anomalyco/opencode
1
0
Fork 0
mirror of https://github.com/anomalyco/opencode.git synced 2026-07-22 10:15:31 -04:00
Code Issues 5.9k Packages Projects Releases 19 Wiki Activity

[FEATURE]: Advanced Dual-Scope Memory System for OpenCode #6674

New Issue
Closed
opened 2026-02-16 18:04:55 -05:00 by yindo · 1 comment
yindo commented 2026-02-16 18:04:55 -05:00
Owner
Copy Link
Copy Source

Originally created by @furqanahmadrao on GitHub (Jan 18, 2026).

Originally assigned to: @thdxr on GitHub.

Feature hasn't been suggested before.

  • I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

📋 Summary

I have developed and tested a robust Memory System for OpenCode that solves the issues of context pollution, manual file management, and unsafe edits. This system introduces a Dual-Scope Architecture (Global vs. Project), ID-Based Management, and Automatic Templating.

I propose integrating this system (or similar logic) into OpenCode to transform it from a stateless CLI into a persistent, context-aware Engineer.

🛑 The Problem

Currently, managing long-term context in OpenCode is manual and fragile:

  1. Global Pollution: Saving project-specific facts (e.g., "This app uses Vite") to the global AGENTS.md clutters the agent's brain for all future sessions.
  2. Unsafe Edits: Asking the AI to "remove the memory about testing" often leads to fuzzy text matching that deletes the wrong lines.
  3. Context Amnesia: Starting a new project requires manually teaching the agent the architecture every time, or manually creating a ./AGENTS.md file.

🛠️ The Solution: Dual-Scope Memory Engine

I have built a custom toolset (memory.ts) and Agent Protocol that implements the following architecture:

1. Dual-Scope Context

The system distinguishes between User Wisdom and Project Facts.

Scope Location Purpose Example
Global ~/.config/.../AGENTS.md User preferences, Life lessons "Always use Dark Mode", "I prefer concise code"
Project ./AGENTS.md Architecture, Tech Stack, patterns "Auth is handled by Clerk", "API is at /v1"

Why this matters:

  • Switching projects gives the Agent a "Fresh Brain" for code, but keeps the "Personality" of the user.
  • No more "hallucinating" React code while working in a Vue project.

2. ID-Based Management (CRUD)

Instead of fuzzy text search, every memory is assigned a deterministic short-hash ID (e.g., a1b2).

  • Read: Returns [ID:a1b2] - Content...
  • Update: memory_update(id='a1b2', content='...') → Atomic, safe updates.
  • Remove: memory_remove(target='a1b2') → 100% precision deletion.

3. Smart Auto-Templating

If a user asks to save a project fact ("Remember this stack") and no ./AGENTS.md exists, the tool automatically initializes the file with a professional template:

# 📂 Project Context: [Name]
## 🏗️ Architecture
## 📝 Code Standards
## 🧠 Project Context System (Auto-managed)

4. Protected "Safe Zones"

The tool respects a specific block in the markdown file:

<!-- MANAGED BY TOOLS - DO NOT EDIT MANUALLY -->
   [... Memory Tool only touches lines here ...]
<!-- END MEMORY SYSTEM -->

This ensures the tool never overwrites the user's manual instructions or custom rules outside this block.

🧠 The Workflow

  1. User: "Remember this is a Next.js app."
  2. Agent: Detects this is a code fact → Selects scope='project'.
  3. System: Checks for ./AGENTS.md. If missing, creates it with the Template. Saves memory.
  4. Result: Next session, the Agent reads ./AGENTS.md and instantly knows the stack.

🎯 Impact

This transforms OpenCode from a "Smart CLI" into a "Senior Partner" that:

  1. Remembers your preferences everywhere (Global).
  2. Understands your specific project deeply (Project).
  3. Manages its own knowledge base safely (ID/Protected Blocks).

I have been using this system successfully to manage complex refactors and architecture planning. I believe it would be a valuable standard feature or official plugin for the community.

Originally created by @furqanahmadrao on GitHub (Jan 18, 2026). Originally assigned to: @thdxr on GitHub. ### Feature hasn't been suggested before. - [x] I have verified this feature I'm about to request hasn't been suggested before. ### Describe the enhancement you want to request ## 📋 Summary I have developed and tested a robust **Memory System** for OpenCode that solves the issues of context pollution, manual file management, and unsafe edits. This system introduces a **Dual-Scope Architecture** (Global vs. Project), **ID-Based Management**, and **Automatic Templating**. I propose integrating this system (or similar logic) into OpenCode to transform it from a stateless CLI into a persistent, context-aware Engineer. --- ## 🛑 The Problem Currently, managing long-term context in OpenCode is manual and fragile: 1. **Global Pollution**: Saving project-specific facts (e.g., "This app uses Vite") to the global `AGENTS.md` clutters the agent's brain for *all* future sessions. 2. **Unsafe Edits**: Asking the AI to "remove the memory about testing" often leads to fuzzy text matching that deletes the wrong lines. 3. **Context Amnesia**: Starting a new project requires manually teaching the agent the architecture every time, or manually creating a `./AGENTS.md` file. --- ## 🛠️ The Solution: Dual-Scope Memory Engine I have built a custom toolset (`memory.ts`) and Agent Protocol that implements the following architecture: ### 1. Dual-Scope Context The system distinguishes between **User Wisdom** and **Project Facts**. | Scope | Location | Purpose | Example | |-------|----------|---------|---------| | **Global** | `~/.config/.../AGENTS.md` | User preferences, Life lessons | "Always use Dark Mode", "I prefer concise code" | | **Project** | `./AGENTS.md` | Architecture, Tech Stack, patterns | "Auth is handled by Clerk", "API is at /v1" | **Why this matters:** - Switching projects gives the Agent a "Fresh Brain" for code, but keeps the "Personality" of the user. - No more "hallucinating" React code while working in a Vue project. ### 2. ID-Based Management (CRUD) Instead of fuzzy text search, every memory is assigned a deterministic short-hash ID (e.g., `a1b2`). - **Read**: Returns `[ID:a1b2] - Content...` - **Update**: `memory_update(id='a1b2', content='...')` → Atomic, safe updates. - **Remove**: `memory_remove(target='a1b2')` → 100% precision deletion. ### 3. Smart Auto-Templating If a user asks to save a project fact ("Remember this stack") and no `./AGENTS.md` exists, the tool **automatically initializes** the file with a professional template: ```markdown # 📂 Project Context: [Name] ## 🏗️ Architecture ## 📝 Code Standards ## 🧠 Project Context System (Auto-managed) ``` ### 4. Protected "Safe Zones" The tool respects a specific block in the markdown file: ```html <!-- MANAGED BY TOOLS - DO NOT EDIT MANUALLY --> [... Memory Tool only touches lines here ...] <!-- END MEMORY SYSTEM --> ``` This ensures the tool **never** overwrites the user's manual instructions or custom rules outside this block. --- ## 🧠 The Workflow 1. **User**: "Remember this is a Next.js app." 2. **Agent**: Detects this is a *code fact* → Selects `scope='project'`. 3. **System**: Checks for `./AGENTS.md`. If missing, creates it with the Template. Saves memory. 4. **Result**: Next session, the Agent reads `./AGENTS.md` and instantly knows the stack. --- ## 🎯 Impact This transforms OpenCode from a "Smart CLI" into a **"Senior Partner"** that: 1. Remembers your preferences everywhere (Global). 2. Understands your specific project deeply (Project). 3. Manages its own knowledge base safely (ID/Protected Blocks). I have been using this system successfully to manage complex refactors and architecture planning. I believe it would be a valuable standard feature or official plugin for the community.
yindo added the discussion label 2026-02-16 18:04:55 -05:00
yindo closed this issue 2026-02-16 18:04:55 -05:00
yindo commented 2026-02-16 18:04:55 -05:00
Author
Owner
Copy Link
Copy Source

@github-actions[bot] commented on GitHub (Jan 18, 2026):

This issue might be a duplicate of existing issues. Please check:

  • #8948: Server Registry & Auto-Discovery for Persistent Sessions - addresses persistent agent state across sessions
  • #8932: Allow pinning important instructions to keep them after compaction - similar context preservation concept
  • #8919: Adopt .agents as default project directory - related to standardized project-scoped agent configuration
  • #8751: Hot-reload agents, skills and commands - addresses dynamic knowledge management
  • #9062: Support config.d/ directory for modular configuration - addresses project-level configuration management
  • #8533: Allow custom skills directories - related to project-scoped skill management

Feel free to ignore if none of these address your specific case.

@github-actions[bot] commented on GitHub (Jan 18, 2026): This issue might be a duplicate of existing issues. Please check: - #8948: Server Registry & Auto-Discovery for Persistent Sessions - addresses persistent agent state across sessions - #8932: Allow pinning important instructions to keep them after compaction - similar context preservation concept - #8919: Adopt .agents as default project directory - related to standardized project-scoped agent configuration - #8751: Hot-reload agents, skills and commands - addresses dynamic knowledge management - #9062: Support config.d/ directory for modular configuration - addresses project-level configuration management - #8533: Allow custom skills directories - related to project-scoped skill management Feel free to ignore if none of these address your specific case.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
acp
bad
bug
discussion
docs
documentation
enhancement
extension
good first issue
growth
help-wanted
mcp
model-problem
needs:compliance
nix
opentui
perf
pull-request
Mirrored from GitHub Pull Request
 tui
web
windows
zen
No Label discussion
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#6674
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?