Remove opencode-rag-plugin: hangs startup without Ollama embedding backend
The RAG plugin probes a local embedding backend (Ollama on localhost:11434 by default) at init with no timeout. When Ollama isn't running, opencode hangs after loading plugins and never starts. Dropped it from the template and documented re-enable steps + the --pure diagnostic in the README. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
@@ -66,7 +66,7 @@ echo 'export BRAVE_API_KEY=BSA...' >> ~/.bashrc
|
||||
|
||||
## What's Synced
|
||||
|
||||
- **9 Plugins** - oh-my-openagent, opencode-mem, opencode-pty, dynamic context pruning, RAG, command-inject, notify, agent-identity, tokenscope
|
||||
- **8 Plugins** - oh-my-openagent, opencode-mem, opencode-pty, dynamic context pruning, command-inject, notify, agent-identity, tokenscope
|
||||
- **32 Skills** - Coding standards, frontend/backend patterns, security review, etc.
|
||||
- **3 Agents** - explorer, reviewer, docs-researcher
|
||||
- **14 MCP Servers** - codegraph, brave-search, playwright, github, context7, exa, etc.
|
||||
@@ -100,6 +100,15 @@ git pull
|
||||
**Git MCP fails?**
|
||||
- Install Python package: `pip install mcp-server-git`
|
||||
|
||||
**opencode hangs on startup (no output)?**
|
||||
- The `opencode-rag-plugin` was removed because its startup probes a local
|
||||
embedding backend (Ollama on `localhost:11434` by default) with no timeout —
|
||||
if Ollama isn't running, opencode hangs after loading plugins.
|
||||
- To re-enable it, first start a backend (`ollama serve` + `ollama pull embeddinggemma`,
|
||||
or configure an OpenAI/Cohere embedding key), then add `"opencode-rag-plugin"`
|
||||
back to the `plugin` array in `opencode.json.template`.
|
||||
- Diagnose plugin hangs with `opencode run --pure "ok"` (runs without external plugins).
|
||||
|
||||
## Structure
|
||||
|
||||
```
|
||||
|
||||
Reference in New Issue
Block a user