Commits

Branch selector

main

User selector

All users

All time

Commit history

Commits on Jul 17, 2026

• docs: fix google-ads-mcp-server links to correct google-ads-mcp repo

  Show description for 2066a01
  

  itallstartedwithaidea

  and

  cursoragent

  committed
  2066a01View commit details

  Copy full SHA for 2066a01

  Browse repository at this point

Commits on Apr 12, 2026

• feat: Add multi-language READMEs (FR/ES/ZH/NL/RU/KO)

  Show description for b5cdbe8
  

  John Williams

  committed
  b5cdbe8View commit details

  Copy full SHA for b5cdbe8

  Browse repository at this point

Commits on Mar 31, 2026

• fix: rename mcp/ to mcp_servers/ to resolve namespace collision (#3)

  Show description for ff170d2
  

  itallstartedwithaidea

  and

  John Williams

  authored
  ff170d2View commit details

  Copy full SHA for ff170d2

  Browse repository at this point

Commits on Mar 23, 2026

• Add MiniAgent-494M (Qwen2.5 multilingual base), update README with all 3 models

  Show description for ad414f9
  

  John Williams

  committed
  ad414f9View commit details

  Copy full SHA for ad414f9

  Browse repository at this point

• Add continuous learning, RAG, blog training data, Qwen2.5-0.5B support

  Show description for aa9331b
  

  John Williams

  committed
  aa9331bView commit details

  Copy full SHA for aa9331b

  Browse repository at this point

• Update README with real trained models, training results, and honest quality assessment

  Show description for e4ac279
  

  John Williams

  committed
  e4ac279View commit details

  Copy full SHA for e4ac279

  Browse repository at this point

• Fix combined mode OOM: reduce batch_size to 8 and max_length to 320 for 104M model on 20GB

  Show description for 2f1198c
  

  John Williams

  committed
  2f1198cView commit details

  Copy full SHA for 2f1198c

  Browse repository at this point

• Add combined training mode: minimind base + advertising pretrain + SFT

  Show description for 1ca8180
  

  John Williams

  committed
  1ca8180View commit details

  Copy full SHA for 1ca8180

  Browse repository at this point

• Add practitioner expertise from skills, hub, and agent specs to training data

  Show description for 110e360
  

  John Williams

  committed
  110e360View commit details

  Copy full SHA for 110e360

  Browse repository at this point

• Add 60+ platform training data: Search, Social, DSP, Retail, CTV, Mobile, Native, Analytics

  Show description for fcf6a39
  

  John Williams

  committed
  fcf6a39View commit details

  Copy full SHA for fcf6a39

  Browse repository at this point

• Add GGUF converter, update README with HuggingFace model link

  Show description for a20c197
  

  John Williams

  committed
  a20c197View commit details

  Copy full SHA for a20c197

  Browse repository at this point

• Fix SFT OOM: reduce batch_size to 8 and max_length to 512 for 20GB GPUs

  Show description for d8c4476
  

  John Williams

  committed
  d8c4476View commit details

  Copy full SHA for d8c4476

  Browse repository at this point

• Move import math to top of pretrain.py (was only inside main())

  Show description for e6f978b
  

  John Williams

  committed
  e6f978bView commit details

  Copy full SHA for e6f978b

  Browse repository at this point

• Rewrite GQAttention.forward with consistent dimension ordering

  Show description for 9e2b0c3
  

  John Williams

  committed
  9e2b0c3View commit details

  Copy full SHA for 9e2b0c3

  Browse repository at this point

• Fix RoPE apply_rope: index by seq_len (dim 2) not n_heads (dim 1)

  Show description for 098b876
  

  John Williams

  committed
  098b876View commit details

  Copy full SHA for 098b876

  Browse repository at this point

• Fix device detection: send GPU info to stderr, only return device string

  Show description for d5b5c1a
  

  John Williams

  committed
  d5b5c1aView commit details

  Copy full SHA for d5b5c1a

  Browse repository at this point

• Fix training pipeline: real tokenizer, expanded data, one-command GPU training

  Show description for 875fb21
  

  John Williams

  committed
  875fb21View commit details

  Copy full SHA for 875fb21

  Browse repository at this point

• 🎯 Enhancement Plan: pushy skill descriptions, enforcement skill, workflow composition, eval framework

  Show description for 531cba0
  

  John Williams

  committed
  531cba0View commit details

  Copy full SHA for 531cba0

  Browse repository at this point

• 🚀 v0.2.0 — Complete repo: 14 MCP servers, hub connectors, tests, diagrams, templates, Makefile, examples

  Show description for 99782a5
  

  John Williams

  committed
  99782a5View commit details

  Copy full SHA for 99782a5

  Browse repository at this point

Commits on Mar 22, 2026

• 🔧 v0.1.1 — Add web demo, data downloader, DPO/LoRA stubs, CLI, fix missing files

  

  John Williams

  committed
  eebfcfaView commit details

  Copy full SHA for eebfcfa

  Browse repository at this point

• 🚀 MiniAgent v0.1.0

  

  John Williams

  committed
  ed73a26View commit details

  Copy full SHA for ed73a26

  Browse repository at this point