🎄 Holiday ToDo List 🎄

[mostlygeek]

Here'a a Holiday feature blitz for 2025. These are just early designs. :)

✅ 1. Support for API keys

This has been requested (#50, #251) a few times. It's time to add it. It'll be really helpful to support adding support for remote models.

This is can look in the config:

apiKeys: 
   - "sk-api-key-1"
   - "sk-api-key-2"
   - "sk-api-key-3"
   - "sk-api-key-4"
   - "sk-api-key-5"

It'll be a top level configuration item that is just an array of strings. llama-swap has always been default-allow. This will remain true for backwards compatibility.

When the apiKeys list has elements these http-headers will be searched:

• Authorization: Bearer {apiKey}
• x-api-key: {apiKey} (anthropic support)

If a valid API key is found then:

• the Authorization and x-api-key headers will be removed

If a valid API key is NOT found then:

• send a 401 Unauthorized response

✅ 2. Add support for remote models (#299)

I've seen some settings where people are using workarounds for llama-swap to proxy to another server. Perhaps this can be done with a few small changes to the configuration. For example:

# peers: object defining other providers
# - optional, default: {} (empty)
peers:
   # minimal configuration to proxy request to another host
   peer_1:  
      # anything that supports the APIs supports by llama-swap
      # can be another llama-swap instance
      proxy: http://192.168.1.131:8080
      # models: a list models on the remote host
      # - local models are search first before peer models 
      # - if multiple peers provide a model, they are searched in the order they were defined
      models: 
        - llama-8B
        - qwen-coder-30B
        - another/model_id

   # support non-local remote models
   openrouter:
      proxy: https://openrouter.ai/api
      # apiKey is injected into the request automatically
      apiKey: sk-this-is-injected-into-request
      models: 
        - qwen/qwen3-235b-a22b-2507
        - meta-llama/llama-3.1-8b-instruct

When apiKey is present in the configuration it will be injected into the request. Based on these rules:

• it will set or replace the http headers: Authorization and x-api-key to be compatible with HTTP authorization and Anthropic's proprietary header.
• if no apiKey is set then no key is set on the headers

3. /v1/images - with example using sd-server!

• llama-swap wiki pages: https://github.com/mostlygeek/llama-swap/wiki/use-llama%E2%80%90swap-with-stable%E2%80%90diffusion.cpp
• feat: simple openai image generation api compatiple server leejet/stable-diffusion.cpp#1037

4. Create llama-swap.wiki

• A new documentation based website for llama-swap
• a configuration generator - let's vibe code this

[mostlygeek]

with https://llama-swap.wiki launched, wrapping up the holiday sprint.