NineRouter Picker

Group models by vendor, auto-pick latest version, filter junk — for anyone using 9Router or a multi-provider OpenAI-compatible API.

Problem

9Router /v1/models returns all models from all providers in one flat list. For example, 41 models from BAi + Nvidia + Alibaba etc — mixed up, hard to choose, can't tell which vendor each is from, which version is the latest, or which are just free trials.

Solution

This picker:

• Group by vendor — BAi, Nvidia, etc. are automatically separated by model id prefix
• Auto-pick — just say pick("glm") → instantly get bai/glm-5.2 (latest version)
• Skip free/junk — models with "free" substring are auto-filtered
• Registered filter — only show models you have access to (manual curation)
• Version-aware sort — gpt-5.6 > gpt-5.5 > gpt-5.4 > gpt-5-mini
• 5-minute cache — doesn't spam 9Router on every render

Usage

1. Have your own 9Router

git clone https://github.com/Mftrferdinand/NineRouter-Picker
cd ninerouter-picker
python3 -m venv venv
./venv/bin/pip install -r requirements.txt
cp .env.example .env
# Edit .env → fill in YOUR 9Router URL & key
./venv/bin/python picker.py list         # view your models
./venv/bin/python picker.py pick glm     # auto-pick latest glm

2. Don't have 9Router — but have an OpenAI-compatible API (Groq, OpenRouter, etc.)

Edit .env:

NINEROUTER_URL=https://api.groq.com/openai/v1
NINEROUTER_KEY=gsk_YOUR_KEY_HERE

Picker automatically uses that endpoint.

CLI

./venv/bin/python picker.py list              # picker (filter registered)
./venv/bin/python picker.py list-all          # all models
./venv/bin/python picker.py show              # current model
./venv/bin/python picker.py test              # test current (ping)
./venv/bin/python picker.py register bai/glm-5.2
./venv/bin/python picker.py unregister bai/glm-5.2
./venv/bin/python picker.py use bai/glm-5.2   # set current + test
./venv/bin/python picker.py pick glm          # auto-pick by preference
./venv/bin/python picker.py pick              # auto-pick (priority default)
./venv/bin/python picker.py refresh           # bust cache + render

Python API

from picker import pick, list_registered, current, set_current, render, add_registered

# Auto-pick
pick()              # → first match from PICK_PRIORITY (e.g. 'bai/minimax-m3')
pick("glm")         # → latest glm (e.g. 'bai/glm-5.2')
pick("kimi")        # → latest kimi
pick("gpt")         # → latest gpt
pick("claude")      # → latest claude

# Curation
list_registered()   # → ['bai/glm-5.2', 'bai/minimax-m3', ...]
add_registered('bai/qwen3.6-27b')
set_current('bai/glm-5.2')
current()           # → 'bai/minimax-m3' or None

# Render markdown
print(render("model"))    # grouped, filter registered
print(render("register")) # grouped, all models
print(render("full"))     # current + registered + grouped

Auto-Pick Logic

1. Pool = list_registered() if available, else all non-free models.
2. If preference is passed:
   • exact match (full id or tail) → return immediately
   • substring match → sort by version DESC (latest wins)
3. Else: rank by PICK_PRIORITY list (exact tail > prefix > substring), return top.

Version sort: flatten all numeric runs in the tail, padded to 6 segments.

gpt-5.6-sol > gpt-5.5 > gpt-5.4-pro > gpt-5.4 > gpt-5-mini

Customize

Edit picker.py:

# Label vendor (default: capitalize prefix)
VENDOR_LABELS = {"bai": "BAi", "nvidia": "Nvidia", "alibaba": "Aliyun"}

# Label model (default: tail of id)
MODEL_LABELS = {"bai/glm-5.2": "Glm-5.2 (Premium)"}

# Priority fallback (default + curation)
PICK_PRIORITY = ["minimax-m3", "glm-5.2", "kimi-k2.6", ...]

Security

• .env (contains YOUR real URL & key) is git-ignored — never pushed
• .env.example on GitHub is just a dummy template (CHANGE_ME)
• Each user uses THEIR OWN 9Router/endpoint — no key sharing
• localhost = only your device can access

File Structure

ninerouter-picker/
├── picker.py              # core lib + CLI (single file)
├── requirements.txt       # httpx only
├── .env.example           # template (don't fill with real key)
├── .env                   # your real key (gitignored!)
├── .gitignore
├── registered.json        # models you selected (auto-generated)
├── current.json           # active model (auto-generated)
└── README.md

License

MIT