Veda — FreeLLM Router

Only the models that work.

Veda (वेद = knowledge) makes your FreeLLM API actually usable.

The Problem

FreeLLM shows all models from your API keys and says you have 1.6B tokens or something. In reality half are dead — Google Flash, DeepSeek etc show Payment Required or No longer available. You literally can't tell which ones work without trying each one.

Veda fixes that. It tests every model and keeps only the ones that really work.

What Veda Does

• Checks all models on first start and saves only working ones
• Shares work fairly — each request uses a different model, so no single model runs out
• If one fails, tries the next instantly — you still get an answer
• Dashboard to see everything live
• Refresh button — add a new API key in FreeLLM, hit Refresh, new working models are added automatically

Setup (2 minutes)

1. Install:

pip install -r requirements.txt

2. Create config:

copy .env.example .env

Open .env and add your FreeLLM details:

FREELLM_API_BASE=https://api.freellm.ai
FREELLM_API_KEY=your-key-here

3. Start:

python veda.py

Open http://localhost:3000 in your browser. Done.

Using the Dashboard

Open http://localhost:3000

• Top cards — how many models are working, how many requests, speed
• Model Pool — list of working models. Green ON = ready, Red COOLDOWN = resting after an error (comes back automatically)
• Live log — shows which model was used for each request
• Quick test — type a message, click Send to test instantly
• Refresh models — top-right button. Use it after you add a new API key in FreeLLM. It fetches all models again, tests them, and updates the list.

Use with Cline or any other extension (VS Code)

1. Start Veda (python veda.py)
2. In VS Code, open Cline settings
3. Set Base URL to http://localhost:3000/v1
4. Set API Key to the same key from your .env
5. Use Cline normally — Veda will handle model switching for you.

Files

• veda.py — the app
• preferred_models.json — your working models (created automatically, don't edit)
• .env — your keys
• static/index.html — dashboard

Need Help?

• If a model shows COOLDOWN, it will recover on its own in ~15-60 seconds
• If you added a new key and don't see new models, hit Refresh models
• If nothing works, check your .env keys and restart python veda.py