My Wiki!

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
programming:0_ide:kilocode_vscode [2026/05/12 11:20] – [Agent Memory] tdprogramming:0_ide:kilocode_vscode [2026/05/12 22:07] (current) – [memvid mcp for vscode] td
Line 18: Line 18:
 ## Agent Memory ## Agent Memory
   * https://developers.openai.com/cookbook/examples/agents_sdk/building_reliable_agents_memory_compaction   * https://developers.openai.com/cookbook/examples/agents_sdk/building_reliable_agents_memory_compaction
 +  * https://voxos.ai/blog/how-to-give-ai-coding-agents-long-term-m/index.html
   * https://github.com/Kilo-Org/kilocode/discussions/2022   * https://github.com/Kilo-Org/kilocode/discussions/2022
   * https://github.com/Kilo-Org/kilocode/discussions/377   * https://github.com/Kilo-Org/kilocode/discussions/377
 +
 +
 +Folder and Manifest Best Practices
 +
 +Put source documents, manifests, helper files, and output directories in the Manifest instead of pasting large content into the prompt.
 +
 +Put longer task instructions in workspace files such as README.md, task.md, or AGENTS.md; keep agent instructions focused on behavior and boundaries.
 +
 +Use stable document IDs and a machine-readable manifest file so generated memos can cite sources and reviewers can inspect the path back to evidence.
 +
 +Let Memory() manage its own memory artifacts. By default, sandbox memory uses memories/ and sessions/ under the workspace.
 +
 +Keep generated artifacts under outputs/ so the application can inspect, copy, validate, or archive them after the run.
 +
 +Keep mount scopes narrow. If you mount a data room, mount only what the agent should read or write.
 +
 +Treat secrets as runtime configuration injected by your application or sandbox provider, not as prompt text or committed manifest content.
 +
 +Prefer a small synthetic File(...) or Dir(...) entry for a tutorial, then switch to LocalDir, GitRepo, or storage mounts for production-sized datasets
 +.
 +### This is a visual preview of the sandbox workspace structure
 +.
 +### The next cell builds the actual Manifest entries manually.
 +```
 +WORKSPACE_TREE = """
 +/workspace/
 +  README.md
 +  manifest.csv
 +  docs/
 +    batch_1/
 +    batch_2/
 +    batch_3/
 +  outputs/
 +  memories/    # Generated by Memory()
 +  sessions/    # Generated by Memory()
 +""".strip()
 +
 +print(WORKSPACE_TREE)
 +
 +```
 +
 +Prepare a Small Evidence Workspace
 +### memvid mcp for vscode
 +
 +#### Install npm, uv on windows
 +  * https://wiki.visrc.com/doku.php?id=windows:winadmin:npm_user
 +
 +
 +#### Setup memvid mcp server
 +
 +  * https://github.com/khgs2411/memvid_mcp
 +
 +
 +Running local mpx server may not work due to not available PATH for non admin user. 
 +
 +Install memvid-mcp-server locally
 +
 +If you have Node.js installed but npx isn't in PATH, install the package globally or locally:
 +
 +     npm install -g memvid-mcp-server
 +
 +Then update the ~/.config/kilo.json (~/.config/kilo.jsonc for global settings) configuration to use the direct path:
 +
 +```
 +{
 +  "command": ["memvid-mcp-server"],
 +  ...
 +}
 +
 +```
 +
 +Example:
 +
 +```
 +"$schema": "https://app.kilo.ai/config.json",
 +  "mcp": {
 +    "memvid": {
 +      "type": "local",
 +      "command": [
 +        "npx",
 +        "-y",
 +        "memvid-mcp-server@latest"
 +      ],
 +      "environment": {
 +        "OPENAI_API_KEY": "sk-proj-qoGHEBwSHb41x8slxaziGjq1RSp5Sz3FT3BlbkFJYfp15vPobwEm0e-Ib0zdRhOTntpQsGf2qSY8kA3FR6T4np3tMlrc79yjyXLWtcpEd1BaE-IbkA",
 +        "MEMVID_LOCAL_STORAGE": "0"
 +      },
 +      
 +```
 +
 +```
 + "$schema": "https://app.kilo.ai/config.json",
 +  "mcp": {
 +    "memvid": {
 +      "type": "local",
 +      "command": ["memvid-mcp-server"],
 +      "environment": {
 +        "OPENAI_API_KEY": "sk-proj-NEZJMF3Yljd-lfp15vPobwEm0e-Ib0zdRhOTntpQsGf2qSY8kA3FR6T4np3tMlrc79yjyXLWtcpEd1BaE-IbkA",
 +        "MEMVID_LOCAL_STORAGE": "0"
 +      },
 +```
 +
 +Note: If you do not provide OPENAI_API_KEY, the ask_memory tool will not be available, but you can still use create_or_open_memory, add_content, and search_memory (lexical mode).
 +
 +MEMVID_LOCAL_STORAGE Set to "1" to store memory files in ./memvid_mcp (current directory) instead of ~/.memvid_mcp.
 +
 +
 +**This only work in external powershell not in VS Code sandbox. See troubleshooting**
 +If PATH only defined in non admin user $profile, use this command
 +
 +     ["powershell", "-Command", ". $PROFILE; npx -y memvid-mcp-server@latest"]
 +     
 +#### Troubleshooting
 +
 +VS Code use sandbox environment where non admin user $profile is not available.
 +
 +##### Set PATH
 +     powershell -Command "[Environment]::SetEnvironmentVariable('Path', 'D:\devfs\python;D:\devfs\node-v24.15.0-win-x64;' + [Environment]::GetEnvironmentVariable('Path', 'User'), 'User')" 2>&1; echo "PATH_UPDATED:$?"
 +     
 +Edit PATH as non admin user:
 +
 +     rundll32 sysdm.cpl,EditEnvironmentVariables
 +
 +Add full PATH:
 +
 +     D:\devfs\python;D:\devfs\node-v24.15.0-win-x64;D:\devfs\node-v24.15.0-win-x64;D:\devfs\node-v24.15.0-win-x64;C:\Users\13223\.local\bin;C:\Users\13223\AppData\Local\Microsoft\WindowsApps;C:\Users\13223\AppData\Local\Programs\Git\cmd;C:\Users\13223\AppData\Local\Programs\Microsoft VS Code Insiders\bin;C:\Users\13223\AppData\Local\Programs\DockerDesktop\resources\bin;D:\devfs\jdk-21.0.10+7\bin;D:\devfs\apache-maven-3.9.15\bin;D:\devfs\node-v24.15.0-win-x64;.local\bin;D:\devfs\vim92;
 +     
 +##### Set $env
 +
 +     [Environment]::SetEnvironmentVariable("OPENAI_API_KEY", "sk-your-key-here", "User")
 +
 +Then restart your computer (not just VS Code) to ensure it propagates system-wide.
 +
 +Check $env in terminal:
 +  
 +     $env:OPENAI_API_KEY
 +
 +Alternative: Set it in VS Code's terminal settings If you need it available immediately in the integrated terminal regardless, add this to your VS Code settings.json:
 +
 +"terminal.integrated.env.windows": {
 +    "OPENAI_API_KEY": "sk-your-key-here"
 +}
 +

Trace:

Navigation