ryupold/agentification
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
messing around with multiple (local) LLMs using lm studio sdk
94 commits 2 branches 0 tags 1.8 MiB
Find a file
Exact
Michael Scherbakow 43cbf9de09 Improves agent coding capabilities 
- Updates agent model to 'google/gemma-3-12b' for coding tasks, increasing token limit.
- Integrates new file and web tools for enhanced coding capabilities.
- Streamlines agent interaction with filesystem by including common file operations like reading, writing, deleting, and existence checks as tools.
- Adds web interaction capabilities to agents with a new tool.
- Modifies visual snapshot specification for the visual worker
2026-02-13 17:30:34 +01:00
act Improves agent coding capabilities 2026-02-13 17:30:34 +01:00
chat Adds agent message chain loop 2026-02-01 19:00:05 +01:00
coding Improves agent coding capabilities 2026-02-13 17:30:34 +01:00
common Improves agent coding capabilities 2026-02-13 17:30:34 +01:00
ssr Improves agent coding capabilities 2026-02-13 17:30:34 +01:00
.gitignore Adds visual inspection of the web app to the agent 2026-01-30 20:48:51 +01:00
bun.lock introduce ssr server to allow observing the rendered content 2026-01-23 18:56:31 +01:00
esbuild.config.ts setup project 2026-01-17 12:48:13 +01:00
package.json introduce ssr server to allow observing the rendered content 2026-01-23 18:56:31 +01:00
README.md Adds arena combat simulation between agents 2026-01-29 19:01:50 +01:00

README.md

agentification

playing around with multiple LLMs

  1. Download and install LM Studio from https://lmstudio.ai/
  2. Make sure you have Bun installed from https://bun.sh/
  3. Run: 
    bun i

# get models you want to use
lms get openai/gpt-oss-20b
lms get mistralai/mistral-nemo-instruct-2407
lms get liquid/lfm2-1.2b
...

Run examples:

1. turn based conversation between multiple agents

bun chat/agents-just-talk-loop.ts

2. turn based arena battle between multiple agents

bun act/agents-arena-loop.ts

3. Simulate a standup meeting between multiple agents then work separately on tasks

bun coding/agents-agile-loop.ts

#...in a seperate shell start the bundler for watching the SPA
cd coding/agile
bun run watch

#...in a seperate shell start serving the SPA
cd coding/agile
bun run serve

#...in a seperate shell start the SSR server
bun ssr/ssr-server.ts

4. One agent building a vanilla JS SPA

bun coding/agents-vanilla-loop.ts

#...in a seperate shell start serving the SPA
cd coding/vanilla
bun run serve

#...in a seperate shell start the SSR server
bun ssr/ssr-server.ts

5. One agent building a JS SPA with just visual feedback

bun coding/agents-visual-worker-loop.ts

#...in a seperate shell start serving the SPA
cd coding/visual-worker
bun run serve

#...in a seperate shell start the SSR server
bun ssr/ssr-server.ts