
||||
contextBuilder.AppendLine($"Content path: {retrievalContext.Path}");
|
||||
|
||||
if(retrievalContext.Links.Count > 0)
|
||||
{
|
||||
sb.AppendLine("Additional links:");
|
||||
contextBuilder.AppendLine("Additional links:");
|
||||
foreach(var link in retrievalContext.Links)
|
||||
sb.AppendLine($"- {link}");
|
||||
contextBuilder.AppendLine($"- {link}");
|
||||
}
|
||||
|
||||
|
||||
var guardService = Program.SERVICE_PROVIDER.GetRequiredService<PromptInjectionGuardService>();
|
||||
var source = PromptInjectionSource.RetrievalContext(retrievalContext.DataSourceName, retrievalContext.Path);
|
||||
|
||||
switch(retrievalContext)
|
||||
{
|
||||
case RetrievalTextContext textContext:
|
||||
sb.AppendLine();
|
||||
sb.AppendLine("Matched text content:");
|
||||
sb.AppendLine("````");
|
||||
sb.AppendLine(textContext.MatchedText);
|
||||
sb.AppendLine("````");
|
||||
|
||||
contextBuilder.AppendLine();
|
||||
contextBuilder.AppendLine("Matched text content:");
|
||||
contextBuilder.AppendLine("````");
|
||||
contextBuilder.AppendLine(textContext.MatchedText);
|
||||
contextBuilder.AppendLine("````");
|
||||
|
||||
if(textContext.SurroundingContent.Count > 0)
|
||||
{
|
||||
sb.AppendLine();
|
||||
sb.AppendLine("Surrounding text content:");
|
||||
contextBuilder.AppendLine();
|
||||
contextBuilder.AppendLine("Surrounding text content:");
|
||||
foreach(var surrounding in textContext.SurroundingContent)
|
||||
{
|
||||
sb.AppendLine();
|
||||
sb.AppendLine("````");
|
||||
sb.AppendLine(surrounding);
|
||||
sb.AppendLine("````");
|
||||
contextBuilder.AppendLine();
|
||||
contextBuilder.AppendLine("````");
|
||||
contextBuilder.AppendLine(surrounding);
|
||||
contextBuilder.AppendLine("````");
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
|
||||
await FilterWhatWeHaveSoFar();
|
||||
break;
|
||||
|
||||
|
||||
case RetrievalImageContext imageContext:
|
||||
sb.AppendLine();
|
||||
sb.AppendLine("Matched image content as base64-encoded data:");
|
||||
sb.AppendLine("````");
|
||||
sb.AppendLine(await imageContext.TryAsBase64(token) is (success: true, { } base64Image)
|
||||
? base64Image
|
||||
//
|
||||
// Filtering happens before the image is appended, and only covers the text
|
||||
// around it. Base64 image data is not prose, and running it through the filter
|
||||
// would have it treated as one enormous encoded carrier.
|
||||
//
|
||||
await FilterWhatWeHaveSoFar();
|
||||
contextBuilder.AppendLine();
|
||||
contextBuilder.AppendLine("Matched image content as base64-encoded data:");
|
||||
contextBuilder.AppendLine("````");
|
||||
contextBuilder.AppendLine(await imageContext.TryAsBase64(token) is (success: true, { } base64Image)
|
||||
? base64Image
|
||||
: string.Empty);
|
||||
sb.AppendLine("````");
|
||||
contextBuilder.AppendLine("````");
|
||||
break;
|
||||
|
||||
|
||||
default:
|
||||
await FilterWhatWeHaveSoFar();
|
||||
LOGGER.LogWarning($"The retrieval content type '{retrievalContext.Type}' of data source '{retrievalContext.DataSourceName}' at location '{retrievalContext.Path}' is not supported yet.");
|
||||
break;
|
||||
}
|
||||
|
||||
sb.AppendLine();
|
||||
contextBuilder.AppendLine();
|
||||
sb.Append(contextBuilder);
|
||||
return sb.ToString();
|
||||
|
||||
//
|
||||
// Replaces what has been built so far with its filtered version. A data source is as
|
||||
// untrusted as any other external content: it may serve text written to steer the model
|
||||
// rather than to answer the query.
|
||||
//
|
||||
async Task FilterWhatWeHaveSoFar()
|
||||
{
|
||||
var sanitized = await guardService.SanitizeAsync(contextBuilder.ToString(), source);
|
||||
contextBuilder.Clear();
|
||||
contextBuilder.Append(sanitized);
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -104,7 +104,7 @@ public sealed class AISrcSelWithRetCtxVal : IRagProcess
|
||||
else
|
||||
{
|
||||
var previousDataSecurity = chatThread.DataSecurity;
|
||||
var previousDataConfidenceLevel = chatThread.DataConfidenceLevel;
|
||||
var previousRequiredProviderConfidence = chatThread.RequiredProviderConfidence;
|
||||
|
||||
//
|
||||
// Update the data security of the chat thread. We consider the current data security
|
||||
@@ -155,11 +155,10 @@ public sealed class AISrcSelWithRetCtxVal : IRagProcess
|
||||
LOGGER.LogInformation($"The data security of the chat thread was updated from '{previousDataSecurity}' to '{chatThread.DataSecurity}'.");
|
||||
|
||||
foreach (var dataSource in selectedDataSources.OfType<IInternalDataSource>())
|
||||
if (dataSource.ConfidenceLevel > chatThread.DataConfidenceLevel)
|
||||
chatThread.DataConfidenceLevel = dataSource.ConfidenceLevel;
|
||||
chatThread.RequireProviderConfidence(dataSource.ConfidenceLevel);
|
||||
|
||||
if (previousDataConfidenceLevel != chatThread.DataConfidenceLevel)
|
||||
LOGGER.LogInformation($"The data confidence level of the chat thread was updated from '{previousDataConfidenceLevel.GetName()}' to '{chatThread.DataConfidenceLevel.GetName()}'.");
|
||||
if (previousRequiredProviderConfidence != chatThread.RequiredProviderConfidence)
|
||||
LOGGER.LogInformation($"The required provider confidence of the chat thread was updated from '{previousRequiredProviderConfidence.GetName()}' to '{chatThread.RequiredProviderConfidence.GetName()}'.");
|
||||
}
|
||||
|
||||
//
|
||||
|
||||
Reference in new issue
Block a user