lancedb-opencode-pro

LanceDB-backed long-term memory provider for OpenCode

Welcome to lancedb-opencode-pro! This plugin empowers OpenCode with a durable, long-term memory system powered by LanceDB.

To help you find what you need quickly, please select the guide that best fits your needs:

🗺️ Choose Your Path

⚠️ Experiencing Issues?

You see "Memory store unavailable" error or plugin not loading 👉 Read the Compatibility Guide (10 min)

• Fix: rm -rf ~/.cache/opencode/node_modules/ then restart
• OpenCode v1.4.3 fully supported (NAPI issue was a cache problem, not a loader bug)
• Diagnosis checklist and solutions

🚀 First-Time Users

You are new to this project and want to get it running quickly. 👉 Read the Quick Start Guide (15 min)

• Complete installation steps & examples
• Basic memory operations
• Troubleshooting common issues

⚙️ Advanced Users

You have it running and want to tune retrieval, use OpenAI, or share memory across projects. 👉 Read the Advanced Configuration (30 min)

• Hybrid retrieval (RRF, recency boost, importance weighting)
• Memory injection controls (budget/adaptive modes)
• OpenAI Embedding setup
• Cross-project memory sharing (global scope)

🛠️ Developers & Contributors

You want to understand the architecture, run tests, or contribute to the source code. 👉 Read the Development Workflow (20 min)

• Local environment setup
• OpenCode skills usage guide
• Testing and validation processes
• Release workflow

🎯 Quick Start (5 Minutes)

1. Registration

Register the plugin in ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": [
    "lancedb-opencode-pro"
  ]
}

2. Configuration

Create ~/.config/opencode/lancedb-opencode-pro.json:

{
  "provider": "lancedb-opencode-pro",
  "dbPath": "~/.opencode/memory/lancedb",
  "embedding": {
    "provider": "ollama",
    "model": "nomic-embed-text",
    "baseUrl": "http://127.0.0.1:11434"
  }
}

3. Start Ollama and Restart OpenCode

# Verify Ollama is accessible
curl http://127.0.0.1:11434/api/tags

# Restart OpenCode

That's it! OpenCode will now automatically capture key decisions and inject them into future conversations.

✨ Core Features

Auto Memory Capture

• Automatically extracts decisions, lessons, and patterns from assistant responses.
• Configurable minimum capture length (default: 80 chars).
• Auto-categorization into project or global scope.

Hybrid Retrieval (v0.1.4+)

• Vector Search + BM25 Lexical Search
• Reciprocal Rank Fusion (RRF)
• Recency boost & Importance weighting

Memory Management Tools

| Tool | Description | Documentation | |------|-------------|---------------| | memory_search | Hybrid search for long-term memory | Doc | | memory_delete | Delete a single memory entry | Doc | | memory_clear | Clear all memories in a specific scope | Doc | | memory_stats | View memory statistics | Doc | | memory_remember | Manually store a memory | Doc | | memory_forget | Remove or disable a memory | Doc | | memory_what_did_you_learn | Show recent learning summaries | Doc | | memory_dashboard | Weekly learning dashboard with trends | Doc | | memory_kpi | Learning effectiveness KPIs (retry-to-success, memory lift) | Doc |

(For full details on tools like Effectiveness Feedback, Cross-Project Sharing, Deduplication, Citations, and Episodic Learning, please refer to the Advanced Configuration.)

⚙️ Configuration Overview

Basic Setup

{
  "provider": "lancedb-opencode-pro",
  "dbPath": "~/.opencode/memory/lancedb",
  "embedding": {
    "provider": "ollama",
    "model": "nomic-embed-text",
    "baseUrl": "http://127.0.0.1:11434"
  },
  "retrieval": {
    "mode": "hybrid",
    "vectorWeight": 0.7,
    "bm25Weight": 0.3,
    "minScore": 0.2
  }
}

Full configuration options including injection control, deduplication, and OpenAI setup are available in docs/ADVANCED_CONFIG.md.

Priority Order

1. Environment Variables (LANCEDB_OPENCODE_PRO_*)
2. LANCEDB_OPENCODE_PRO_CONFIG_PATH
3. Project sidecar: .opencode/lancedb-opencode-pro.json
4. Global sidecar: ~/.config/opencode/lancedb-opencode-pro.json
5. Legacy sidecar or built-in defaults

🧪 Validation & Testing

Quick validation using Docker:

docker compose build --no-cache && docker compose up -d
docker compose exec opencode-dev npm run verify

Full pre-release validation:

docker compose exec opencode-dev npm run verify:full

Performance benchmark (optional):

# Mock mode (fast)
docker compose exec opencode-dev ./scripts/run-perf-benchmark.sh

# Real Ollama mode
docker compose exec opencode-dev ./scripts/run-perf-benchmark.sh --real

Check docs/memory-validation-checklist.md for more details.

📦 Installation Options

Primary method: npm package (recommended) Alternatively, install via .tgz release asset or build from source. See Install Options.

🗺️ Version History

• v0.8.3: @opencode-ai/sdk and plugin upgrade to 1.4.7, Effect type mock fix, SDK compatibility docs
• v0.8.2: Plugin version mismatch fix (internal)
• v0.8.1: @opencode-ai/sdk and plugin upgrade to 1.4.3, bunfig.toml test isolation fix
• v0.8.0: Structured Logging via client.app.log() per OpenCode best practices, Test Environment Isolation Fix
• v0.7.0: OpenCode SDK v1.3.14 Compatibility, Node 22 memory_search Race Condition Fix
• v0.6.3: Index Creation Guard (defer on empty/insufficient tables, fix #70), LanceDB 0.27.2
• v0.6.2: Index Race Condition Fix (concurrent-process conflict handling, jitter backoff)
• v0.6.1: Event TTL/Archival, Index Creation Resilience, Duplicate Consolidation Performance
• v0.6.0: Learning Dashboard, KPI Pipeline, Feedback-Driven Ranking, Task-Type Aware Injection

[older versions removed - see CHANGELOG.md for full history]

See CHANGELOG.md for all changes.

🤝 Contributing

1. Read docs/DEVELOPMENT_WORKFLOW.md
2. Understand OpenSpec specs in openspec/specs/
3. Use backlog-to-openspec to create proposals

📞 Support & License

• Issues: Submit errors or requests on GitHub Issues.
• License: MIT License - see LICENSE.

Last Updated: 2026-04-17 Latest Version: v0.8.3