repo stringlengths 5 53 | pr_number int32 1 321k | task_type stringclasses 2
values | issue_text stringlengths 0 81.2k | pr_title stringlengths 1 319 | pr_body stringlengths 0 105k | base_sha stringlengths 40 40 | head_sha stringlengths 40 40 | gold_diff stringlengths 0 202M | changed_files listlengths 0 100 | review_threads listlengths 0 100 | test_patch stringlengths 0 23.4M | merged bool 1
class |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
MathewSachin/Captura | 445 | issue_to_patch | Migrate WPF project to the new SDK format | afe3ef62a452a2e83bb85001dc6092dfcc9b065d | 8506770e16ed6aefe72828a88139a26e7fbb36ff | diff --git a/src/Captura.Loc/Captura.Loc.csproj b/src/Captura.Loc/Captura.Loc.csproj
index 2ffe52af5..90880a39e 100644
--- a/src/Captura.Loc/Captura.Loc.csproj
+++ b/src/Captura.Loc/Captura.Loc.csproj
@@ -2,11 +2,6 @@
<PropertyGroup>
<TargetFramework>netstandard2.0</TargetFramework>
</PropertyGroup>
- <Item... | [
"src/Captura.Loc/Captura.Loc.csproj",
"src/Captura.NAudio/Captura.NAudio.csproj",
"src/Captura.SharpAvi/Captura.SharpAvi.csproj",
"src/Captura.ViewCore/Captura.ViewCore.csproj",
"src/Captura/Captura.csproj",
"src/PostBuild.targets"
] | [] | true | |||
MathewSachin/Captura | 444 | issue_to_patch | New Crowdin translations | Enjoy! | 29d4985aedd74bdf116af5d99c017667216380ab | 3b86d68b0925290501e9c807d3fee50e8650808c | diff --git a/src/Captura.Loc/Languages/ca.json b/src/Captura.Loc/Languages/ca.json
new file mode 100644
index 000000000..0f0be985c
--- /dev/null
+++ b/src/Captura.Loc/Languages/ca.json
@@ -0,0 +1,160 @@
+{
+ "About": "En quant a",
+ "AccentColor": "Color d'èmfasi",
+ "Add": "Afegir",
+ "AlwaysOnTop": "Sempre a la p... | [
"src/Captura.Loc/Languages/ca.json"
] | [] | true | ||
MathewSachin/Captura | 391 | issue_to_patch | Steps Recorder | Requested in #333.
Adding a record steps mode to Captura.
Similar consecutive events are merged into a single frame.
e.g. If you're typing **hello**, then it will be shown in a single frame instead of taking 5 frames with each character on separate frame.
## Status
- Steps
- [x] Click
- [x] Show modi... | 8a0cabf298ff7c30951f952183812ca2a3c08bac | 03af9ed7da1784ae0c8a51f8c1d6e08ba5ad24fa | diff --git a/docs/Cmdline/Verb-Start.md b/docs/Cmdline/Verb-Start.md
index e86806dce..0cdecf449 100644
--- a/docs/Cmdline/Verb-Start.md
+++ b/docs/Cmdline/Verb-Start.md
@@ -73,4 +73,19 @@ e.g. Stream to Twitch
```
captura-cli start --encoder stream:rtmp://live.twitch.tv/app/TWITCH_KEY
+```
+
+### Steps
+Use `steps:... | [
"docs/Cmdline/Verb-Start.md",
"licenses/System.Reactive.txt",
"src/Captura.Base/Images/IEditableFrame.cs",
"src/Captura.Base/Images/RepeatFrame.cs",
"src/Captura.Base/Services/IPlatformServices.cs",
"src/Captura.Base/Settings/StepsSettings.cs",
"src/Captura.Base/Settings/VideoSettings.cs",
"src/Captur... | [] | true | ||
MathewSachin/Captura | 441 | issue_to_patch | Fix dependency properties registration | According to PVS-Studio, there is a potential error related to dependency property registration.
see https://www.viva64.com/en/w/v3047 | 2d7e19025c388c3721191d26f5b2ed77ae37e69e | 861a5218c647adf553645449bbc64a0c6c6a9a6f | diff --git a/src/Captura/Presentation/PuncturedRect.cs b/src/Captura/Presentation/PuncturedRect.cs
index 0ca4888ce..50b2057aa 100644
--- a/src/Captura/Presentation/PuncturedRect.cs
+++ b/src/Captura/Presentation/PuncturedRect.cs
@@ -11,7 +11,7 @@ public class PuncturedRect : Shape
DependencyProperty.Regist... | [
"src/Captura/Presentation/PuncturedRect.cs"
] | [] | true | ||
MathewSachin/Captura | 383 | issue_to_patch | Developer Documentation | 3352832fcf0a676a499a4cde9ca067887e0c981f | 253af3a121316d3eebc75d41f5b0ff9f8ef76197 | diff --git a/.github/ISSUE_TEMPLATE/bug_report.md b/.github/ISSUE_TEMPLATE/bug_report.md
index 35c83fb21..db20099dd 100644
--- a/.github/ISSUE_TEMPLATE/bug_report.md
+++ b/.github/ISSUE_TEMPLATE/bug_report.md
@@ -27,7 +27,7 @@ If applicable, add screenshots to help explain your problem.
- RAM: [e.g. 4GB]
- CPU: [e.... | [
".github/ISSUE_TEMPLATE/bug_report.md",
".github/PULL_REQUEST_TEMPLATE.md",
"CODE_OF_CONDUCT.md",
"CONTRIBUTING.md",
"README.md",
"SUPPORT.md",
"docs/Acknowledgements.md",
"docs/Build.md",
"docs/CI.md",
"docs/Cake.md",
"docs/Changelogs/README.md",
"docs/Changelogs/v7.0.1.md",
"docs/Changelog... | [] | true | |||
MathewSachin/Captura | 420 | issue_to_patch | Audio subsystem redesign | - Remove BASS support.
- Added a `Captura.Audio` project which is referenced by NAudio and ~BASS projects~.
- Redesigned `IAudioProvider`.
- Only 1 microphone + 1 speaker at a time.
- Added an Audio page to settings.
- Added `Default` devices for microphone and speaker.
- `Include Cursor` and other overlay toggle... | 815dc86609cd58df581611431b5058a16a75fff6 | e5fc7dc770a59a9460c5075d805b518c4c7c4d87 | diff --git a/Inno.iss b/Inno.iss
index 2744b6940..029aa8371 100644
--- a/Inno.iss
+++ b/Inno.iss
@@ -44,29 +44,7 @@ Name: "desktopicon"; Description: "{cm:CreateDesktopIcon}"; GroupDescription: "{
; Remove Assemblies from previous installation to prevent conflicts
[InstallDelete]
-Type: files; Name: "{app}\bass.dll... | [
"Inno.iss",
"appveyor.yml",
"build.cake",
"licenses/BASS.txt",
"licenses/Captura, Screna.txt",
"scripts/bass.cake",
"src/Captura.Audio/Captura.Audio.csproj",
"src/Captura.Audio/DataAvailableEventArgs.cs",
"src/Captura.Audio/IAudioFileWriter.cs",
"src/Captura.Audio/IAudioItem.cs",
"src/Captura.Au... | [] | true | ||
MathewSachin/Captura | 388 | issue_to_patch | New Crowdin translations | Enjoy! | e0678d71e042e62a5c309c65b3d658fe2e26266c | 18b7e047e172ccb2892abff559db26d4b6d9e4cf | diff --git a/src/Captura.Loc/Languages/he.json b/src/Captura.Loc/Languages/he.json
index 6211ef2b2..92a91c00c 100644
--- a/src/Captura.Loc/Languages/he.json
+++ b/src/Captura.Loc/Languages/he.json
@@ -10,17 +10,17 @@
"BorderColor": "צבע גבולות",
"BorderThickness": "עובי גבולות",
"Bottom": "תחתון",
- "CaptureD... | [
"src/Captura.Loc/Languages/he.json",
"src/Captura.Loc/Languages/ko.json"
] | [] | true | ||
MathewSachin/Captura | 411 | issue_to_patch | Convert from RGB to NV12 on GPU before sending frames to FFmpeg | Size of RGB32 frame: Width * Height * 4
Size of NV12 frame: Width * Height * 1.5
Lesser data copied from GPU to CPU, lesser CPU usage.
Need to use `Media Foundation .NET` due to a bug in `SharpDX.MediaFoundation` because `SharpDX` is no longer maintained.
Replaced `IImageProvider.EditorType` with `DummyFrame`... | 78a1d15367de562c1f1b352dd324ea438eeaa879 | 90d2f6f8844f20700110ad2ecbaa2b6a2eecd9f1 | diff --git a/licenses/Media Foundation .NET.txt b/licenses/Media Foundation .NET.txt
new file mode 100644
index 000000000..c645affcb
--- /dev/null
+++ b/licenses/Media Foundation .NET.txt
@@ -0,0 +1,481 @@
+ GNU LIBRARY GENERAL PUBLIC LICENSE
+ Version 2, June 1991
+
+ Copyright (... | [
"licenses/Media Foundation .NET.txt",
"src/Captura.Base/Images/IImageProvider.cs",
"src/Captura.Base/Images/INV12Frame.cs",
"src/Captura.Core/WebcamImageProvider.cs",
"src/Captura.FFmpeg/Video/FFmpegVideoWriter.cs",
"src/Captura.Windows/Captura.Windows.csproj",
"src/Captura.Windows/Capture/DxgiTargetDev... | [] | true | ||
MathewSachin/Captura | 418 | issue_to_patch | Post-processing video converters | Post-processing writers are now handled differently.
These implement `IVideoConverter`.
`FFmpegPostProcessingWriter` and `FFmpegGifWriter` have been changed.
Progress is displayed in the notifications and after conversion filename is updated.
Video Encoder options have been moved to the Settings page which now in... | e0678d71e042e62a5c309c65b3d658fe2e26266c | 0a48862dab252ae68d5f98499fd5f23c018332a8 | diff --git a/src/Captura.Base/Video/IVideoConverter.cs b/src/Captura.Base/Video/IVideoConverter.cs
new file mode 100644
index 000000000..214c05b52
--- /dev/null
+++ b/src/Captura.Base/Video/IVideoConverter.cs
@@ -0,0 +1,14 @@
+using System;
+using System.Threading.Tasks;
+
+namespace Captura.Models
+{
+ public inte... | [
"src/Captura.Base/Video/IVideoConverter.cs",
"src/Captura.Base/Video/VideoConverterParams.cs",
"src/Captura.Core/Models/Notifications/FileSaveNotification.cs",
"src/Captura.Core/Models/Recents/FileRecentItem.cs",
"src/Captura.Core/Settings/Models/VideoSettings.cs",
"src/Captura.FFmpeg/FFMpegLogItem.cs",
... | [] | true | ||
MathewSachin/Captura | 417 | issue_to_patch | Screen picker dialog | Shows a dialog box with previews and names of screens instead of a fullscreen selector | 2e201d8333d6ab9e2bc5854c7c4a7ae9b3d14090 | 80cab9296061f1beefbf9c6a09b54b3de88b7205 | diff --git a/src/Captura.ViewCore/ViewModels/VideoSourcesViewModel.cs b/src/Captura.ViewCore/ViewModels/VideoSourcesViewModel.cs
index d3eebaedb..09f8bd75f 100644
--- a/src/Captura.ViewCore/ViewModels/VideoSourcesViewModel.cs
+++ b/src/Captura.ViewCore/ViewModels/VideoSourcesViewModel.cs
@@ -55,18 +55,8 @@ void ChangeS... | [
"src/Captura.ViewCore/ViewModels/VideoSourcesViewModel.cs",
"src/Captura/Captura.csproj",
"src/Captura/Controls/VideoSourceKindList.xaml",
"src/Captura/Models/VideoSourcePicker.cs",
"src/Captura/ViewModels/ScreenPickerViewModel.cs",
"src/Captura/Windows/ScreenPickerWindow.xaml",
"src/Captura/Windows/Scr... | [] | true | ||
MathewSachin/Captura | 415 | issue_to_patch | Update Hebrew language | 9c15e1f7bcc49f408f883b74d29497f6eb37d14d | 750507fc0b9d2a8e66b88419d31daec0d43b800d | diff --git a/src/Captura.Loc/Languages/he.json b/src/Captura.Loc/Languages/he.json
index aa4586d03..6211ef2b2 100644
--- a/src/Captura.Loc/Languages/he.json
+++ b/src/Captura.Loc/Languages/he.json
@@ -1,10 +1,10 @@
{
"About": "אודות",
- "AccentColor": "",
- "Add": "",
+ "AccentColor": "צבע הדגשה",
+ "Add": "הוס... | [
"src/Captura.Loc/Languages/he.json"
] | [] | true | |||
MathewSachin/Captura | 413 | issue_to_patch | Fix: Negative Window handles not supported by CLI | #412 | 8c3096f5b3e5b8dfb67c3b6a8d6ea465c11e4bf7 | 15b89041489263e1238b3040288beb49869e4590 | diff --git a/src/Screna/VideoSourceProviders/WindowSourceProvider.cs b/src/Screna/VideoSourceProviders/WindowSourceProvider.cs
index 80bc49339..c956e07e8 100644
--- a/src/Screna/VideoSourceProviders/WindowSourceProvider.cs
+++ b/src/Screna/VideoSourceProviders/WindowSourceProvider.cs
@@ -75,7 +75,7 @@ public override b... | [
"src/Screna/VideoSourceProviders/WindowSourceProvider.cs"
] | [] | true | ||
MathewSachin/Captura | 403 | issue_to_patch | Move to .Net 4.7.2 | Moving to .Net 4.7.1 or above reduces the number of dependency DLLs which need to be distributed with Captura.
.Net 4.7.2 is available for Windows 7 SP1, Windows 8.1 and Windows 10 Anniversary update and above so that might not be a problem.
Number of dlls depended on by Captura is reduced by approx. 100 and 4 MB... | eb73ec4201161db6dfede153b1d8afa1f6480e83 | 6f304eed707c0df0fad8fd12b9f0ad9087c4f151 | diff --git a/README.md b/README.md
index 98928a234..6d28cc44a 100644
--- a/README.md
+++ b/README.md
@@ -78,7 +78,7 @@ Dev builds can be unstable and should be used for testing purposes only.
## Build Notes
- [Visual Studio 2017](https://visualstudio.com) or greater is required.
-- .Net Framework v4.6.1 is required... | [
"README.md",
"src/Captura.Console/App.config",
"src/Captura.Console/Captura.Console.csproj",
"src/Captura.Core/Captura.Core.csproj",
"src/Captura.Hotkeys/Captura.Hotkeys.csproj",
"src/Captura.MouseKeyHook/Captura.MouseKeyHook.csproj",
"src/Captura.NAudio/Captura.NAudio.csproj",
"src/Captura.SharpAvi/C... | [] | diff --git a/src/Tests/Tests.csproj b/src/Tests/Tests.csproj
index 6bf4f9705..ec2ef0937 100644
--- a/src/Tests/Tests.csproj
+++ b/src/Tests/Tests.csproj
@@ -1,7 +1,7 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<AssemblyName>Captura.Tests</AssemblyName>
- <TargetFramework>net461</TargetFramework>
+... | true | |
MathewSachin/Captura | 397 | issue_to_patch | Using Direct2DEditor with GDI capture | Using `Direct2DEditor` along with the `Media Foundation` encoder further increases performance for GDI captures.
- [x] Implement for `RegionProvider`
- [x] Implement for `WindowProvider`
- [x] Support Mouse Pointer | 56c18ac694cd3168d16995dae53bfc1fa864f166 | fca665b2129a9ddeedd1195a89bfd12dd6e1e931 | diff --git a/src/Captura.Core/CoreModule.cs b/src/Captura.Core/CoreModule.cs
index 04e15d542..4080832af 100644
--- a/src/Captura.Core/CoreModule.cs
+++ b/src/Captura.Core/CoreModule.cs
@@ -12,7 +12,7 @@ public void OnLoad(IBinder Binder)
{
Binder.Bind<IAudioWriterItem, WaveItem>();
- ... | [
"src/Captura.Core/CoreModule.cs",
"src/Captura.Windows/Capture/DxgiTargetDeviceContext.cs",
"src/Captura.Windows/Capture/GdiTargetDeviceContext.cs",
"src/Captura.Windows/Capture/ITargetDeviceContext.cs",
"src/Captura.Windows/Capture/MouseCursor.cs",
"src/Captura.Windows/Capture/RegionProvider.cs",
"src/... | [] | true | ||
MathewSachin/Captura | 393 | issue_to_patch | Around Mouse Recording | A new mode for recording the region around mouse cursor. | 52900a2a8c8ec752652936b953131911c5f580c4 | 7abe448b49d3052d0f995bf9ea5130d0fdc2020d | diff --git a/src/Captura.Core/CoreModule.cs b/src/Captura.Core/CoreModule.cs
index 5d81a9517..04268929b 100644
--- a/src/Captura.Core/CoreModule.cs
+++ b/src/Captura.Core/CoreModule.cs
@@ -84,6 +84,7 @@ static void BindAudioSource(IBinder Binder)
static void BindVideoSourceProviders(IBinder Binder)
{
... | [
"src/Captura.Core/CoreModule.cs",
"src/Captura.Core/Models/AroundMouseItem.cs",
"src/Captura.Core/Models/AroundMouseSourceProvider.cs",
"src/Captura.Core/Settings/Settings.cs",
"src/Captura.Core/ViewModels/ScreenShotModel.cs",
"src/Captura.ViewCore/ViewModels/ViewConditionsModel.cs",
"src/Captura/Pages/... | [] | true | ||
MathewSachin/Captura | 395 | issue_to_patch | MediaFoundation SinkWriter | Using `MediaFoundation` `SinkWriter` considerably reduces CPU usage.
- Media Foundation made default writer.
- Works best with `Desktop Duplication`.
### Api Changes
- Added `IBitmapFrame.CopyTo` overload accepting `IntPtr`.
- Added `IFrameWrapper` interface. This is implemented by `MultiDisposeFrame`.
- Adde... | 0f093d1685aa0afacbc3cdc73009a0da484f644b | 575691eae4104b9ee47e5313101677ae41fa3b50 | diff --git a/src/Captura.Base/Images/IBitmapFrame.cs b/src/Captura.Base/Images/IBitmapFrame.cs
index cc5fb1105..e97685eac 100644
--- a/src/Captura.Base/Images/IBitmapFrame.cs
+++ b/src/Captura.Base/Images/IBitmapFrame.cs
@@ -9,5 +9,7 @@ public interface IBitmapFrame : IDisposable
int Height { get; }
... | [
"src/Captura.Base/Images/IBitmapFrame.cs",
"src/Captura.Base/Images/IFrameWrapper.cs",
"src/Captura.Base/Images/IImageProvider.cs",
"src/Captura.Base/Images/RepeatFrame.cs",
"src/Captura.Core/CoreModule.cs",
"src/Captura.Core/Settings/Models/VideoSettings.cs",
"src/Captura.Core/WebcamImageProvider.cs",
... | [] | true | ||
MathewSachin/Captura | 386 | issue_to_patch | New Crowdin translations | Enjoy! | 53f035d04df5ac537e065db16e47ba7223985816 | 3189e88c6559f41bd94a5454f3cd01b468e02993 | diff --git a/src/Captura.Loc/Languages/pl.json b/src/Captura.Loc/Languages/pl.json
index deeb98ca3..d6ccf026e 100644
--- a/src/Captura.Loc/Languages/pl.json
+++ b/src/Captura.Loc/Languages/pl.json
@@ -1,6 +1,6 @@
{
"About": "O programie",
- "AccentColor": "Kolor akcentujący",
+ "AccentColor": "Kolor interfejsu",
... | [
"src/Captura.Loc/Languages/pl.json"
] | [] | true | ||
MathewSachin/Captura | 385 | issue_to_patch | Rolling Recording (Capture last N seconds) | Requested in #182
Capture last N seconds mode added for Console.
This feature will **NOT** be implemented for the UI version due to the involved complications.
Implemented as a new encoder: `roll:<duration>` where duration is specified in seconds.
```
captura-cli start --encoder roll:60
```
Separate fil... | 53f035d04df5ac537e065db16e47ba7223985816 | 1273568786a15b9e46b4e87e7b3e2b9d67518a88 | diff --git a/src/Captura.Console/ConsoleManager.cs b/src/Captura.Console/ConsoleManager.cs
index 0f0c57988..2aa52e7ba 100644
--- a/src/Captura.Console/ConsoleManager.cs
+++ b/src/Captura.Console/ConsoleManager.cs
@@ -6,6 +6,7 @@
using System.Threading;
using System.Threading.Tasks;
using System.Windows.Forms;
+using... | [
"src/Captura.Console/ConsoleManager.cs",
"src/Captura.Console/Properties/launchSettings.json",
"src/Captura.FFmpeg/FFMpegLogItem.cs",
"src/Captura.FFmpeg/FFmpegModule.cs",
"src/Captura.FFmpeg/FFmpegService.cs",
"src/Captura.FFmpeg/IFFmpegLogEntry.cs",
"src/Captura.FFmpeg/IFFmpegLogRepository.cs",
"src... | [] | true | ||
MathewSachin/Captura | 384 | issue_to_patch | Font setting for Overlays | Requested in #363.
Default font is **Arial**.
- [x] Font picker Control
- [x] Font family setting
- [x] Overlay window preview
- [x] System.Drawing implementation
- [x] DirectWrite implementation
- [x] Same font size units for Preview, System.Drawing and DirectWrite. | 6ed353c76ed14dccd6bb61ee296bd8d6b930da99 | b1c7e0054a8c4e2e596f0896ab5c71335a52a574 | diff --git a/src/Captura.Base/Images/IEditableFrame.cs b/src/Captura.Base/Images/IEditableFrame.cs
index fb99f0812..0285ed0ad 100644
--- a/src/Captura.Base/Images/IEditableFrame.cs
+++ b/src/Captura.Base/Images/IEditableFrame.cs
@@ -21,9 +21,11 @@ public interface IEditableFrame : IBitmapLoader
void DrawElli... | [
"src/Captura.Base/Images/IEditableFrame.cs",
"src/Captura.Base/Images/IFont.cs",
"src/Captura.Base/Images/RepeatFrame.cs",
"src/Captura.Base/Settings/TextOverlaySettings.cs",
"src/Captura.MouseKeyHook/MouseKeyHook.cs",
"src/Captura.Windows/Imaging/DrawingFont.cs",
"src/Captura.Windows/Imaging/GraphicsEd... | [] | true | ||
MathewSachin/Captura | 382 | issue_to_patch | New Crowdin translations | Enjoy! | 750101a38ddabc5b50e71affe0f9cb68a3bbde28 | aa02974dfe6453102e323954734fb2486b502288 | diff --git a/src/Captura.Loc/Languages/de.json b/src/Captura.Loc/Languages/de.json
index 25c3bea51..df7fea340 100644
--- a/src/Captura.Loc/Languages/de.json
+++ b/src/Captura.Loc/Languages/de.json
@@ -52,7 +52,7 @@
"Host": "Host",
"Hotkeys": "Hotkeys",
"ImageEditor": "Fotoeditor",
- "ImgEmpty": "Das Bild wurd... | [
"src/Captura.Loc/Languages/de.json",
"src/Captura.Loc/Languages/vi.json"
] | [] | true | ||
MathewSachin/Captura | 379 | issue_to_patch | New Crowdin translations | Enjoy! | 2144c42ffd445c9e3792eaa68fa2eb7eb5286fb6 | b501a2ca5ee30a48f8ce9ec675625c31575506c9 | diff --git a/src/Captura.Loc/Languages/de.json b/src/Captura.Loc/Languages/de.json
index 6c3b68e75..25c3bea51 100644
--- a/src/Captura.Loc/Languages/de.json
+++ b/src/Captura.Loc/Languages/de.json
@@ -3,7 +3,7 @@
"AccentColor": "Akzentfarbe",
"Add": "Hinzufügen",
"AlwaysOnTop": "Immer im Vordergrund",
- "Audi... | [
"src/Captura.Loc/Languages/de.json",
"src/Captura.Loc/Languages/it.json",
"src/Captura.Loc/Languages/pl.json",
"src/Captura.Loc/Languages/ru.json",
"src/Captura.Loc/Languages/zh-CN.json"
] | [] | true | ||
MathewSachin/Captura | 366 | issue_to_patch | New Crowdin translations | Enjoy! | e01755d70a647e7918f585a7bb8d1d6816c91a7a | e01559bfc8182ff36e739a0f0c88ece64d6ee221 | diff --git a/src/Captura.Loc/Languages/ar.json b/src/Captura.Loc/Languages/ar.json
index 67ec7d44e..5a874a133 100644
--- a/src/Captura.Loc/Languages/ar.json
+++ b/src/Captura.Loc/Languages/ar.json
@@ -72,8 +72,8 @@
"MaxRecent": "عدد العناصر المحفوظة الأقصى",
"MaxTextLength": "طول النص الأقصى",
"MicSource": "مص... | [
"src/Captura.Loc/Languages/ar.json",
"src/Captura.Loc/Languages/be.json",
"src/Captura.Loc/Languages/cs.json",
"src/Captura.Loc/Languages/da.json",
"src/Captura.Loc/Languages/de.json",
"src/Captura.Loc/Languages/es.json",
"src/Captura.Loc/Languages/fi.json",
"src/Captura.Loc/Languages/fr.json",
"src... | [] | true | ||
MathewSachin/Captura | 372 | issue_to_patch | Support %CAPTURA_PATH% | `%CAPTURA_PATH%` will be recognized as app directory for the following:
- [x] `--settings` command-line argument
- [x] Output folder path
- [x] FFmpeg folder path | a8845416f0ea2cb50b0afb1376eeb22d27d068d1 | e38569a59bbe88a87d820a36489a8523a386e21d | diff --git a/src/Captura.Base/Services/ServiceProvider.cs b/src/Captura.Base/Services/ServiceProvider.cs
index b845226ff..eafd06014 100644
--- a/src/Captura.Base/Services/ServiceProvider.cs
+++ b/src/Captura.Base/Services/ServiceProvider.cs
@@ -12,6 +12,8 @@ namespace Captura
{
public static class ServiceProvider... | [
"src/Captura.Base/Services/ServiceProvider.cs",
"src/Captura.Console/Program.cs",
"src/Captura.Core/Settings/Settings.cs",
"src/Captura.Core/ViewModels/MainModel.cs",
"src/Captura.FFmpeg/FFmpegService.cs",
"src/Captura.FFmpeg/FFmpgDownloadViewModel.cs",
"src/Captura.FFmpeg/Settings/FFMpegSettings.cs",
... | [] | diff --git a/src/Tests/Fixtures/AppRunnerFixture.cs b/src/Tests/Fixtures/AppRunnerFixture.cs
index ba8b3783f..68472a012 100644
--- a/src/Tests/Fixtures/AppRunnerFixture.cs
+++ b/src/Tests/Fixtures/AppRunnerFixture.cs
@@ -15,7 +15,7 @@ public AppRunnerFixture()
{
App = Application.Launch(new Proces... | true | |
MathewSachin/Captura | 371 | issue_to_patch | Webcam only mode | a8845416f0ea2cb50b0afb1376eeb22d27d068d1 | 310842687b3b094f0ad2746d5e0d32629c6460da | diff --git a/src/Captura.Core/CoreModule.cs b/src/Captura.Core/CoreModule.cs
index aa4398217..1a2d72c71 100644
--- a/src/Captura.Core/CoreModule.cs
+++ b/src/Captura.Core/CoreModule.cs
@@ -82,6 +82,7 @@ static void BindAudioSource(IBinder Binder)
static void BindVideoSourceProviders(IBinder Binder)
{
... | [
"src/Captura.Core/CoreModule.cs",
"src/Captura.Core/Models/WebcamSourceProvider.cs",
"src/Captura.Core/Models/WebcamVideoItem.cs",
"src/Captura.Core/ViewModels/RecordingModel.cs",
"src/Captura.Core/ViewModels/ScreenShotModel.cs",
"src/Captura.ViewCore/ViewModels/RecordingViewModel.cs",
"src/Captura.View... | [] | true | |||
MathewSachin/Captura | 360 | issue_to_patch | New Crowdin translations | Enjoy! | 413226b1cadf4ecd9d4724228bdca7f8979008d3 | f0cceaf1963904b731c592e753e59b3ed2cf0bda | diff --git a/src/Captura.Loc/Languages/ja.json b/src/Captura.Loc/Languages/ja.json
index 664e46a53..7dabe6d18 100644
--- a/src/Captura.Loc/Languages/ja.json
+++ b/src/Captura.Loc/Languages/ja.json
@@ -1,10 +1,10 @@
{
"About": "バージョン情報",
- "AccentColor": "",
- "Add": "",
+ "AccentColor": "アクセントカラー",
+ "Add": "追加... | [
"src/Captura.Loc/Languages/ja.json"
] | [] | true | ||
MathewSachin/Captura | 359 | issue_to_patch | New Crowdin translations | Enjoy! | e7297fd57ecbd84c40ec2a72d9da6734d06b1638 | 75f9392314d5cd03541ed2cf3cf0c31405eee334 | diff --git a/src/Captura.Loc/Languages/pt.json b/src/Captura.Loc/Languages/pt.json
index ecdd31cfa..6c9342267 100644
--- a/src/Captura.Loc/Languages/pt.json
+++ b/src/Captura.Loc/Languages/pt.json
@@ -1,6 +1,6 @@
{
"About": "Sobre",
- "AccentColor": "",
+ "AccentColor": "Cor de realces",
"Add": "Adicionar",
... | [
"src/Captura.Loc/Languages/pt.json"
] | [] | true | ||
MathewSachin/Captura | 313 | issue_to_patch | YouTube Upload | #### ToDo
- [ ] Show meaningful errors in a friendly way
- [x] Cancel upload with confirmation on exit (handle `Window.Closing` event)
- [x] Resume on error | c543c396a21ed30fc4f2ae302260b3f9f8696b71 | c76aa18b61a1d5057cfd06fd3e30f2e9109945c7 | diff --git a/appveyor.yml b/appveyor.yml
index adec3dcc5..fe56e5bd3 100644
--- a/appveyor.yml
+++ b/appveyor.yml
@@ -15,6 +15,10 @@ configuration:
environment:
imgur_client_id:
secure: Y/2WUSisk7oLSQNY1YzUxw==
+ yt_client_id:
+ secure: CPwUWOhTyZ2IiE+lvVilTKQz1h8sbn7FmkTJvtTlY6mJf2pDlfDhFmhDk1i47M9OPnD4y8a... | [
"appveyor.yml",
"src/Captura.Base/DelegateCommand.cs",
"src/Captura.Base/Services/IMainWindow.cs",
"src/Captura.Core/ApiKeys.cs",
"src/Captura.Core/Captura.Core.csproj",
"src/Captura.Core/CoreModule.cs",
"src/Captura.Core/Models/Recents/FileRecentItem.cs",
"src/Captura.Fakes/FakeWindowProvider.cs",
... | [] | true | ||
MemPalace/mempalace | 1,632 | issue_to_patch | feat: add Cursor IDE support (hooks, plugin, skill, docs, tests) | ## Summary
Adds first-class Cursor IDE support alongside the existing Claude Code and Codex flows: auto-save hook, pre-compaction transcript capture, session-start memory recall, plus a Cursor plugin (skill + slash commands + MCP auto-registration). Same privacy model as the existing Claude hooks — local-only, zero ex... | 2ec4baec3f0a8d19143422f26d2e6ee7735543a4 | 8e19e815cc30577fb37f953e5e828f6168a46cb2 | diff --git a/.claude-plugin/skills/mempalace-recall/SKILL.md b/.claude-plugin/skills/mempalace-recall/SKILL.md
new file mode 100644
index 000000000..749994f89
--- /dev/null
+++ b/.claude-plugin/skills/mempalace-recall/SKILL.md
@@ -0,0 +1,60 @@
+---
+name: mempalace-recall
+description: Recall protocol for MemPalace — s... | [
".claude-plugin/skills/mempalace-recall/SKILL.md",
".cursor-plugin/README.md",
".cursor-plugin/marketplace.json",
".cursor-plugin/mcp.json",
".cursor-plugin/plugin.json",
"CHANGELOG.md",
"README.md",
"commands/mempalace-help.md",
"commands/mempalace-init.md",
"commands/mempalace-mine.md",
"comma... | [
{
"comment": "\n\nUsing a heredoc inside a command substitution `$(...)` can trigger a known parser bug in stock macOS `bash` (version 3.2.57) when nested parentheses are present in the heredoc body. Since macOS is a primary platform for Cursor u... | diff --git a/tests/test_cursor_hooks_install.py b/tests/test_cursor_hooks_install.py
new file mode 100644
index 000000000..652d6558f
--- /dev/null
+++ b/tests/test_cursor_hooks_install.py
@@ -0,0 +1,469 @@
+"""Contract tests for ``hooks/cursor/install.sh``.
+
+The installer's job is to merge MemPalace hook entries into... | true | |
MemPalace/mempalace | 1,771 | issue_to_patch | feat(antigravity): add recall skill, optional rule, shared protocol | ## Summary
Adds the recall layer to the Antigravity plugin, mirroring the three-layer wiring from `feat/cursor-hooks-support`, adapted for Antigravity's native plugin and hook surfaces. Follows on from #1633.
- **`integrations/shared/recall-protocol.md`** — canonical "search before answering" protocol, single source ... | f124bd2c84b9d3790c041262e8840a52e17e451b | 60a4610c6b9dc45a15e3a0013aa011c47d7c4884 | diff --git a/.antigravity-plugin/README.md b/.antigravity-plugin/README.md
index e6fd75d0c..3110cbbc5 100644
--- a/.antigravity-plugin/README.md
+++ b/.antigravity-plugin/README.md
@@ -13,8 +13,12 @@ This directory is the source of truth for what gets installed at
├── mcp_config.json # auto-registers the mempal... | [
".antigravity-plugin/README.md",
".antigravity-plugin/rules/mempalace-recall.md",
".antigravity-plugin/skills/mempalace-recall/SKILL.md",
"hooks/antigravity/install.sh",
"integrations/shared/recall-protocol.md",
"tests/test_antigravity_plugin_manifest.py"
] | [
{
"comment": "\n\nThe protocol steps described here have already drifted from the canonical protocol defined in `integrations/shared/recall-protocol.md`. For example:\n\n- **Step 3**: The canonical protocol includes specific examples `(name, ... | diff --git a/tests/test_antigravity_plugin_manifest.py b/tests/test_antigravity_plugin_manifest.py
index 51fe72468..f3dc94c76 100644
--- a/tests/test_antigravity_plugin_manifest.py
+++ b/tests/test_antigravity_plugin_manifest.py
@@ -34,6 +34,15 @@
SKILL_MD = PLUGIN_DIR / "skills" / "mempalace" / "SKILL.md"
PLUGIN_REA... | true | |
MemPalace/mempalace | 1,679 | issue_to_patch | Add pluggable vector backends | ## Summary
This draft makes the normal MemPalace storage/search paths backend-neutral and adds three first-party non-Chroma backends for the next release:
- `sqlite_exact`: local SQLite + NumPy exact-vector backend with no new dependency
- `qdrant`: opt-in Qdrant REST backend, defaulting to localhost, with no new Pyt... | 15852edbafcec97887f4fa5b632bf3c8d85dfdc8 | 37e34f6cb5f0e96347fc5501d73a6b2728ab625f | diff --git a/README.md b/README.md
index 5259b1143..bd6be43d8 100644
--- a/README.md
+++ b/README.md
@@ -106,6 +106,42 @@ CUDA-accelerated embeddings, build the GPU variant with
`--gpus all`. Customise the bundled extras at build time, e.g.
`docker build --build-arg EXTRAS="extract,spellcheck" -t mempalace .`.
+## ... | [
"README.md",
"mempalace/backends/__init__.py",
"mempalace/backends/base.py",
"mempalace/backends/chroma.py",
"mempalace/backends/embedding_wrapper.py",
"mempalace/backends/pgvector.py",
"mempalace/backends/qdrant.py",
"mempalace/backends/registry.py",
"mempalace/backends/sqlite_exact.py",
"mempala... | [
{
"comment": "\n\nIn `lexical_search`, if the `text_filter` query is successful but returns zero results, `rows` will be empty (`[]`). Because of the `if not rows:` check, the method will fall back to scrolling the entire collection from Qdrant (... | diff --git a/tests/_backend_conformance.py b/tests/_backend_conformance.py
new file mode 100644
index 000000000..9f3a3ac22
--- /dev/null
+++ b/tests/_backend_conformance.py
@@ -0,0 +1,61 @@
+"""Shared backend isolation conformance assertions (RFC 001 isolation contract).
+
+Any backend's test module can import these to... | true | |
MemPalace/mempalace | 1,679 | comment_to_fix | Add pluggable vector backends | 
In `lexical_search`, if the `text_filter` query is successful but returns zero results, `rows` will be empty (`[]`). Because of the `if not rows:` check, the method will fall back to scrolling the entire collection from Qdrant (`self._scroll_all(qdrant... | 15852edbafcec97887f4fa5b632bf3c8d85dfdc8 | 37e34f6cb5f0e96347fc5501d73a6b2728ab625f | diff --git a/mempalace/backends/qdrant.py b/mempalace/backends/qdrant.py
new file mode 100644
index 000000000..f2d348eef
--- /dev/null
+++ b/mempalace/backends/qdrant.py
@@ -0,0 +1,1360 @@
+"""Qdrant REST backend for MemPalace.
+
+Qdrant is an opt-in external-service backend. Chroma remains the default; this
+adapter o... | [
"mempalace/backends/qdrant.py"
] | [
{
"comment": "\n\nIn `lexical_search`, if the `text_filter` query is successful but returns zero results, `rows` will be empty (`[]`). Because of the `if not rows:` check, the method will fall back to scrolling the entire collection from Qdrant (... | true | ||
MemPalace/mempalace | 1,679 | comment_to_fix | Add pluggable vector backends | The module docstring now claims "No API calls" and "local vector backend". That’s no longer true when the configured backend is `qdrant` (dedup will issue remote Qdrant REST calls via `get_collection()`/`query()`), so the docstring is misleading. | 15852edbafcec97887f4fa5b632bf3c8d85dfdc8 | 37e34f6cb5f0e96347fc5501d73a6b2728ab625f | diff --git a/mempalace/dedup.py b/mempalace/dedup.py
index 5e57aff95..82afa28a0 100644
--- a/mempalace/dedup.py
+++ b/mempalace/dedup.py
@@ -7,7 +7,9 @@
are too similar (cosine distance < threshold), keeps the longest/richest
version, and deletes the rest.
-No API calls — uses ChromaDB's built-in embedding similari... | [
"mempalace/dedup.py"
] | [
{
"comment": "The module docstring now claims \"No API calls\" and \"local vector backend\". That’s no longer true when the configured backend is `qdrant` (dedup will issue remote Qdrant REST calls via `get_collection()`/`query()`), so the docstring is misleading.",
"path": "mempalace/dedup.py",
"hunk":... | true | ||
MemPalace/mempalace | 1,633 | issue_to_patch | feat: add Antigravity IDE support (hooks, plugin, skill, docs, tests) | ## Summary
Adds first-class integration with Google's [Antigravity IDE](https://antigravity.google/) as a third sibling to the existing Claude Code and Codex hook integrations.
This is **strictly additive**: no existing files in `main` are restructured. The integration scope is bounded by what Google's [official Anti... | ad67c1a1211bc06a2a300885139e91c09d38b242 | df0e29ac8ba3f052921c4decd8d55c620cd107a7 | diff --git a/.antigravity-plugin/README.md b/.antigravity-plugin/README.md
new file mode 100644
index 000000000..e6fd75d0c
--- /dev/null
+++ b/.antigravity-plugin/README.md
@@ -0,0 +1,44 @@
+# MemPalace — Antigravity plugin
+
+In-repo packaging for the MemPalace integration with Google's [Antigravity IDE](https://antig... | [
".antigravity-plugin/README.md",
".antigravity-plugin/hooks.json.tmpl",
".antigravity-plugin/mcp_config.json",
".antigravity-plugin/plugin.json",
".antigravity-plugin/skills/mempalace/SKILL.md",
"CHANGELOG.md",
"README.md",
"examples/antigravity/README.md",
"examples/antigravity/hooks.json",
"exam... | [
{
"comment": "\n\nThis block contains two issues:\n1. **Sibling Process Wait Bug (Critical):** In bash, a background subshell `( wait \"$MINE_PID\" ... ) &` cannot wait for a sibling process spawned by the parent shell; `wait` will fail immediate... | diff --git a/tests/test_antigravity_hooks_install.py b/tests/test_antigravity_hooks_install.py
new file mode 100644
index 000000000..1735de768
--- /dev/null
+++ b/tests/test_antigravity_hooks_install.py
@@ -0,0 +1,324 @@
+"""End-to-end tests for the Antigravity install.sh.
+
+Covers:
+
+* `--dry-run` is fully side-effe... | true | |
MemPalace/mempalace | 1,737 | issue_to_patch | tool_update_drawer silently normalizes room casing when room arg omitted (e.g. 'ps5-enhanced-player' → 'PS5-enhanced-player'), splitting drawers across two rooms
## Summary
`mempalace_update_drawer` (and its underlying `tool_update_drawer`) silently normalizes the room name's casing when the `room` argument is omitte... | fix(mcp): preserve room/wing casing on case-only update | When `tool_update_drawer` receives a `room` or `wing` argument that differs from the stored value only in casing (e.g. `ps5-enhanced-player` vs `PS5-enhanced-player`), the server now preserves the original stored casing instead of overwriting it.
**Root cause:** LLM clients (Claude, Codex) routinely autocorrect acrony... | 4ceb880d19d46085e3bcd8aa963dc2e75dc8e6e5 | d514787f956264be1451eb437b86e954b95ad715 | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 734ac2863..4714e7e01 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -1696,14 +1696,22 @@ def tool_update_drawer(drawer_id: str, content: str = None, wing: str = None, ro
new_meta = dict(old_meta)
if wing is not... | [
"mempalace/mcp_server.py"
] | [
{
"comment": "\n\nIf `old_meta.get(\"wing\")` contains a non-string value (e.g., due to corrupted metadata or external writes), calling `.lower()` directly on it will raise an `AttributeError` and crash the tool call. Wrapping the retrieved m... | true | |
MemPalace/mempalace | 1,741 | issue_to_patch | feat: expose list_hallways / delete_hallway via the MCP server
**What problem does this solve?**
Hallways shipped in 3.3.6 (#1558) with Python API entry points `list_hallways(wing=None)` and `delete_hallway(...)` in `mempalace/hallways.py`, but neither is registered as an MCP tool in `mempalace/mcp_server.py`. Mining... | feat(mcp): expose list_hallways and delete_hallway tools | ## What does this PR do?
Closes #1739.
Wires the existing `hallways.list_hallways` and `hallways.delete_hallway` Python API into the MCP tool registry, mirroring the existing tunnel-tool pattern in `mcp_server.py`.
After 3.3.6 (#1558), hallways were being produced at mine time but were not retrievable through MCP — ... | 4ceb880d19d46085e3bcd8aa963dc2e75dc8e6e5 | ef2261feb8aaab44fcc4795647051a167bcfac59 | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 734ac2863..6c81d156d 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -88,6 +88,10 @@
delete_tunnel,
follow_tunnels,
)
+from .hallways import ( # noqa: E402
+ list_hallways,
+ delete_hallway,
+)
from .knowledg... | [
"mempalace/mcp_server.py",
"tests/test_mcp_server.py",
"website/reference/mcp-tools.md"
] | [
{
"comment": "\n\nIf `wing` is passed as a non-string (e.g., an integer or boolean) by an MCP client, `_sanitize_optional_name` will attempt to call `.strip()` on it, raising an unhandled `AttributeError`. This error will bubble up as an inte... | diff --git a/tests/test_mcp_server.py b/tests/test_mcp_server.py
index a5e406b0e..a24f496a5 100644
--- a/tests/test_mcp_server.py
+++ b/tests/test_mcp_server.py
@@ -1512,6 +1512,108 @@ def _raise(*args, **kwargs):
assert result == {"error": msg}
+ # ── hallway MCP tools (mirror the tunnel pattern) ──
+
... | true |
MemPalace/mempalace | 1,622 | issue_to_patch | `mempalace mine --limit N` truncates files BEFORE the already-mined skip — fire can finish with 0 new drawers
## Summary
`mempalace mine --limit N` truncates the scanned file list before the already-mined skip, so a fire can produce zero new drawers even when the corpus has unmined files within the time budget. `--li... | fix(miner): count only new work toward --limit, not already-mined skips (#1535) | ## What does this PR do?
Fixes #1535. `--limit N` now counts only files that produce new drawers, not files already mined.
Before this fix, `_mine_impl` (and `_mine_convos_impl`, `mine_formats`) truncated the file list via `files = files[:limit]` before the processing loop. Files that `file_already_mined` would skip ... | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | c9dc4c466d60402f5daa6538cf2b65225e142acc | diff --git a/mempalace/convo_miner.py b/mempalace/convo_miner.py
index a1b61aad1..ad802595d 100644
--- a/mempalace/convo_miner.py
+++ b/mempalace/convo_miner.py
@@ -594,15 +594,14 @@ def _mine_convos_impl(
wing = _resolve_wing(convo_path, wing)
files = scan_convos(convo_dir)
- if limit > 0:
- file... | [
"mempalace/convo_miner.py",
"mempalace/format_miner.py",
"mempalace/miner.py",
"tests/test_convo_miner.py",
"tests/test_format_miner.py",
"tests/test_miner.py"
] | [] | diff --git a/tests/test_convo_miner.py b/tests/test_convo_miner.py
index fd775ace6..22eefcdd0 100644
--- a/tests/test_convo_miner.py
+++ b/tests/test_convo_miner.py
@@ -416,3 +416,41 @@ def test_resolve_wing_empty_string_treated_as_no_wing(tmp_path):
target = tmp_path / ".gemini" / "tmp"
target.mkdir(parents=... | true |
MemPalace/mempalace | 1,692 | issue_to_patch | Feature request: add mempalace_mine tool to MCP server
## Problem
Mining is currently only available as a CLI command (`mempalace mine`). The MCP server does not expose a mine/index tool, which means any MCP client (Desktop Commander, LM Studio, Aionui, etc.) has no way to trigger indexing from within a conversation.... | feat(mcp): add mempalace_mine tool (#1662) | Closes #1662
## What does this PR do?
Adds a `mempalace_mine` MCP tool so MCP clients that cannot shell out (Claude Desktop, LM Studio, Aionui, Desktop Commander) can trigger indexing from within a conversation. Until now mining was reachable only through the `mempalace mine` CLI, so users on non-Claude-Code clients ... | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | 2a66bad49763d359fb7d55479dc9aa5e043fa6d2 | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 8187ccaf2..e5ecb1214 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -1535,6 +1535,213 @@ def tool_delete_drawer(drawer_id: str):
return {"success": False, "error": str(e)}
+def _capture_fd_stdout(fn):
+ """Run ``... | [
"mempalace/mcp_server.py",
"tests/test_mcp_mine.py",
"website/reference/mcp-tools.md"
] | [] | diff --git a/tests/test_mcp_mine.py b/tests/test_mcp_mine.py
new file mode 100644
index 000000000..0f73addf4
--- /dev/null
+++ b/tests/test_mcp_mine.py
@@ -0,0 +1,262 @@
+"""
+test_mcp_mine.py — Tests for the ``mempalace_mine`` MCP tool (#1662).
+
+Mining was previously CLI-only (``mempalace mine``); non-Claude-Code MC... | true |
MemPalace/mempalace | 1,697 | issue_to_patch | fix(backups): add max_backups retention to bound backup disk usage | ## Summary
`mempalace migrate` and `mempalace repair max-seq-id` each write a fresh, full-size, **timestamped** backup every run and never delete the old ones:
- `migrate` → `<palace>.pre-migrate.<timestamp>` (a full `copytree` of the palace)
- `repair max-seq-id` → `chroma.sqlite3.max-seq-id-backup-<timestamp>`... | 02b8753d9759e7c17eca58ba5162e5d2a39ca3d2 | 9be2b97b6ce13bdeebc2ee68b611622348a0063b | diff --git a/CHANGELOG.md b/CHANGELOG.md
index 64d66f2c9..a7fe9cbfd 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,14 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
---
+## [Unreleased]
+
+### Bug Fixes
+
+- **Backup retention to prevent unbounded disk usage.** `mempal... | [
"CHANGELOG.md",
"mempalace/backups.py",
"mempalace/config.py",
"mempalace/migrate.py",
"mempalace/repair.py",
"tests/test_backups.py",
"tests/test_config.py",
"tests/test_migrate.py",
"tests/test_repair.py",
"website/guide/configuration.md"
] | [] | diff --git a/tests/test_backups.py b/tests/test_backups.py
new file mode 100644
index 000000000..b11769c96
--- /dev/null
+++ b/tests/test_backups.py
@@ -0,0 +1,157 @@
+"""Tests for backup retention pruning (mempalace.backups.prune_backups).
+
+These guard the fix for unbounded backup growth: ``mempalace migrate`` and
+... | true | |
MemPalace/mempalace | 1,723 | issue_to_patch | fix(backends): keep post-deletion dim-None HNSW segments instead of quarantining (#1710) | ## What does this PR do?
`quarantine_invalid_hnsw_metadata` renames an HNSW segment to `*.corrupt-*` when its
`index_metadata.pickle` has `dimensionality: None` together with labels, unless
`_missing_dimensionality_appears_recoverable` judges the pickle safe to keep. develop
already keeps such a segment when its bookk... | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | 9a8c87acfe168cd3a45bfbddf6a280064f1b1eec | diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
index b547f806f..a567e8f37 100644
--- a/mempalace/backends/chroma.py
+++ b/mempalace/backends/chroma.py
@@ -925,7 +925,12 @@ def _missing_dimensionality_appears_recoverable(
return False
label_count = len(id_to_label)
- if int(t... | [
"mempalace/backends/chroma.py",
"tests/test_backends.py"
] | [] | diff --git a/tests/test_backends.py b/tests/test_backends.py
index f7845825e..8af4f2bba 100644
--- a/tests/test_backends.py
+++ b/tests/test_backends.py
@@ -1467,6 +1467,35 @@ def test_quarantine_invalid_hnsw_metadata_keeps_consistent_missing_dimensionalit
assert seg.exists()
+def test_quarantine_invalid_hnsw_... | true | |
MemPalace/mempalace | 1,734 | issue_to_patch | Qdrant embedder-identity persistence (fast-follow from #1724)
Fast-follow from the embedder-identity work (#1724, RFC 001).
#1724 implemented embedder-identity persistence + three-state enforcement for the three backends with a natural local metadata slot:
- `sqlite_exact` → `meta` table
- `pgvector` → marker JSON (p... | feat(backends): shared embedder-identity sidecar + qdrant identity (RFC 001) | ## Summary
Completes the embedder-identity contract for the last backend (**qdrant**) and unifies identity persistence across all local-path backends behind a shared sidecar.
## Design: a sidecar, not the marker
Identity is recorded in a small per-palace sidecar (`mempalace_embedder.json`), keyed by collection — **s... | 4ceb880d19d46085e3bcd8aa963dc2e75dc8e6e5 | ed77125ae3e7d58cd362e451e265f4b767bec150 | diff --git a/mempalace/backends/_sidecar.py b/mempalace/backends/_sidecar.py
new file mode 100644
index 000000000..a0485daf2
--- /dev/null
+++ b/mempalace/backends/_sidecar.py
@@ -0,0 +1,71 @@
+"""Shared embedder-identity sidecar (RFC 001).
+
+A small JSON file in the palace directory, keyed by collection name, recordi... | [
"mempalace/backends/_sidecar.py",
"mempalace/backends/chroma.py",
"mempalace/backends/pgvector.py",
"mempalace/backends/qdrant.py",
"tests/test_embedder_identity.py"
] | [
{
"comment": "`QdrantCollection.set_embedder_identity()` delegates to `_set_embedder_identity()`, but `_set_embedder_identity()` no-ops unless the marker file already exists. This means the brand-new/empty palace path in `_enforce_embedder_identity()` (create=True, count==0) cannot actually record identity for ... | diff --git a/tests/test_embedder_identity.py b/tests/test_embedder_identity.py
index 99dda3188..93ddae02f 100644
--- a/tests/test_embedder_identity.py
+++ b/tests/test_embedder_identity.py
@@ -119,16 +119,17 @@ def test_chroma_identity_roundtrip_via_sidecar(tmp_path):
assert os.path.isfile(os.path.join(str(tmp_pat... | true |
MemPalace/mempalace | 1,732 | issue_to_patch | RFC 001 §7.3: implement backend maintenance hooks (observable run_maintenance)
Tracked follow-up from **RFC 001 §7.3** ([#743](https://github.com/MemPalace/mempalace/pull/743)).
The maintenance hooks are specified but **not in `base.py` or the in-tree backends** (#1679 deferred them).
### What §7.3 requires
- `maint... | feat(backends): observable maintenance hooks + pgvector indexed-build path (RFC 001) | ## Summary
Implements the backend maintenance hooks from RFC 001 (#743) — the last of the deferred follow-ups (#1725) — and gives pgvector an **opt-in HNSW index path** with concurrency-safe builds. This closes the loop on the "is pgvector doing real vector search?" thread: it stays exact-by-default (100% recall), and... | 13de7a6743964f3991398367011f455a659f8f1e | fdbafe542045ff144d04e5948ba4df62e6055069 | diff --git a/mempalace/backends/__init__.py b/mempalace/backends/__init__.py
index 1560f2fbe..7432cbe2f 100644
--- a/mempalace/backends/__init__.py
+++ b/mempalace/backends/__init__.py
@@ -27,11 +27,13 @@
HealthStatus,
LexicalHit,
LexicalResult,
+ MaintenanceResult,
PalaceNotFoundError,
Pala... | [
"mempalace/backends/__init__.py",
"mempalace/backends/base.py",
"mempalace/backends/embedding_wrapper.py",
"mempalace/backends/pgvector.py",
"mempalace/backends/sqlite_exact.py",
"tests/test_maintenance_hooks.py"
] | [
{
"comment": "\n\n**Critical Bug**: If the table name is exactly 63 characters long (which is the exact length returned by `_pg_identifier` for any long table name), `_hnsw_index_name` will return the table name itself because `f\"{table}_hnsw_idx... | diff --git a/tests/test_maintenance_hooks.py b/tests/test_maintenance_hooks.py
new file mode 100644
index 000000000..95ac0fd8c
--- /dev/null
+++ b/tests/test_maintenance_hooks.py
@@ -0,0 +1,275 @@
+"""Backend maintenance hooks (RFC 001).
+
+Maintenance is observable, not fire-and-forget: ``run_maintenance(kind)``
+retu... | true |
MemPalace/mempalace | 1,731 | issue_to_patch | RFC 001 §1.5: implement Embedder protocol + identity enforcement
Tracked follow-up from **RFC 001 §1.5 / §5** ([#743](https://github.com/MemPalace/mempalace/pull/743)).
RFC 001 is accepted, but the embedder-identity contract is **not yet implemented** — the in-tree backends from #1679 defer it (`base.py` header notes... | feat(backends): embedder-identity contract + three-state enforcement (RFC 001) | ## Summary
Implements the embedder-identity contract from RFC 001 (#743) — the "hard dependency" tracked in #1724.
The explicit-embedding backends (`pgvector`, `qdrant`, `sqlite_exact`) embed through `EmbeddingCollection` and persisted **nothing** about which model produced their vectors. Swapping two same-dimension ... | 52144704e733b80e4f80d496f01e2b8c33ebaf06 | 212051da35cc38974fb4970bec0f733cf217b70c | diff --git a/mempalace/backends/base.py b/mempalace/backends/base.py
index 640c95fb4..f226f99b6 100644
--- a/mempalace/backends/base.py
+++ b/mempalace/backends/base.py
@@ -15,7 +15,7 @@
from abc import ABC, abstractmethod
from dataclasses import dataclass
-from typing import ClassVar, Optional
+from typing import ... | [
"mempalace/backends/base.py",
"mempalace/backends/chroma.py",
"mempalace/backends/embedding_wrapper.py",
"mempalace/backends/pgvector.py",
"mempalace/backends/sqlite_exact.py",
"mempalace/cli.py",
"mempalace/embedding.py",
"mempalace/palace.py",
"pyproject.toml",
"tests/test_embedder_identity.py"
... | [
{
"comment": "\n\nThere are two issues in this method:\n1. `json.load(f)` can return a non-dictionary type (like a list or string) if the JSON file is corrupted or malformed. If `data` is not a dictionary, calling `(data or {}).get(...)` will rai... | diff --git a/tests/test_embedder_identity.py b/tests/test_embedder_identity.py
new file mode 100644
index 000000000..99dda3188
--- /dev/null
+++ b/tests/test_embedder_identity.py
@@ -0,0 +1,310 @@
+"""Embedder-identity persistence and three-state enforcement (RFC 001).
+
+A same-dimension model swap (e.g. two 384-d mod... | true |
MemPalace/mempalace | 1,727 | issue_to_patch | RFC 001 §10: make searcher.py backend-neutral (metric-aware ranking + lexical fallback)
Tracked follow-up from **RFC 001 §10** ([#743](https://github.com/MemPalace/mempalace/pull/743)).
`searcher.py` is the highest-leakage Chroma coupling left after #1679's seam cleanup. Flagged by two backend authors (kostadis-ntnx ... | feat(searcher): metric-aware distance→similarity conversion (RFC 001 §10) | ## Summary
Makes the searcher's distance→similarity conversion **metric-aware** instead of hard-coding cosine, implementing the `searcher.py` item of RFC 001 §10 (#743).
The searcher hard-coded `max(0, 1 - distance)` for vector similarity in every ranking and `similarity`-reporting site. That formula is correct only ... | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | ffa8ac142cb54c803b812469e04ab78a1be3eb53 | diff --git a/mempalace/backends/base.py b/mempalace/backends/base.py
index 21713d6eb..640c95fb4 100644
--- a/mempalace/backends/base.py
+++ b/mempalace/backends/base.py
@@ -301,6 +301,17 @@ def close(self) -> None:
def health(self) -> HealthStatus:
return HealthStatus.healthy()
+ @property
+ def d... | [
"mempalace/backends/base.py",
"mempalace/backends/chroma.py",
"mempalace/backends/embedding_wrapper.py",
"mempalace/layers.py",
"mempalace/mcp_server.py",
"mempalace/searcher.py",
"tests/test_distance_metric.py",
"tests/test_searcher.py"
] | [
{
"comment": "`BaseCollection.distance_metric` currently always returns \"cosine\". That contradicts the docstring (“Defaults to the owning backend's declared metric”) and also breaks `EmbeddingCollection` delegation: since the property now exists on `BaseCollection`, wrapper collections no longer delegate `dis... | diff --git a/tests/test_distance_metric.py b/tests/test_distance_metric.py
new file mode 100644
index 000000000..c4334b0db
--- /dev/null
+++ b/tests/test_distance_metric.py
@@ -0,0 +1,227 @@
+"""Tests for backend-declared distance metrics (RFC 001) and the
+metric-aware distance→similarity conversion in the searcher.
+... | true |
MemPalace/mempalace | 1,706 | issue_to_patch | Release v3.4.0 — promote develop to main | Promotes `develop` (now at **3.4.0**) to `main` to cut the **v3.4.0** release. 83 commits since v3.3.6.
After this merges, the release flow is:
1. Draft **v3.4.0-rc1** pre-release on `main` → triggers `publish.yml` → approve the `pypi` env gate → `3.4.0rc1` on PyPI.
2. `pip install mempalace==3.4.0rc1` in a clean env ... | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
index 8e8872e32..8f4b0ba71 100644
--- a/.claude-plugin/marketplace.json
+++ b/.claude-plugin/marketplace.json
@@ -9,7 +9,7 @@
"name": "mempalace",
"source": "./.claude-plugin",
"description": "AI memory system — mine pro... | [
".claude-plugin/marketplace.json",
".claude-plugin/plugin.json",
".codex-plugin/plugin.json",
".dockerignore",
".github/workflows/docker-publish.yml",
".github/workflows/publish.yml",
"Dockerfile",
"Dockerfile.gpu",
"README.md",
"docker-compose.yml",
"docker-entrypoint.sh",
"docs/RELEASING.md"... | [
{
"comment": "\n\nIf `documents` is passed as a single string (which is supported by Chroma as `OneOrMany`), `list(documents)` will split the string into individual characters (e.g., `\"abc\"` -> `['a', 'b', 'c']`). This causes `_embed_texts` to ... | diff --git a/tests/_backend_conformance.py b/tests/_backend_conformance.py
new file mode 100644
index 000000000..9f3a3ac22
--- /dev/null
+++ b/tests/_backend_conformance.py
@@ -0,0 +1,61 @@
+"""Shared backend isolation conformance assertions (RFC 001 isolation contract).
+
+Any backend's test module can import these to... | true | |
MemPalace/mempalace | 1,706 | comment_to_fix | Release v3.4.0 — promote develop to main | 
If `documents` is passed as a single string (which is supported by Chroma as `OneOrMany`), `list(documents)` will split the string into individual characters (e.g., `"abc"` -> `['a', 'b', 'c']`). This causes `_embed_texts` to generate embeddings for ea... | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/mempalace/backends/embedding_wrapper.py b/mempalace/backends/embedding_wrapper.py
new file mode 100644
index 000000000..1441deb2d
--- /dev/null
+++ b/mempalace/backends/embedding_wrapper.py
@@ -0,0 +1,145 @@
+"""Core-side embedding adapter for explicit-vector backends."""
+
+from __future__ import annotati... | [
"mempalace/backends/embedding_wrapper.py"
] | [
{
"comment": "\n\nIf `documents` is passed as a single string (which is supported by Chroma as `OneOrMany`), `list(documents)` will split the string into individual characters (e.g., `\"abc\"` -> `['a', 'b', 'c']`). This causes `_embed_texts` to ... | true | ||
MemPalace/mempalace | 1,706 | comment_to_fix | Release v3.4.0 — promote develop to main | 
If `query_texts` is passed as a single string, `list(query_texts)` will split the string into individual characters. Ensure you wrap a single string in a list before passing it to `_embed_texts`.
```suggestion
if query_texts is not None and qu... | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/mempalace/backends/embedding_wrapper.py b/mempalace/backends/embedding_wrapper.py
new file mode 100644
index 000000000..1441deb2d
--- /dev/null
+++ b/mempalace/backends/embedding_wrapper.py
@@ -0,0 +1,145 @@
+"""Core-side embedding adapter for explicit-vector backends."""
+
+from __future__ import annotati... | [
"mempalace/backends/embedding_wrapper.py"
] | [
{
"comment": "\n\nIf `query_texts` is passed as a single string, `list(query_texts)` will split the string into individual characters. Ensure you wrap a single string in a list before passing it to `_embed_texts`.\n\n```suggestion\n if que... | true | ||
MemPalace/mempalace | 1,706 | comment_to_fix | Release v3.4.0 — promote develop to main | 
Qdrant's official API does not support `text_any` in the `match` object (only `text`, `any`, `value`, etc.). Passing `text_any` will cause a `400 Bad Request` on a real Qdrant server, forcing the backend to always fall back to the slow Python-side full... | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/mempalace/backends/qdrant.py b/mempalace/backends/qdrant.py
new file mode 100644
index 000000000..f2d348eef
--- /dev/null
+++ b/mempalace/backends/qdrant.py
@@ -0,0 +1,1360 @@
+"""Qdrant REST backend for MemPalace.
+
+Qdrant is an opt-in external-service backend. Chroma remains the default; this
+adapter o... | [
"mempalace/backends/qdrant.py"
] | [
{
"comment": "\n\nQdrant's official API does not support `text_any` in the `match` object (only `text`, `any`, `value`, etc.). Passing `text_any` will cause a `400 Bad Request` on a real Qdrant server, forcing the backend to always fall back to t... | true | ||
MemPalace/mempalace | 1,706 | comment_to_fix | Release v3.4.0 — promote develop to main | 
For backends that return a `GetResult` dataclass (which does not implement `__getitem__`), `hasattr(result, "__getitem__")` evaluates to `False`. This causes `existing_ids` and `existing_metas` to always be empty, silently bypassing the pre-mining coll... | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/mempalace/collision_scan.py b/mempalace/collision_scan.py
new file mode 100644
index 000000000..53819f2b6
--- /dev/null
+++ b/mempalace/collision_scan.py
@@ -0,0 +1,122 @@
+"""Pre-mining defense against drawer_id collisions.
+
+Runs immediately before a batched chromadb upsert. Computes the union of
+incom... | [
"mempalace/collision_scan.py"
] | [
{
"comment": "\n\nFor backends that return a `GetResult` dataclass (which does not implement `__getitem__`), `hasattr(result, \"__getitem__\")` evaluates to `False`. This causes `existing_ids` and `existing_metas` to always be empty, silently byp... | true | ||
MemPalace/mempalace | 1,706 | comment_to_fix | Release v3.4.0 — promote develop to main | 
`tempfile` is used on line 495 but is never imported in `_apply_topics_by_wing_renames` or at the module level, which will raise a `NameError` at runtime. Import `tempfile` alongside `json`.
```suggestion
import json
import tempfile
from ... | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/mempalace/migrate.py b/mempalace/migrate.py
index 9e809da22..36e18622d 100644
--- a/mempalace/migrate.py
+++ b/mempalace/migrate.py
@@ -360,3 +360,218 @@ def migrate(palace_path: str, dry_run: bool = False, confirm: bool = False):
print(f"\n{'=' * 60}\n")
return True
+
+
+# --------------------... | [
"mempalace/migrate.py"
] | [
{
"comment": "\n\n`tempfile` is used on line 495 but is never imported in `_apply_topics_by_wing_renames` or at the module level, which will raise a `NameError` at runtime. Import `tempfile` alongside `json`.\n\n```suggestion\n import json\n ... | true | ||
MemPalace/mempalace | 1,706 | comment_to_fix | Release v3.4.0 — promote develop to main | 
Calling `row.keys()` repeatedly inside a dictionary comprehension for every row in `meta_rows` introduces a significant performance bottleneck. Since `value_columns` are explicitly queried in the SELECT statement, they are guaranteed to exist in `r... | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
index 0730a7b74..b547f806f 100644
--- a/mempalace/backends/chroma.py
+++ b/mempalace/backends/chroma.py
@@ -2,10 +2,14 @@
import contextlib
import datetime as _dt
+import json
import logging
+import math
import os
import pickle
+import re
... | [
"mempalace/backends/chroma.py"
] | [
{
"comment": "\n\nCalling `row.keys()` repeatedly inside a dictionary comprehension for every row in `meta_rows` introduces a significant performance bottleneck. Since `value_columns` are explicitly queried in the SELECT statement, they are g... | true | ||
MemPalace/mempalace | 1,706 | comment_to_fix | Release v3.4.0 — promote develop to main | EmbeddingCollection computes embeddings with `list(documents)`. If a caller passes a single document as a bare `str` (ChromaDB's OneOrMany shape), this splits the string into characters, producing the wrong number of embeddings and causing downstream length-mismatch errors in explicit-vector backends (pgvector/sqlite_e... | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/mempalace/backends/embedding_wrapper.py b/mempalace/backends/embedding_wrapper.py
new file mode 100644
index 000000000..1441deb2d
--- /dev/null
+++ b/mempalace/backends/embedding_wrapper.py
@@ -0,0 +1,145 @@
+"""Core-side embedding adapter for explicit-vector backends."""
+
+from __future__ import annotati... | [
"mempalace/backends/embedding_wrapper.py"
] | [
{
"comment": "EmbeddingCollection computes embeddings with `list(documents)`. If a caller passes a single document as a bare `str` (ChromaDB's OneOrMany shape), this splits the string into characters, producing the wrong number of embeddings and causing downstream length-mismatch errors in explicit-vector backe... | true | ||
MemPalace/mempalace | 1,706 | comment_to_fix | Release v3.4.0 — promote develop to main | EmbeddingCollection.update has the same OneOrMany issue as add/upsert: if `documents` is provided as a bare `str`, `list(documents)` will embed per-character and the backend update will see a length mismatch vs `ids`. Normalize scalar `ids`/`documents`/`metadatas` to lists before computing embeddings and delegating. | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/mempalace/backends/embedding_wrapper.py b/mempalace/backends/embedding_wrapper.py
new file mode 100644
index 000000000..1441deb2d
--- /dev/null
+++ b/mempalace/backends/embedding_wrapper.py
@@ -0,0 +1,145 @@
+"""Core-side embedding adapter for explicit-vector backends."""
+
+from __future__ import annotati... | [
"mempalace/backends/embedding_wrapper.py"
] | [
{
"comment": "EmbeddingCollection.update has the same OneOrMany issue as add/upsert: if `documents` is provided as a bare `str`, `list(documents)` will embed per-character and the backend update will see a length mismatch vs `ids`. Normalize scalar `ids`/`documents`/`metadatas` to lists before computing embeddi... | true | ||
MemPalace/mempalace | 1,706 | comment_to_fix | Release v3.4.0 — promote develop to main | The comment claims `'|'` "cannot appear in source paths on any supported platform", but on POSIX filesystems `|` is a valid filename character. The delimiter choice may still be fine, but the rationale should be accurate so future contributors don't rely on a false guarantee. | db1fbe888b59514a66c43e745f095d762b9bf276 | 939a076baf0b349e1f5b3a7e27ad1d545364f18b | diff --git a/mempalace/ids.py b/mempalace/ids.py
new file mode 100644
index 000000000..ff20946f4
--- /dev/null
+++ b/mempalace/ids.py
@@ -0,0 +1,128 @@
+"""Centralized drawer/triple ID construction with collision-safe delimiter.
+
+Drawer IDs and content-addressed identifiers built by concatenating strings
+without a d... | [
"mempalace/ids.py"
] | [
{
"comment": "The comment claims `'|'` \"cannot appear in source paths on any supported platform\", but on POSIX filesystems `|` is a valid filename character. The delimiter choice may still be fine, but the rationale should be accurate so future contributors don't rely on a false guarantee.",
"path": "memp... | true | ||
MemPalace/mempalace | 1,709 | issue_to_patch | ci(publish): build + publish in one job (fix BlobNotFound artifact handoff) | ## Why
The v3.3.6 publish runs failed **twice**, both at the publish job's `download-artifact` step with `BlobNotFound`. The build job built `mempalace-3.3.6.tar.gz`+`.whl` and uploaded a valid 25 MB `dist` artifact (confirmed in the run's artifact list), but the **cross-job download** consistently failed — a GitHub v... | 72a69e24765c78a9599dc3db87f66e5baecb18d8 | 19d2f9216ddd3005944ef7f2d6616dee4d2924bb | diff --git a/.github/workflows/publish.yml b/.github/workflows/publish.yml
index d656615fd..7688ebe6d 100644
--- a/.github/workflows/publish.yml
+++ b/.github/workflows/publish.yml
@@ -10,8 +10,12 @@ name: Publish to PyPI
# so this lets a maintainer run the publish for any existing tag from the
# Actions tab ... | [
".github/workflows/publish.yml"
] | [] | true | ||
MemPalace/mempalace | 1,707 | issue_to_patch | fix(backends): wrap bare-str OneOrMany inputs before embedding | Fixes the one real correctness bug from the #1706 review (Gemini + Copilot, HIGH).
`EmbeddingCollection` did `_embed_texts(list(documents))`. ChromaDB's `OneOrMany` allows a single document as a bare `str` — and `list("abc")` → `['a','b','c']`, so each **character** got embedded, producing the wrong number of vectors ... | 9694e2cb572bbc15fb5f88e4a4c52125acb77fe3 | 051f0cc1ecb7075fcda2f9ce2429999cfbea3cab | diff --git a/mempalace/backends/embedding_wrapper.py b/mempalace/backends/embedding_wrapper.py
index f4a8ccda0..1441deb2d 100644
--- a/mempalace/backends/embedding_wrapper.py
+++ b/mempalace/backends/embedding_wrapper.py
@@ -18,6 +18,23 @@ def _embed_texts(texts: list[str]) -> list[list[float]]:
return [list(v) fo... | [
"mempalace/backends/embedding_wrapper.py",
"tests/test_embedding_wrapper.py"
] | [
{
"comment": "\n\nCurrently, `_as_list` only handles `str` by wrapping it in a list. However, ChromaDB's `OneOrMany` shape also allows a single dictionary for `metadatas` (e.g., `{\"source\": \"web\"}`). If a single dictionary is passed, calling ... | diff --git a/tests/test_embedding_wrapper.py b/tests/test_embedding_wrapper.py
new file mode 100644
index 000000000..5d592d6af
--- /dev/null
+++ b/tests/test_embedding_wrapper.py
@@ -0,0 +1,126 @@
+"""EmbeddingCollection OneOrMany handling (PR #1706 review).
+
+A bare ``str`` passed as ``documents``/``query_texts`` (Ch... | true | |
MemPalace/mempalace | 1,708 | issue_to_patch | ci(publish): add workflow_dispatch manual trigger | ## Why
Publishing the v3.3.6 release fired **no** `publish.yml` run — GitHub doesn't reliably emit a `release` event when a **draft tied to a pre-existing tag** is published (v3.3.6 was tagged ~2 weeks ago). Without a manual trigger there's no way to run the publish, and no escape hatch if a future release event also ... | 9694e2cb572bbc15fb5f88e4a4c52125acb77fe3 | d51e9434f9dfeab3377f6fe015826c2f3ee45010 | diff --git a/.github/workflows/publish.yml b/.github/workflows/publish.yml
index ca1813cc2..d656615fd 100644
--- a/.github/workflows/publish.yml
+++ b/.github/workflows/publish.yml
@@ -2,10 +2,16 @@ name: Publish to PyPI
# Publishes mempalace to PyPI via Trusted Publishing (OIDC — no stored token).
#
-# Trigger: a ... | [
".github/workflows/publish.yml"
] | [
{
"comment": "`actions/checkout` resolves an unqualified `ref` name as a branch before a tag if both exist. Using the bare tag name here can accidentally (or maliciously) check out a branch named like the release tag instead of the tag object, which undermines the intent of publishing exactly what was tagged.",... | true | ||
MemPalace/mempalace | 1,705 | issue_to_patch | chore(release): 3.4.0 | Bumps all five version manifests `3.3.6 -> 3.4.0` **plus the README version badge** (a sixth version location enforced by `test_readme_claims.py::TestVersionBadge`, outside version-guard's 5-file set — that's what failed #1704).
Replaces #1704, which was on a `release/3.4.0` branch that the "Protect release branches" ... | ab716d02a53e93b8afbafdb1c30de0a0fa6485e7 | 73d4e43b22ce5809d4e2206133d1bd1b3dda5c93 | diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
index 8e8872e32..8f4b0ba71 100644
--- a/.claude-plugin/marketplace.json
+++ b/.claude-plugin/marketplace.json
@@ -9,7 +9,7 @@
"name": "mempalace",
"source": "./.claude-plugin",
"description": "AI memory system — mine pro... | [
".claude-plugin/marketplace.json",
".claude-plugin/plugin.json",
".codex-plugin/plugin.json",
"README.md",
"mempalace/version.py",
"pyproject.toml"
] | [] | true | ||
MemPalace/mempalace | 1,675 | issue_to_patch | fix(config): strip leading/trailing separators in normalize_wing_name | ## Problem
MemPalace's own Claude Code save hook files conversations into wings that its own MCP write tools then reject.
`hooks/mempal_save_hook.sh` runs `mempalace mine "$(dirname "$TRANSCRIPT_PATH")" --mode convos` with no `--wing`, so the convo miner defaults the wing to `normalize_wing_name(dir.name)` (`convo_mi... | 708ef4a9fd3808c13fc57cb7ee8cebffda3989ba | ed772a01f9e159b246e7006d8ee51e5b0484d5f3 | diff --git a/mempalace/config.py b/mempalace/config.py
index 2cf83349d..36d7d82e9 100644
--- a/mempalace/config.py
+++ b/mempalace/config.py
@@ -37,8 +37,13 @@ def normalize_wing_name(name: str) -> str:
The same rule is applied by ``init`` when persisting `topics_by_wing`
and when writing `mempalace.yaml`, so... | [
"mempalace/config.py",
"tests/test_config.py",
"tests/test_miner.py"
] | [] | diff --git a/tests/test_config.py b/tests/test_config.py
index c125bf111..19847b6dd 100644
--- a/tests/test_config.py
+++ b/tests/test_config.py
@@ -196,6 +196,16 @@ def test_normalize_wing_name_mixed():
assert normalize_wing_name("My-Cool App") == "my_cool_app"
+def test_normalize_wing_name_strips_leading_sep... | true | |
MemPalace/mempalace | 1,702 | issue_to_patch | feat(migrate): mempalace migrate-wings — normalize legacy wing names (#1675 follow-up) | ## What
Adds `mempalace migrate-wings [--dry-run] [--yes]` — a one-time, idempotent migration that normalizes legacy wing names (strips leading/trailing separators) so palaces built before the wing-name rule (#1675) keep their memories discoverable under the new name.
The problem: a Claude-Code path-encoded dir like ... | 6ac2de8e041f3745dc89823bc277c7a0475efc21 | 8ec438ef577265716b453bc7f65a29b849f3e480 | diff --git a/docs/recovery/wing-name-migration.md b/docs/recovery/wing-name-migration.md
new file mode 100644
index 000000000..e63a34fa5
--- /dev/null
+++ b/docs/recovery/wing-name-migration.md
@@ -0,0 +1,63 @@
+# Recovery: legacy wing names split after the normalization change
+
+**Companion to #1675.** `normalize_win... | [
"docs/recovery/wing-name-migration.md",
"mempalace/cli.py",
"mempalace/migrate.py",
"tests/test_migrate_wings.py"
] | [
{
"comment": "\n\nIf the backend collection does not have metadata for any of the retrieved records, `metas` could be `None`. Zipping `ids` with `None` will raise a `TypeError`. Adding a defensive guard to default `metas` to a list of `None` ... | diff --git a/tests/test_migrate_wings.py b/tests/test_migrate_wings.py
new file mode 100644
index 000000000..229823ad9
--- /dev/null
+++ b/tests/test_migrate_wings.py
@@ -0,0 +1,153 @@
+"""Tests for the wing-name normalization migration (migrate-wings).
+
+normalize_wing_name strips leading/trailing separators (#1675);... | true | |
MemPalace/mempalace | 1,700 | issue_to_patch | fix(mcp): repair diary_write content alias + restore -32602 diagnostic (fixes red develop) | ## Why
#1245 merged into `develop` as a no-CI fork PR, and its first real CI (the develop merge) went **red**. This restores `develop` to green while keeping #1245's useful filtered-search recall fallback.
## Three defects in #1245
1. **Broke the `-32602` diagnostic.** `tool_diary_write` was given `entry: str = None... | 334fc6d0bef0161c5f0b643382d6af11143bd04a | b130301bd5ed51d38db5b23544a5874b009eb1d3 | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 4906ad8e2..8187ccaf2 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -1864,7 +1864,7 @@ def tool_kg_stats():
# ==================== AGENT DIARY ====================
-def tool_diary_write(agent_name: str, entry: str = None, t... | [
"mempalace/mcp_server.py",
"mempalace/searcher.py",
"tests/test_mcp_server.py"
] | [
{
"comment": "\n\nThe fallback query returns a `QueryResult` object (which uses attribute access) rather than a plain dictionary. Attempting to access its fields via dict-like subscription (e.g., `raw[\"documents\"]`) will raise a `TypeError` at ... | diff --git a/tests/test_mcp_server.py b/tests/test_mcp_server.py
index be34e616b..a5e406b0e 100644
--- a/tests/test_mcp_server.py
+++ b/tests/test_mcp_server.py
@@ -3323,6 +3323,88 @@ def test_two_missing_required_lists_both_names(self):
assert "'entry'" in message
assert " and " not in message.split(... | true | |
MemPalace/mempalace | 1,696 | issue_to_patch | feat(docker): add Docker image for MCP server and CLI | ## Description
Adds a Docker image to run MemPalace (MCP server or CLI) without a local python setup.
## What's included
- `Dockerfile` — CPU image (multi-stage, uv). Default extras: extract, spellcheck.
- `Dockerfile.gpu` — CUDA variant for GPU-accelerated embeddings.
- `docker-entrypoint.sh` — runs... | 02b8753d9759e7c17eca58ba5162e5d2a39ca3d2 | eb3ff2e61a72bce60c9fc17bf886fff495d128c9 | diff --git a/.dockerignore b/.dockerignore
new file mode 100644
index 000000000..544641ed0
--- /dev/null
+++ b/.dockerignore
@@ -0,0 +1,55 @@
+# Keep the build context lean — only what `uv sync` + the package need.
+
+# VCS
+.git/
+.gitignore
+
+# Python build artifacts / caches
+*.egg-info/
+dist/
+build/
+__pycache__... | [
".dockerignore",
".github/workflows/docker-publish.yml",
"Dockerfile",
"Dockerfile.gpu",
"README.md",
"docker-compose.yml",
"docker-entrypoint.sh"
] | [
{
"comment": "\n\nUsing a single-stage build for the GPU image leaves `build-essential` and other build-time dependencies in the final production image, which increases image size and introduces unnecessary security risks (compilers in produc... | true | ||
MemPalace/mempalace | 1,696 | comment_to_fix | feat(docker): add Docker image for MCP server and CLI | 
The documented `docker run` command for `mine` will fail inside the container because the host directory `/work` is not mounted. To make this command work, the host directory containing the project files must be mounted into the container (e.g., us... | 02b8753d9759e7c17eca58ba5162e5d2a39ca3d2 | eb3ff2e61a72bce60c9fc17bf886fff495d128c9 | diff --git a/README.md b/README.md
index 48208e4e9..5259b1143 100644
--- a/README.md
+++ b/README.md
@@ -69,6 +69,43 @@ python -m venv .venv && source .venv/bin/activate
pip install mempalace
```
+### Docker
+
+A container image is also available for running the MCP server or the CLI
+without a local Python toolcha... | [
"README.md"
] | [
{
"comment": "\n\nThe documented `docker run` command for `mine` will fail inside the container because the host directory `/work` is not mounted. To make this command work, the host directory containing the project files must be mounted into... | true | ||
MemPalace/mempalace | 1,673 | issue_to_patch | fix(backends): strip lone surrogates from documents at the ChromaDB chokepoint | ## Problem
A single lone UTF-16 surrogate (U+D800–U+DFFF) in a document raises
`UnicodeEncodeError` inside chromadb and aborts the **entire** add/upsert batch
with a `-32000` Internal Error — silently dropping every other row in the same
batch. For the sweeper (batches of 64 messages from Claude Code sessions, the
path... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | e54896699b9749102c3e7a8afec36e6f23f27dde | diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
index 426f21cb5..92d89b73a 100644
--- a/mempalace/backends/chroma.py
+++ b/mempalace/backends/chroma.py
@@ -1038,8 +1038,40 @@ def _sanitize_metadatas_for_chromadb(metadatas):
for m in metadatas
]
+ @staticmethod
+ de... | [
"mempalace/backends/chroma.py",
"tests/test_clean_lone_surrogates.py"
] | [
{
"comment": "\n\nIf `documents` is passed as a single string (which is natively supported by ChromaDB's API as `OneOrMany[Document]`), treating it as an iterable in the list comprehension will silently split the string into individual characters... | diff --git a/tests/test_clean_lone_surrogates.py b/tests/test_clean_lone_surrogates.py
index 60b6c1232..650f800bb 100644
--- a/tests/test_clean_lone_surrogates.py
+++ b/tests/test_clean_lone_surrogates.py
@@ -177,3 +177,88 @@ def test_diary_write(self, monkeypatch, collection, config, kg):
topic="log",
... | true | |
MemPalace/mempalace | 1,635 | issue_to_patch | MCP tool_search retry does not classify stale-index error variants as transient
**What happened?**
`mempalace_search` can surface stale-index transient errors that do not trigger the existing retry path in `mempalace.mcp_server.tool_search`.
Current retry classification in `_is_transient_index_error()` handles messa... | fix(mcp): retry stale-index transient in tool_search | ## What does this PR do?
Extends `tool_search` transient-index retry classification to also catch stale-index variants that were not previously retried.
- Updates `_is_transient_index_error()` in `mempalace/mcp_server.py` to treat `stale-index` / `stale index` (case-insensitive) as transient.
- Keeps existing retry b... | 6957c7e599747c744baaa49148df2b3aaacd5b50 | 2d4fae1ac32a1cd059889289c075ef033bd8a616 | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 5bfda3f20..d0f9e71ae 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -300,7 +300,15 @@ def _is_transient_index_error(result) -> bool:
if not isinstance(result, dict):
return False
err = result.get("error", "")
... | [
"mempalace/mcp_server.py",
"tests/test_mcp_server.py"
] | [] | diff --git a/tests/test_mcp_server.py b/tests/test_mcp_server.py
index 1870de55a..0b39f507a 100644
--- a/tests/test_mcp_server.py
+++ b/tests/test_mcp_server.py
@@ -1025,6 +1025,34 @@ def fake_search(*args, **kwargs):
assert "error" in result
assert "index_recovered" not in result
+ def test_sear... | true |
MemPalace/mempalace | 1,687 | issue_to_patch | Fix crash when tool_use input is a list instead of dict | ## Summary
- Guard `_format_tool_use` against `input` being a list instead of dict in Claude Code JSONL transcripts
- Prevents `AttributeError: 'list' object has no attribute 'get'` at `normalize.py:554`
## Problem
When mining Claude Code conversation transcripts (`--mode convos`), some `tool_use` content blocks hav... | 02b8753d9759e7c17eca58ba5162e5d2a39ca3d2 | 5ba8399bb27cd44144222eba34747bfb79dd368f | diff --git a/mempalace/normalize.py b/mempalace/normalize.py
index ca62cca3d..c34a22555 100644
--- a/mempalace/normalize.py
+++ b/mempalace/normalize.py
@@ -523,6 +523,8 @@ def _format_tool_use(block: dict) -> str:
"""Format a tool_use block into a human-readable one-liner."""
name = block.get("name", "Unknow... | [
"mempalace/normalize.py"
] | [
{
"comment": "\n\nInstead of only checking if `inp` is a `list`, it is safer to check if it is not a `dict`. This protects against other unexpected types (such as `None`, `str`, or other non-dictionary types) that would also cause an `Attribu... | true | ||
MemPalace/mempalace | 1,694 | issue_to_patch | Hook diary checkpoints are never visible to diary_read — wing derivation mismatch
## Environment
- OS: Windows 11 Pro
- MemPalace: 3.3.6 (plugin install)
- Claude Code: 2.1.161
## Problem
Stop hook diary checkpoints are saved to a **project-derived wing** (e.g. `wing_.claude`, `wing_portable_claude_paseo_kit`), but ... | fix(hooks): file stop-hook diary checkpoints under the harness agent identity (#1693) | ## What does this PR do?
Fixes #1693.
Stop-hook diary checkpoints were never visible to `mempalace_diary_read`. The Stop hook saves through `_save_diary_direct` (`mempalace/hooks_cli.py`), which called `tool_diary_write` with a hardcoded `agent_name="session-hook"`. `tool_diary_read` filters Chroma metadata by `agent... | 02b8753d9759e7c17eca58ba5162e5d2a39ca3d2 | 4e0fce010ecdca648f3300819e9200f4fdd7bae9 | diff --git a/mempalace/hooks_cli.py b/mempalace/hooks_cli.py
index 7275ef41c..3b86477e2 100644
--- a/mempalace/hooks_cli.py
+++ b/mempalace/hooks_cli.py
@@ -651,12 +651,16 @@ def _save_diary_direct(
session_id: str,
wing: str = "",
toast: bool = False,
+ *,
+ agent_name: str,
) -> dict:
"""Wr... | [
"mempalace/hooks_cli.py",
"tests/test_hooks_cli.py"
] | [] | diff --git a/tests/test_hooks_cli.py b/tests/test_hooks_cli.py
index 723238dfe..e09e2ba21 100644
--- a/tests/test_hooks_cli.py
+++ b/tests/test_hooks_cli.py
@@ -13,6 +13,7 @@
from mempalace.hooks_cli import (
SAVE_INTERVAL,
_count_human_messages,
+ _diary_agent_for_harness,
_extract_recent_messages,
... | true |
MemPalace/mempalace | 1,245 | issue_to_patch | fix: filtered search fallback + diary_write content alias | ## What this fixes
Two bugs found in production use with a ChromaDB palace of 1,200+ drawers ingested via mixed paths (bulk import + MCP tool calls).
### 1. Filtered search crashes on HNSW/SQLite index mismatch (`searcher.py`)
When the HNSW vector index is out of sync with the SQLite metadata store, any filtered sea... | cabc3afb6eb23b81b6d2f4a1858517daa428f994 | d12c96e43eea664e5bc873dcd2ca52a9e22011dc | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 43897c8ca..2b3eb621f 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -1095,7 +1095,7 @@ def tool_kg_stats():
# ==================== AGENT DIARY ====================
-def tool_diary_write(agent_name: str, entry: str, topic: s... | [
"mempalace/mcp_server.py",
"mempalace/searcher.py"
] | [] | true | ||
MemPalace/mempalace | 1,137 | issue_to_patch | fix(convo_miner): preserve blank lines and indentation in AI responses | ## What and Why
`_chunk_by_exchange` collapses every AI response during ingest:
- `.strip()` on each line destroys indentation (breaks code fences and nested lists)
- blank lines are skipped, fusing paragraphs
- `" ".join(ai_lines)` replaces every newline with a space
This violates the **verbatim-always** design prin... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 038a7e0e15267324285e52cfc9832f3550dd50eb | diff --git a/mempalace/convo_miner.py b/mempalace/convo_miner.py
index ee82e36c9..48221d3ff 100644
--- a/mempalace/convo_miner.py
+++ b/mempalace/convo_miner.py
@@ -192,11 +192,15 @@ def _chunk_by_exchange(lines: list, chunk_size: int, min_chunk_size: int) -> lis
next_line = lines[i]
i... | [
"mempalace/convo_miner.py",
"tests/test_convo_miner_unit.py"
] | [] | diff --git a/tests/test_convo_miner_unit.py b/tests/test_convo_miner_unit.py
index b14f8ab8f..38033d5de 100644
--- a/tests/test_convo_miner_unit.py
+++ b/tests/test_convo_miner_unit.py
@@ -186,6 +186,64 @@ def test_paragraph_loop_preserves_slice_order(self):
assert chunks[1]["content"] == "b" * CHUNK_SIZE
... | true | |
MemPalace/mempalace | 1,602 | issue_to_patch | HNSW quarantine never re-arms after in-place palace writes — _quarantined_paths gate skips mtime-change reconnects
## Summary
The HNSW segment quarantine (`quarantine_stale_hnsw` / `quarantine_invalid_hnsw_metadata`) runs once per palace per process, gated by the class-level set `ChromaBackend._quarantined_paths`. Af... | fix(backends): re-arm HNSW quarantine gate on mtime change and explicit reconnect (#1573) | ## What does this PR do?
Fixes #1573 -- `ChromaBackend._quarantined_paths` gate prevented HNSW quarantine from re-running after external in-place palace writes.
The gate was only cleared on inode change (full palace replacement). In-place writes from closet_llm, mine, or compress bump mtime without changing the inode... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 22804b8e08a05069e5aa256f7e1320dc7fb72fd6 | diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
index 426f21cb5..780490e73 100644
--- a/mempalace/backends/chroma.py
+++ b/mempalace/backends/chroma.py
@@ -1383,12 +1383,17 @@ def _client(self, palace_path: str):
)
if cached is None or inode_changed or mtime_changed or mtime... | [
"mempalace/backends/chroma.py",
"mempalace/mcp_server.py",
"tests/test_backends.py",
"tests/test_mcp_server.py"
] | [
{
"comment": "\n\nThe current logic triggers a `discard` on the first call of every `ChromaBackend` instance because `mtime_appeared` is true when `cached_mtime` is initialized to `0.0`. This causes redundant HNSW quarantine checks even if th... | diff --git a/tests/test_backends.py b/tests/test_backends.py
index 95ad69f1f..0125b75ca 100644
--- a/tests/test_backends.py
+++ b/tests/test_backends.py
@@ -1127,6 +1127,42 @@ def _spy(path, stale_seconds=300.0):
)
+def test_client_rearms_quarantine_on_mtime_change(tmp_path, monkeypatch):
+ """When the DB f... | true |
MemPalace/mempalace | 1,602 | comment_to_fix | fix(backends): re-arm HNSW quarantine gate on mtime change and explicit reconnect (#1573) | 
The current logic triggers a `discard` on the first call of every `ChromaBackend` instance because `mtime_appeared` is true when `cached_mtime` is initialized to `0.0`. This causes redundant HNSW quarantine checks even if they were already performe... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 22804b8e08a05069e5aa256f7e1320dc7fb72fd6 | diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
index 426f21cb5..780490e73 100644
--- a/mempalace/backends/chroma.py
+++ b/mempalace/backends/chroma.py
@@ -1383,12 +1383,17 @@ def _client(self, palace_path: str):
)
if cached is None or inode_changed or mtime_changed or mtime... | [
"mempalace/backends/chroma.py"
] | [
{
"comment": "\n\nThe current logic triggers a `discard` on the first call of every `ChromaBackend` instance because `mtime_appeared` is true when `cached_mtime` is initialized to `0.0`. This causes redundant HNSW quarantine checks even if th... | true | ||
MemPalace/mempalace | 1,618 | issue_to_patch | Mines smaller than hnsw:sync_threshold never persist index_metadata.pickle
**What happened?**
After mining any directory smaller than the `_HNSW_BLOAT_GUARD` `sync_threshold` (50,000 records), the very next palace open quarantines the segment as drifted:
```
Quarantined corrupt HNSW segment <palace>/<UUID> (sqlite N... | fix(backends): lower HNSW bloat-guard thresholds to fix sub-50k persist (#1579) | ## What does this PR do?
Lowers `_HNSW_BLOAT_GUARD` thresholds from 50,000 to 2 (both `hnsw:batch_size` and `hnsw:sync_threshold`).
chromadb's Rust HNSW segment only writes `index_metadata.pickle` and `link_lists.bin` when both internal counters cross their thresholds. With 50k/50k, any mine under 50,000 drawers (the... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 7b1f6247d063d6210045791b3215b52e1391ca1f | diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
index 426f21cb5..f93aaf19c 100644
--- a/mempalace/backends/chroma.py
+++ b/mempalace/backends/chroma.py
@@ -105,36 +105,38 @@ def _hnsw_payload_appears_sane(seg_dir: str) -> bool:
return ratio is None or ratio <= _HNSW_LINK_TO_DATA_MAX_RATIO
... | [
"mempalace/backends/chroma.py",
"tests/test_backends.py"
] | [
{
"comment": "\n\nFor better maintainability and consistency with other tests in this file (e.g., lines 746, 756, 771), please use the `_HNSW_MISSING_METADATA_DATA_FLOOR` constant instead of the magic number `1024`.\n\n```suggestion\n ... | diff --git a/tests/test_backends.py b/tests/test_backends.py
index 95ad69f1f..24c56a94d 100644
--- a/tests/test_backends.py
+++ b/tests/test_backends.py
@@ -30,6 +30,9 @@
quarantine_stale_hnsw,
)
+# embeddinggemma-300m Matryoshka truncation (first 384 of 768 dims).
+_TEST_EMBED_DIM = 384
+
class _FakeCollect... | true |
MemPalace/mempalace | 1,618 | comment_to_fix | fix(backends): lower HNSW bloat-guard thresholds to fix sub-50k persist (#1579) | 
For better maintainability and consistency with other tests in this file (e.g., lines 746, 756, 771), please use the `_HNSW_MISSING_METADATA_DATA_FLOOR` constant instead of the magic number `1024`.
```suggestion
if data.exists() and data.s... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 7b1f6247d063d6210045791b3215b52e1391ca1f | diff --git a/tests/test_backends.py b/tests/test_backends.py
index 95ad69f1f..24c56a94d 100644
--- a/tests/test_backends.py
+++ b/tests/test_backends.py
@@ -30,6 +30,9 @@
quarantine_stale_hnsw,
)
+# embeddinggemma-300m Matryoshka truncation (first 384 of 768 dims).
+_TEST_EMBED_DIM = 384
+
class _FakeCollect... | [
"tests/test_backends.py"
] | [
{
"comment": "\n\nFor better maintainability and consistency with other tests in this file (e.g., lines 746, 756, 771), please use the `_HNSW_MISSING_METADATA_DATA_FLOOR` constant instead of the magic number `1024`.\n\n```suggestion\n ... | true | ||
MemPalace/mempalace | 1,100 | issue_to_patch | [Windows] UnicodeEncodeError when running `mempalace mine` on GBK console
## Environment
- OS: Windows 11
- Python: 3.12
- mempalace version: 3.3.0
- Console: PowerShell/CMD with GBK encoding
## Problem
When running `mempalace mine`, the command fails with UnicodeEncodeError:
```
UnicodeEncodeError: 'gbk' codec can'... | fix(closet_llm): replace non-ASCII symbols in progress output (#1034) | ## What and Why
`closet_llm.py` prints `✓`, `✗`, and `—` (U+2713, U+2717, U+2014) during `mempalace closets regen`. These characters cannot be encoded in GBK (Windows PowerShell/CMD default), causing `UnicodeEncodeError` on the same code path fixed in `miner.py` via #681.
`cli.py` has the same problem on the `sweep` ... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 29db125b9b0eeffaebaf7b9aba76b1211ed9ab71 | diff --git a/mempalace/closet_llm.py b/mempalace/closet_llm.py
index a85d5178b..801f7119f 100644
--- a/mempalace/closet_llm.py
+++ b/mempalace/closet_llm.py
@@ -264,7 +264,7 @@ def regenerate_closets(
f"Regenerating closets for {len(sources)} source files via {cfg.endpoint} ({cfg.model})..."
)
if dry... | [
"mempalace/closet_llm.py"
] | [] | true | |
MemPalace/mempalace | 1,102 | issue_to_patch | [Windows] UnicodeEncodeError when running `mempalace mine` on GBK console
## Environment
- OS: Windows 11
- Python: 3.12
- mempalace version: 3.3.0
- Console: PowerShell/CMD with GBK encoding
## Problem
When running `mempalace mine`, the command fails with UnicodeEncodeError:
```
UnicodeEncodeError: 'gbk' codec can'... | fix(normalize): handle UTF-8 BOM in transcript files | ## What and Why
`normalize()` opens transcript files with `encoding='utf-8'`. When a file has a UTF-8 BOM prefix (`\xef\xbb\xbf`) — common from Windows exports of Claude Code JSONL sessions — `json.loads()` raises `JSONDecodeError` on every line because the first line starts with `{`. `_try_claude_code_jsonl` silentl... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 818e4f3b31f34d2b6d7cb6d9674fe5432ac505d6 | diff --git a/mempalace/normalize.py b/mempalace/normalize.py
index ca62cca3d..e3023e559 100644
--- a/mempalace/normalize.py
+++ b/mempalace/normalize.py
@@ -122,7 +122,7 @@ def normalize(filepath: str) -> str:
if file_size > 500 * 1024 * 1024: # 500 MB safety limit
raise IOError(f"File too large ({file_s... | [
"mempalace/normalize.py"
] | [] | true | |
MemPalace/mempalace | 1,682 | issue_to_patch | status cold-loads the HNSW vector index just to count drawers (~60s/call on large palaces)
## What happens
On a large palace (~398k drawers, ~3.5 GB `chroma.sqlite3`), `mempalace status` takes **~60s of CPU on every call**. `mempalace repair-status` returns the same drawer counts in **<1s**.
## Why
`miner.status()`... | fix(status): count drawers from sqlite instead of cold-loading the HNSW index | Closes #1681.
## What does this PR do?
`mempalace status` opened the ChromaDB collection purely to tally drawers by wing/room — and opening it cold-loads the ~586 MB HNSW vector index. On a 398k-drawer / 3.5 GB palace that costs **~60s of CPU on every call**, even though the counts live in `chroma.sqlite3`'s relation... | 02b8753d9759e7c17eca58ba5162e5d2a39ca3d2 | ce46747282b8196f0ecaa733d97ef6c477bef6ca | diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
index 426f21cb5..ed3d94ef0 100644
--- a/mempalace/backends/chroma.py
+++ b/mempalace/backends/chroma.py
@@ -7,6 +7,7 @@
import os
import pickle
import sqlite3
+from collections import defaultdict
from numbers import Integral
from pathlib impo... | [
"mempalace/backends/chroma.py",
"mempalace/miner.py",
"tests/test_miner.py"
] | [
{
"comment": "\n\nThe function's return type is annotated as `Optional[tuple[int, dict[str, dict[str, int]]]]`, but it currently returns a nested `defaultdict`. Returning a `defaultdict` can lead to unexpected side effects (such as silently m... | diff --git a/tests/test_miner.py b/tests/test_miner.py
index 93f0e80c0..b296b6e43 100644
--- a/tests/test_miner.py
+++ b/tests/test_miner.py
@@ -810,6 +810,133 @@ def get(self, *args, **kwargs):
assert "WING: proj" in out
+def test_status_does_not_cold_load_vector_index(palace_path, seeded_collection, capsys):... | true |
MemPalace/mempalace | 1,666 | issue_to_patch | fix(ids): delimit hash inputs to prevent drawer_id collisions (#80) | ## Summary
Closes the drawer_id collision defect class deferred from PR #1572 (Copilot finding #13). Six call sites built ID hash inputs by concatenating strings without a delimiter — under ChromaDB's primary-key constraint a hash collision silently overwrites the prior write, losing content with no error raised. Two ... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 206cfbb6ca9c660cd8d738b11c57ea7be7d8ba7a | diff --git a/mempalace/collision_scan.py b/mempalace/collision_scan.py
new file mode 100644
index 000000000..53819f2b6
--- /dev/null
+++ b/mempalace/collision_scan.py
@@ -0,0 +1,122 @@
+"""Pre-mining defense against drawer_id collisions.
+
+Runs immediately before a batched chromadb upsert. Computes the union of
+incom... | [
"mempalace/collision_scan.py",
"mempalace/convo_miner.py",
"mempalace/format_miner.py",
"mempalace/ids.py",
"mempalace/knowledge_graph.py",
"mempalace/mcp_server.py",
"mempalace/miner.py",
"tests/test_collision_scan.py",
"tests/test_convo_miner_unit.py",
"tests/test_ids.py"
] | [
{
"comment": "\n\nIf `result` is a dict-like object but `result[\"metadatas\"]` or `result[\"ids\"]` is `None` (which can happen in ChromaDB under certain conditions or with custom backends), zipping them directly on line 94 will raise a `TypeErr... | diff --git a/tests/test_collision_scan.py b/tests/test_collision_scan.py
new file mode 100644
index 000000000..3ac315d2a
--- /dev/null
+++ b/tests/test_collision_scan.py
@@ -0,0 +1,228 @@
+"""Tests for mempalace.collision_scan — pre-mining defense against
+drawer_id collisions.
+
+The scan runs immediately before batch... | true | |
MemPalace/mempalace | 1,678 | issue_to_patch | Bug: MCP server module-level mkdir defeats _palace_root_exists() kill switch
## Summary
The kill switch added in #1305 (`hooks_cli._palace_root_exists()`) is silently
defeated by a module-level side effect in `mcp_server.py`. Importing the MCP
server creates `~/.mempalace/wal/` unconditionally — including when the us... | fix(mcp): defer WAL setup so import no longer recreates ~/.mempalace (#1676) | ## What does this PR do?
Fixes #1676. Importing `mempalace.mcp_server` recreated `~/.mempalace`, which defeated the kill-switch added in #1305.
The write-ahead-log setup ran at module scope:
```python
_WAL_DIR = Path(os.path.expanduser("~/.mempalace/wal"))
_WAL_DIR.mkdir(parents=True, exist_ok=True) # creates ~/.m... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 0dd0ade3b9256812fe703ed14231dbba8fc4604a | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 33dafd219..1ff993861 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -387,21 +387,51 @@ def _refresh_vector_disabled_flag() -> None:
# This provides an audit trail for detecting memory poisoning and
# enables review/rollback of... | [
"mempalace/mcp_server.py",
"tests/test_mcp_server.py"
] | [
{
"comment": "\n\nCalling `mkdir(parents=True, exist_ok=True)` on every single write operation introduces unnecessary system call overhead, especially since the directory will almost always exist after the first write.\n\nWe can optimize this... | diff --git a/tests/test_mcp_server.py b/tests/test_mcp_server.py
index 1870de55a..9232403f1 100644
--- a/tests/test_mcp_server.py
+++ b/tests/test_mcp_server.py
@@ -2459,6 +2459,66 @@ def always_fails():
assert col is None
+class TestImportKillSwitchSafety:
+ """Importing mcp_server must not recreate ~/... | true |
MemPalace/mempalace | 1,678 | comment_to_fix | fix(mcp): defer WAL setup so import no longer recreates ~/.mempalace (#1676) | 
Calling `mkdir(parents=True, exist_ok=True)` on every single write operation introduces unnecessary system call overhead, especially since the directory will almost always exist after the first write.
We can optimize this by using an EAFP (Easier ... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 0dd0ade3b9256812fe703ed14231dbba8fc4604a | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 33dafd219..1ff993861 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -387,21 +387,51 @@ def _refresh_vector_disabled_flag() -> None:
# This provides an audit trail for detecting memory poisoning and
# enables review/rollback of... | [
"mempalace/mcp_server.py"
] | [
{
"comment": "\n\nCalling `mkdir(parents=True, exist_ok=True)` on every single write operation introduces unnecessary system call overhead, especially since the directory will almost always exist after the first write.\n\nWe can optimize this... | true | ||
MemPalace/mempalace | 1,678 | comment_to_fix | fix(mcp): defer WAL setup so import no longer recreates ~/.mempalace (#1676) | 
Calling `chmod` on every single write operation introduces an unnecessary I/O system call overhead on the hot path of logging. We can optimize this by caching the initialized directory path in a module-level variable. This avoids redundant `chmod` ... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 0dd0ade3b9256812fe703ed14231dbba8fc4604a | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 33dafd219..1ff993861 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -387,21 +387,51 @@ def _refresh_vector_disabled_flag() -> None:
# This provides an audit trail for detecting memory poisoning and
# enables review/rollback of... | [
"mempalace/mcp_server.py"
] | [
{
"comment": "\n\nCalling `chmod` on every single write operation introduces an unnecessary I/O system call overhead on the hot path of logging. We can optimize this by caching the initialized directory path in a module-level variable. This a... | true | ||
MemPalace/mempalace | 1,678 | comment_to_fix | fix(mcp): defer WAL setup so import no longer recreates ~/.mempalace (#1676) | 
If `chmod` fails with a persistent `OSError` (such as `PermissionError` or `Operation not permitted` on restricted filesystems, network shares, or certain Docker environments), `_WAL_INITIALIZED_DIR` is never set. Consequently, every subsequent wri... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 0dd0ade3b9256812fe703ed14231dbba8fc4604a | diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
index 33dafd219..1ff993861 100644
--- a/mempalace/mcp_server.py
+++ b/mempalace/mcp_server.py
@@ -387,21 +387,51 @@ def _refresh_vector_disabled_flag() -> None:
# This provides an audit trail for detecting memory poisoning and
# enables review/rollback of... | [
"mempalace/mcp_server.py"
] | [
{
"comment": "\n\nIf `chmod` fails with a persistent `OSError` (such as `PermissionError` or `Operation not permitted` on restricted filesystems, network shares, or certain Docker environments), `_WAL_INITIALIZED_DIR` is never set. Consequent... | true | ||
MemPalace/mempalace | 1,631 | issue_to_patch | fix(embedding): add embed_query/embed_documents to EmbeddinggemmaONNX for ChromaDB 1.5.x | ## Bug
ChromaDB 1.5.x calls `embedding_function.embed_query(input=...)` via keyword argument during `collection.query()`. `EmbeddinggemmaONNX` lacked both `embed_query` and `embed_documents` methods, causing:
```
TypeError: embed_query() got an unexpected keyword argument input
```
whenever semantic search was trigg... | 6957c7e599747c744baaa49148df2b3aaacd5b50 | 48f4eb6dfe5a5830579eb65dda770c0db442cad2 | diff --git a/mempalace/embedding.py b/mempalace/embedding.py
index b22b1f52b..85bf1b928 100644
--- a/mempalace/embedding.py
+++ b/mempalace/embedding.py
@@ -191,6 +191,9 @@ def _lazy_load(self) -> None:
model_path = hf_hub_download(
_EMBEDDINGGEMMA_REPO, subfolder="onnx", filename=_EMBEDDINGGEMMA_... | [
"mempalace/embedding.py",
"tests/test_embeddinggemma.py"
] | [
{
"comment": "\n\nIf `embed_query` is called with a single string (which is the standard signature for `embed_query` in LangChain and other common integrations), `input` will be a `str` instead of a `list[str]`. \n\nSince `EmbeddinggemmaONNX.__ca... | diff --git a/tests/test_embeddinggemma.py b/tests/test_embeddinggemma.py
index 3100fce7a..f108ff969 100644
--- a/tests/test_embeddinggemma.py
+++ b/tests/test_embeddinggemma.py
@@ -126,7 +126,7 @@ def test_lazy_load_runs_once(patched_lazy_load):
ef(["one"])
ef(["two"])
ef(["three"])
- assert patched_l... | true | |
MemPalace/mempalace | 1,698 | issue_to_patch | ci: add PyPI trusted-publishing workflow | ## What
Adds automated PyPI publishing via **Trusted Publishing (OIDC)** — no API token stored anywhere. Releases publish when a **GitHub Release is published**, gated by a manual approval on the `pypi` environment.
## How it works
- **`build` job** (read-only token): verifies the release tag's commit is reachable f... | 02b8753d9759e7c17eca58ba5162e5d2a39ca3d2 | dc7d0be10c6da79415d493ba983768b02268ca00 | diff --git a/.github/workflows/publish.yml b/.github/workflows/publish.yml
new file mode 100644
index 000000000..ca1813cc2
--- /dev/null
+++ b/.github/workflows/publish.yml
@@ -0,0 +1,102 @@
+name: Publish to PyPI
+
+# Publishes mempalace to PyPI via Trusted Publishing (OIDC — no stored token).
+#
+# Trigger: a GitHub ... | [
".github/workflows/publish.yml",
"docs/RELEASING.md"
] | [
{
"comment": "\n\nCommitting directly to `main` is often blocked by branch protection rules. Additionally, doing so causes `develop` to drift behind `main` unless a back-merge is performed.\n\nIt is cleaner and safer to bump the version on `d... | true | ||
MemPalace/mempalace | 1,680 | issue_to_patch | compute_hallways_for_wing crashes (too many SQL variables) on wings > ~32k drawers
## Summary
The within-wing hallways post-mine step added in #1558 fetches the entire wing in a single `col.get(where={"wing": wing})`. ChromaDB binds one SQL variable per matching id, so on a wing larger than SQLite's `SQLITE_MAX_VARIAB... | fix(hallways): paginate drawer fetch to avoid SQLite variable overflow on large wings (#1619) | compute_hallways_for_wing fetched the whole wing in a single
col.get(where={"wing": wing}). ChromaDB binds one SQL variable per matched id,
so on a wing larger than SQLITE_MAX_VARIABLE_NUMBER (32766) the call raised
"too many SQL variables" inside chromadb. The exception was caught, so the mine
completed — but the wing... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 300e824e94a01700d5c4296ad5967bf3b0e3fdc7 | diff --git a/mempalace/hallways.py b/mempalace/hallways.py
index 907527a07..9bee6c60b 100644
--- a/mempalace/hallways.py
+++ b/mempalace/hallways.py
@@ -179,7 +179,12 @@ def compute_hallways_for_wing(
Args:
wing: wing name to scan.
- col: ChromaDB collection — must support ``.get(where=..., inclu... | [
"mempalace/hallways.py",
"tests/test_hallways.py",
"tests/test_hallways_pagination.py"
] | [
{
"comment": "\n\nInstead of fetching the entire collection and filtering by wing client-side (which is highly inefficient and can cause significant performance/memory overhead on large databases), you can paginate with the `where` filter directl... | diff --git a/tests/test_hallways.py b/tests/test_hallways.py
index 94b7ed26c..92ba1865a 100644
--- a/tests/test_hallways.py
+++ b/tests/test_hallways.py
@@ -27,11 +27,21 @@ def _use_tmp_hallway_file(monkeypatch, tmp_path):
def _fake_collection(drawers):
- """Build a MagicMock collection whose .get() returns the... | true |
MemPalace/mempalace | 1,680 | comment_to_fix | fix(hallways): paginate drawer fetch to avoid SQLite variable overflow on large wings (#1619) | 
Instead of fetching the entire collection and filtering by wing client-side (which is highly inefficient and can cause significant performance/memory overhead on large databases), you can paginate with the `where` filter directly in the `col.get` query... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 300e824e94a01700d5c4296ad5967bf3b0e3fdc7 | diff --git a/mempalace/hallways.py b/mempalace/hallways.py
index 907527a07..9bee6c60b 100644
--- a/mempalace/hallways.py
+++ b/mempalace/hallways.py
@@ -179,7 +179,12 @@ def compute_hallways_for_wing(
Args:
wing: wing name to scan.
- col: ChromaDB collection — must support ``.get(where=..., inclu... | [
"mempalace/hallways.py"
] | [
{
"comment": "\n\nInstead of fetching the entire collection and filtering by wing client-side (which is highly inefficient and can cause significant performance/memory overhead on large databases), you can paginate with the `where` filter directl... | true | ||
MemPalace/mempalace | 1,680 | comment_to_fix | fix(hallways): paginate drawer fetch to avoid SQLite variable overflow on large wings (#1619) | 
Update the mock collection's `_get` implementation to support filtering by `where` to accurately reflect the paginated wing-filtered query.
```suggestion
def _get(limit=None, offset=0, include=None, where=None, ids=None, **kwargs):
fil... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 300e824e94a01700d5c4296ad5967bf3b0e3fdc7 | diff --git a/tests/test_hallways.py b/tests/test_hallways.py
index 94b7ed26c..92ba1865a 100644
--- a/tests/test_hallways.py
+++ b/tests/test_hallways.py
@@ -27,11 +27,21 @@ def _use_tmp_hallway_file(monkeypatch, tmp_path):
def _fake_collection(drawers):
- """Build a MagicMock collection whose .get() returns the... | [
"tests/test_hallways.py"
] | [
{
"comment": "\n\nUpdate the mock collection's `_get` implementation to support filtering by `where` to accurately reflect the paginated wing-filtered query.\n\n```suggestion\n def _get(limit=None, offset=0, include=None, where=None, ids=N... | true | ||
MemPalace/mempalace | 1,680 | comment_to_fix | fix(hallways): paginate drawer fetch to avoid SQLite variable overflow on large wings (#1619) | 
Update the mock to only raise an error when `where` is queried without a `limit` (which is when the SQLite variable overflow actually occurs in ChromaDB), and support filtering by `where` when `limit` is provided.
```suggestion
def _get(limit=... | 9b7cfc9940846f9bb06bdfe40476ca34b504ab4b | 300e824e94a01700d5c4296ad5967bf3b0e3fdc7 | diff --git a/tests/test_hallways_pagination.py b/tests/test_hallways_pagination.py
new file mode 100644
index 000000000..8847071ae
--- /dev/null
+++ b/tests/test_hallways_pagination.py
@@ -0,0 +1,56 @@
+"""Regression test for #1619.
+
+``compute_hallways_for_wing`` must fetch drawers by paginating
+(``count()`` + ``get... | [
"tests/test_hallways_pagination.py"
] | [
{
"comment": "\n\nUpdate the mock to only raise an error when `where` is queried without a `limit` (which is when the SQLite variable overflow actually occurs in ChromaDB), and support filtering by `where` when `limit` is provided.\n\n```sugg... | true | ||
MemPalace/mempalace | 1,572 | issue_to_patch | fix(extract): polish PR — address bot review feedback on PR #1555 | Polish PR addressing the inline bot review feedback (Copilot + gemini-code-assist) on PR #1555 that wasn't load-bearing enough to block the original ship.
## Scope: 12 of 13 bot polish items
One item — **Copilot #13 (drawer ID delimiter collision)** — is intentionally **deferred** to its own dedicated PR because ... | 95caf80fe34fb3a572e82b66ce5d127b557108e9 | b6dc122acb3ea653cd783775bea01de4eed02535 | diff --git a/docs/format-coverage.md b/docs/format-coverage.md
index 178658e09..8b166a1f1 100644
--- a/docs/format-coverage.md
+++ b/docs/format-coverage.md
@@ -1,9 +1,9 @@
# Format coverage — `mempalace mine --mode extract`
-**Proposed for mempalace 3.3.6.**
-**Target file (new):** `mempalace/format_miner.py`
-**Ta... | [
"docs/format-coverage.md",
"mempalace/format_miner.py",
"mempalace/searcher.py",
"tests/test_format_miner.py",
"tests/test_line_numbers.py"
] | [
{
"comment": "\n\nThis summary call will be skipped if an exception occurs in the outer loop because the `except` blocks currently swallow the exception but the code execution proceeds past the `finally` block only if no re-raising occurs. If... | diff --git a/tests/test_format_miner.py b/tests/test_format_miner.py
index 4c4e28c13..a9ed81682 100644
--- a/tests/test_format_miner.py
+++ b/tests/test_format_miner.py
@@ -1,14 +1,12 @@
-"""Tests for format_miner (proposed for mempalace 3.3.6).
+"""Tests for format_miner (mempalace 3.3.6, integrated in PR #1555).
-C... | true | |
MemPalace/mempalace | 1,216 | issue_to_patch | fix(migrate): close SQLite connection and clean temp palace on exception | ## What and Why
Two resource leaks in `mempalace.migrate` that bite hardest on Windows:
### 1. SQLite connection leak in `extract_drawers_from_sqlite()`
`mempalace/migrate.py:55` opened a connection at the top of the function and only closed it on the happy path. Any exception during query/iteration leaked the handl... | 1247e175b518aedafbc31f6bc201e67bae72b82a | cb1303694d9caa52bbc21ffa4e87bd294d9576a9 | diff --git a/mempalace/migrate.py b/mempalace/migrate.py
index 5b74591b3..9e809da22 100644
--- a/mempalace/migrate.py
+++ b/mempalace/migrate.py
@@ -22,8 +22,10 @@
import os
import shutil
import sqlite3
+import tempfile
import uuid
from collections import defaultdict
+from contextlib import closing
from datetime ... | [
"mempalace/migrate.py",
"tests/test_migrate.py"
] | [] | diff --git a/tests/test_migrate.py b/tests/test_migrate.py
index ba12ff528..5d9255a9f 100644
--- a/tests/test_migrate.py
+++ b/tests/test_migrate.py
@@ -1,10 +1,16 @@
"""Tests for destructive-operation safety in mempalace.migrate."""
import os
+import sqlite3
from types import SimpleNamespace
from unittest.mock i... | true | |
MemPalace/mempalace | 1,617 | issue_to_patch | _fix_blob_seq_ids does not repair collection config_json_str missing _type field (chromadb 1.5.9+ regression)
## Summary
When `_fix_blob_seq_ids` (mempalace/backends/chroma.py:816) runs on a palace whose `collections.config_json_str` was written by older chromadb (lacks `_type` field), the migration repairs the BLOB ... | fix(backends): repair missing _type in collection config (#1611) | ## What does this PR do?
chromadb <= 1.5.8 writes `config_json_str = '{}'` (empty JSON) when creating
collections. chromadb 1.5.9 introduced a strict `_type` check in the
collection config deserialization path -- its absence raises
`KeyError: '_type'` on palace open. Since the pin allows `>=1.5.4,<2`,
any `uv tool upg... | d0d011eb33c25bba87fb530047e5ce2066fd6ba7 | c4d8804b448eaac29fbffa0c2d12e5fa02d2cf6e | diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
index 0730a7b74..426f21cb5 100644
--- a/mempalace/backends/chroma.py
+++ b/mempalace/backends/chroma.py
@@ -2,6 +2,7 @@
import contextlib
import datetime as _dt
+import json
import logging
import os
import pickle
@@ -665,6 +666,7 @@ def _pi... | [
"mempalace/backends/chroma.py",
"tests/test_backends.py"
] | [
{
"comment": "\n\nThe `sqlite3` connection is not explicitly closed. In Python, using a `sqlite3.Connection` as a context manager only handles transaction management (committing on success or rolling back on exception); it does not close the conn... | diff --git a/tests/test_backends.py b/tests/test_backends.py
index e0c3b2b5d..95ad69f1f 100644
--- a/tests/test_backends.py
+++ b/tests/test_backends.py
@@ -23,6 +23,7 @@
ChromaCollection,
_HNSW_MISSING_METADATA_DATA_FLOOR,
_fix_blob_seq_ids,
+ _fix_missing_collection_type,
_pin_hnsw_threads,
... | true |
MemPalace/mempalace | 1,617 | comment_to_fix | fix(backends): repair missing _type in collection config (#1611) | 
The `sqlite3` connection is not explicitly closed. In Python, using a `sqlite3.Connection` as a context manager only handles transaction management (committing on success or rolling back on exception); it does not close the connection. Since this migra... | d0d011eb33c25bba87fb530047e5ce2066fd6ba7 | c4d8804b448eaac29fbffa0c2d12e5fa02d2cf6e | diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
index 0730a7b74..426f21cb5 100644
--- a/mempalace/backends/chroma.py
+++ b/mempalace/backends/chroma.py
@@ -2,6 +2,7 @@
import contextlib
import datetime as _dt
+import json
import logging
import os
import pickle
@@ -665,6 +666,7 @@ def _pi... | [
"mempalace/backends/chroma.py"
] | [
{
"comment": "\n\nThe `sqlite3` connection is not explicitly closed. In Python, using a `sqlite3.Connection` as a context manager only handles transaction management (committing on success or rolling back on exception); it does not close the conn... | true | ||
MemPalace/mempalace | 1,647 | issue_to_patch | build(deps-dev): bump ruff from 0.15.14 to 0.15.15 | > [!WARNING]
> Dependabot will stop supporting `python v3.9`!
>
> Please upgrade to one of the following versions: `v3.9`, `v3.10`, `v3.11`, `v3.12`, `v3.13`, or `v3.14`.
>
Bumps [ruff](https://github.com/astral-sh/ruff) from 0.15.14 to 0.15.15.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="... | 6957c7e599747c744baaa49148df2b3aaacd5b50 | a0508cb06586a27d119296edd06865d42b6df153 | diff --git a/pyproject.toml b/pyproject.toml
index fdf326b2e..1828a8453 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -71,7 +71,7 @@ chroma = "mempalace.backends.chroma:ChromaBackend"
dev = [
"pytest>=7.0",
"pytest-cov>=4.0",
- "ruff==0.15.14",
+ "ruff==0.15.15",
"psutil>=5.9",
# Proper... | [
"pyproject.toml"
] | [] | true | ||
MemPalace/mempalace | 1,652 | issue_to_patch | file_already_mined non-deterministic under additive mining when multiple parent_drawer_id groups share a source_file
### Summary
After PR #1593's additive-mining changes, re-mining a source file no longer destroys prior drawers — each mining pass writes its own `parent_drawer_id` group. As a side effect, a single `so... | fix(palace): file_already_mined iterates all groups when source_file has multiple parent_drawer_id mining passes | ## Summary
Under the additive-mining model (drawer history preserved across re-mines), a single `source_file` can have multiple `parent_drawer_id` groups in the palace — one per mining pass — each with its own stored `source_mtime` and `normalize_version`.
`file_already_mined` (the function the project miner uses to ... | 6957c7e599747c744baaa49148df2b3aaacd5b50 | 13c38ac588c9d9296f09c400c228a70bbce66f64 | diff --git a/mempalace/palace.py b/mempalace/palace.py
index 8ffa88fbc..128019ac1 100644
--- a/mempalace/palace.py
+++ b/mempalace/palace.py
@@ -741,47 +741,51 @@ def file_already_mined(
treated as exchange-mode drawers.
"""
try:
- stored_meta = None
- if extract_mode is None:
- ... | [
"mempalace/palace.py",
"tests/test_miner.py"
] | [] | diff --git a/tests/test_miner.py b/tests/test_miner.py
index 1baadd14e..93f0e80c0 100644
--- a/tests/test_miner.py
+++ b/tests/test_miner.py
@@ -1976,3 +1976,79 @@ def test_two_digit_year_00_is_2000(self, tmp_path):
content = "Y2K reference: 25/01/00.\n"
f.write_text(content)
assert _extract_... | true |
MemPalace/mempalace | 1,615 | issue_to_patch | chore: sync main back to develop after v3.3.6 release | ## Summary
Closes the release cycle for v3.3.6 by bringing the three main-only commits back to develop:
- \`62a555c\` fix(release): align ruff pin to 0.15.14 + hoist COCA imports out of hot paths (#1614)
- \`a1cf052\` Merge pull request #1614 from MemPalace/fix/release-3.3.6-ruff-pin-align
- \`db1fbe8\` Merge pull re... | d0d011eb33c25bba87fb530047e5ce2066fd6ba7 | 753df201ac1f7eb443d04c4b3913cb25a1c79e76 | diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 5aee4e34e..671e1da54 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -53,6 +53,6 @@ jobs:
# Keep this pin identical to the `ruff==` pin in pyproject.toml
# ([project.optional-dependencies].dev and [dependency-gr... | [
".github/workflows/ci.yml"
] | [] | true | ||
MemPalace/mempalace | 1,610 | issue_to_patch | Release v3.3.6 | ## Summary
Promotes `develop` to `main` for the v3.3.6 release. Mirrors the v3.3.5 release flow.
Once merged, an annotated `v3.3.6` tag will be pushed to the merge commit on `main`, followed by a `chore/sync-3.3.6-back-to-develop` PR to fast-forward develop with any main-side fixes.
### Contents
- 224 commits ahead... | d0163a7bec5ee6faa5e86169ed407a7bf41d5582 | a1cf0529151dedb32527f142f86c08b027b81bc1 | diff --git a/.claude-plugin/hooks/hooks.json b/.claude-plugin/hooks/hooks.json
index b80a78504..9960beda9 100644
--- a/.claude-plugin/hooks/hooks.json
+++ b/.claude-plugin/hooks/hooks.json
@@ -6,7 +6,8 @@
"hooks": [
{
"type": "command",
- "command": "bash \"${CLAUDE_PLUGIN_RO... | [
".claude-plugin/hooks/hooks.json",
".claude-plugin/marketplace.json",
".claude-plugin/plugin.json",
".codex-plugin/plugin.json",
".github/workflows/ci.yml",
".github/workflows/deploy-docs.yml",
".gitignore",
".pre-commit-config.yaml",
"CHANGELOG.md",
"CONTRIBUTING.md",
"README.md",
"benchmarks... | [
{
"comment": "The ruff version pinned in this workflow (0.15.9) does not match the repo’s other pins (pyproject.toml and .pre-commit-config.yaml both pin 0.15.14). This can cause CI to fail `ruff format --check` even when contributors format locally. Align the workflow pin with the single canonical ruff version... | diff --git a/tests/conftest.py b/tests/conftest.py
index 9a3b02d4a..bd653632c 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -33,6 +33,23 @@
from mempalace.config import MempalaceConfig # noqa: E402
from mempalace.knowledge_graph import KnowledgeGraph # noqa: E402
+# Redirect ChromaDB's ONNX model cach... | true | |
MemPalace/mempalace | 1,610 | comment_to_fix | Release v3.3.6 | The ruff version pinned in this workflow (0.15.9) does not match the repo’s other pins (pyproject.toml and .pre-commit-config.yaml both pin 0.15.14). This can cause CI to fail `ruff format --check` even when contributors format locally. Align the workflow pin with the single canonical ruff version used elsewhere (or up... | d0163a7bec5ee6faa5e86169ed407a7bf41d5582 | a1cf0529151dedb32527f142f86c08b027b81bc1 | diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 7c6bcaad9..671e1da54 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -50,6 +50,9 @@ jobs:
with:
python-version: "3.11"
cache: 'pip'
- - run: pip install "ruff>=0.4.0,<0.5"
+ # Keep thi... | [
".github/workflows/ci.yml"
] | [
{
"comment": "The ruff version pinned in this workflow (0.15.9) does not match the repo’s other pins (pyproject.toml and .pre-commit-config.yaml both pin 0.15.14). This can cause CI to fail `ruff format --check` even when contributors format locally. Align the workflow pin with the single canonical ruff version... | true | ||
MemPalace/mempalace | 1,613 | issue_to_patch | feat(entity): known-systems lexicon keeps multi-word product names atomic | Adds a curated list of multi-word product/system names ("Claude Code", "GitHub Copilot", "Visual Studio Code", "GPT-4", …) and a compound pre-pass that detects them atomically before the existing single-word extraction runs. Without this, the regex-based detector decomposes "Claude Code" into "Claude" + "Code" — and th... | c9d8a40a370d9112b13fd2eda4912f7d1c110344 | dbbf4c5af5c99e9e3da45132f08b63d72268bccf | diff --git a/mempalace/data/known_systems.json b/mempalace/data/known_systems.json
new file mode 100644
index 000000000..6cdd480ac
--- /dev/null
+++ b/mempalace/data/known_systems.json
@@ -0,0 +1,68 @@
+{
+ "schema_version": 1,
+ "description": "Multi-word product/system names that must be detected atomically — not d... | [
"mempalace/data/known_systems.json",
"mempalace/entity_detector.py",
"mempalace/miner.py",
"mempalace/palace.py",
"tests/test_entity_detector.py"
] | [
{
"comment": "\n\nThe current implementation re-compiles regex patterns for every compound on every call to `_apply_known_systems_prepass`. Since the lexicon is static and cached, it is much more efficient to pre-compile these patterns once i... | diff --git a/tests/test_entity_detector.py b/tests/test_entity_detector.py
index 11278f939..cc7483122 100644
--- a/tests/test_entity_detector.py
+++ b/tests/test_entity_detector.py
@@ -166,6 +166,148 @@ def test_coca_wordlist_contains_all_known_aya_false_positives():
assert not missing, f"COCA wordlist missing kno... | true | |
MemPalace/mempalace | 1,612 | issue_to_patch | perf(miner,palace): hoist COCA filter imports out of per-drawer hot paths | The COCA content-word filter shipped in PR #1605 imported `_get_coca_filter` and `_candidate_entity_words` locally inside two hot paths:
- `palace.build_closet_lines` — runs per source file during mine
- `miner._extract_entities_for_metadata` — runs per drawer during mine
Both imports are now at module top, ... | f5ea021cecdd8308816861368ebda12e1ce53b5b | 5fe54df4c8cfb8cc317e350661e52ce64094e212 | diff --git a/mempalace/miner.py b/mempalace/miner.py
index db5f3894c..2937e660b 100644
--- a/mempalace/miner.py
+++ b/mempalace/miner.py
@@ -19,10 +19,12 @@
from collections import defaultdict
from typing import Optional
+from .entity_detector import _get_coca_filter
from .palace import (
NORMALIZE_VERSION,
... | [
"mempalace/miner.py",
"mempalace/palace.py"
] | [] | true | ||
MemPalace/mempalace | 1,608 | issue_to_patch | docs(readme): move scam/retention alerts below header, soften tone | ## Summary
- The CAUTION (scam alert) and IMPORTANT (Claude Code retention) admonitions sat above the logo + title, so the first content on the repo page was loud red warnings rather than the project introduction. Moves both below the header.
- Softens the scam block: drops the `# 🚨 CRITICAL SECURITY WARNING: BEWARE ... | df36eb301bb1eda1e43b9d69732463d988d1105a | f51ecc162163589ac7ad081f45059506b84d7730 | diff --git a/README.md b/README.md
index 9b7ce1171..f069ed438 100644
--- a/README.md
+++ b/README.md
@@ -1,21 +1,3 @@
-> [!CAUTION]
-> # 🚨 CRITICAL SECURITY WARNING: BEWARE OF SCAMS
-> **MemPalace has NO other official websites.**
->
-> The **ONLY** official sources are:
-> 1. This **[GitHub repository](https://githu... | [
"README.md"
] | [
{
"comment": "The phrasing “MemPalace has no other official websites” is potentially confusing since the paragraph immediately lists an official docs website (mempalaceofficial.com). Consider using “domains” (or similar) to avoid the apparent contradiction while keeping the meaning the same.\n",
"path": "RE... | true | ||
MemPalace/mempalace | 1,614 | issue_to_patch | fix(release): align ruff pin to 0.15.14 + hoist COCA imports out of hot paths | The COCA content-word filter shipped in PR https://github.com/MemPalace/mempalace/pull/1605 imported _get_coca_filter and _candidate_entity_words locally inside two hot paths:
palace.build_closet_lines — runs per source file during mine
miner._extract_entities_for_metadata — runs per drawer during mine
Both import... | f5ea021cecdd8308816861368ebda12e1ce53b5b | 62a555cd80ade346bc415302d8c7f8cedfc33b5a | diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 5aee4e34e..671e1da54 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -53,6 +53,6 @@ jobs:
# Keep this pin identical to the `ruff==` pin in pyproject.toml
# ([project.optional-dependencies].dev and [dependency-gr... | [
".github/workflows/ci.yml",
"mempalace/miner.py",
"mempalace/palace.py"
] | [
{
"comment": "\n\nHoisting `_get_coca_filter` to the module level is a good improvement for performance in hot paths and adheres to pylint's C0415 rule. However, please ensure that this does not introduce a circular dependency between `miner.... | true | ||
MiniMax-AI/skills | 75 | issue_to_patch | fix: fix some problems | fixed some problems | e61fdd53925614a4db33ba6394fdc3a93aa053fc | 62870bece6d63c2e2fbd92f6c03ae26c4ee5cf8f | diff --git a/skills/minimax-music-playlist/SKILL.md b/skills/minimax-music-playlist/SKILL.md
index 49f2a07f..61bf62fe 100644
--- a/skills/minimax-music-playlist/SKILL.md
+++ b/skills/minimax-music-playlist/SKILL.md
@@ -14,7 +14,7 @@ metadata:
# MiniMax Music Playlist — Personalized Playlist Generator
-Scan the use... | [
"skills/minimax-music-playlist/SKILL.md"
] | [] | true |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.