drjones/nexus-mcp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fully migrate off Gemini: shared Ollama client (gemma4:26b @ shadow-death), health reports llm backend, drop @google/genai

Browse Source
This commit is contained in:
drjones
2026-08-14 00:23:03 +00:00
parent d1210773e7
commit 121570fb67
10 changed files with 80 additions and 478 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
README.md
View File

@@ -40,11 +40,18 @@ npm run dev # tsx server.ts with Vite HMR
## Configuration
Set `GEMINI_API_KEY` in `.env.local` to enable the AI prompt-to-tool generator. All other features run without any external key.
AI features (tool generation, web-search synthesis, text summarization) run on a **local Ollama model** — `gemma4:26b` on `shadow-death` (`http://10.30.20.128:11434`) by default. No cloud API key required.
Override via environment variables:
| Variable | Default |
|----------|---------|
| `OLLAMA_HOST` | `http://10.30.20.128:11434` |
| `OLLAMA_MODEL` | `gemma4:26b` |
## Stack
React 19 · Vite 6 · Express 4 · TypeScript · Tailwind CSS 4 · Recharts · `@google/genai`
React 19 · Vite 6 · Express 4 · TypeScript · Tailwind CSS 4 · Recharts · Ollama (local LLM)
---