Clojurians Log v2

I'm not sure there is an easy way, but one way could be to extend the reader for example, like :

(ns my-reader-tags
   (:require [clojure.string :as str]))

(defn read-multi-line-string-tag [multi-line-str-vec]
  (str/join "\n" multi-line-str-vec))

{multi-line-str my-reader-tags/read-multi-line-string-tag}

(def ml
  #multi-line-str
  ["This is a"
   "multi line string"])

user> ml
"This is a\nmulti line string"

@zikajk true, done

@zikajk btw I've now added deepseek support

oh nice! Curious about your todo: Ability to let GPT decide which context files it needs!! Also I may be misunderstanding, but adding project files to the context will not risk to fill the context window? Doesn't that makes the LLM become less precise as well (afaik beginning and end of context are used better than what is in the middle)? How does it work in your usage?

Currently the context window for claude is about https://support.anthropic.com/en/articles/7996856-what-is-the-maximum-prompt-length pages, for chatgpt it's a bit more than https://platform.openai.com/docs/models/gp. So far I haven't had to use that many project files to provide enough context to get meaningful responses. I've been using it to build a project with #biff, and it's working incredibly well. I generally provide the database schema as context and then either some example files (e.g., the source code file if I want to generate a test for it or some file highlighting some similar functionality). It's also a blaze for writing UI components with hiccup. The ability to let gpt decide which context files it needs would just speed up the process. The idea would be to first send a request with all available context files and the query it needs to answer and to ask it which context files it needs to provide the best possible answer.

would it be possible to add support for other providers like gpt-el offers? I personally use DeepSeek and would like to see if this lib would suit me better than other similar tools (Aider, Elysium).

Yeah sure. Can you open an issue and I’ll look into adding it?

@anselm.coogan I think you have to enable Issues