dotfiles
User-specific application configuration is traditionally stored in so called dotfiles, these are my own.
Star
Fork
Watch

Commits on main
[pi] Default to deepseek v4.1 flash, cleanup models
committed
[bin/gufo-dev] Allocate the TTY from the descriptors
committed
[dlmodel] Restart a part whose sidecar cannot be read
committed
[pi] Merge auth.json so pi writes its own credentials too
committed
[mise] Edit ~/.ssh/config with a block so unapply works
committed
[pi] Add the Grafana Cloud MCP server
committed
[mise] Use absolute dotfile sources
committed
[mise] Drop desktop host build packages
committed
[mise] Remove retired llama-server and its references
committed
[pi] Manage settings.json with merge edits instead of copy
committed
[pi] Keep the starting model list to models the plans actually serve
committed
[mise] Move go to base mise.toml
committed
[pi] Install pi-pkg from npm instead of git
committed
[pi] Remove expired qwen-token-plan provider
committed
[git] Sign tags by default
committed
[mise] Add the tg Tangled CLI
committed
[pi] Only configure deepseek-flash through the DeepSeek API
committed
[pi] Configure deepseek api key and models
committed
[pi] Trim the Go list to the ladder Step 5's free preview makes honest
committed
[mise] Use version file for rust
committed
[pi] Rebuild the Linux model list around the Go value frontier
committed
[nvim] Enable rust_analyzer
committed
[mise] Install glab on Linux for the gitlab pi skill
committed
[pi] Read Pi api keys from the Developer 1Password vault
committed
[pi] Restrict enabled models to the no-train test tier
committed
[bin/gufo-dev] Fix no-op guards and Strix Halo detection
committed
[bin/gufo-dev] Add helper script to work on gufo inference server
committed
[pi] Use native context window for gufo/qwen3.8-flash
committed
[jj] Add jj-vcs to test drive it
committed
[pi] Fix defaultModel on macOS and add typesafe api key on linux
committed
[AGENTS,pi] Empty the shared AGENTS.md; commit style moves to a skill
committed
[pi] Shrink gufo context and maxTokens config
committed
[pi] 1.0 setting updates
committed
[pi] Gate gufo thinking levels to the four it does not collapse
committed
[pi] Add gufo provider over its OpenAI chat endpoint
committed
[pi] Install pi-review extension
committed
[fish,pi] Re-add mcp-adapter on macOS, and disable MCP ui
committed
[pi] Gate opencode-go qwen3.8-flash thinking levels to off/low/medium/xhigh
committed
[pi] Drop opencode-go /v1 reroute; use stock Anthropic compat + allowEmptySignature
committed
[llama.cpp,pi] Fully drop llama.cpp and ds4
committed
[pi] Trim opencode-go qwen compat to fields doing real work
committed
[pi] Replace os() templates with per-platform copied configs
committed
[pi] Drop ds4 and llama.cpp configs
committed
[pi] Add Qwen plan, patch opencode-go's qwen3.8-flash model
committed
[pi] Route qwen3.8-flash over opencode's OpenAI endpoint
committed
[pi] Drop pi-mcp-adapter for smaller pi-pkg extension
committed
[mise] Drop Fedora ROCm RPMs now installed by amferd
committed
[mise] Provide rocm and mise tools to non-interactive bash
committed
[fish] pi-kagi now uses the OS keyring, rather than env vars
committed
[nvim] Format shell files with shfmt
committed
[nvim] Restore blink.cmp LSP capabilities and YAML schema store
committed
[nvim] Replace hand-rolled Helm setup with helm-ls.nvim
committed
[nvim] Fold lsp/ server configs into init.lua
committed
[nvim] Strip kickstart boilerplate; manage tools via mise and vim.pack
committed
[nvim] Base configuration on kickstart.nvim
committed
[pi] Setup ds4 with vision supported models
committed
[pi] Document ds4 provider config and thinking-level rationale
committed
[bin] Symlink all contents from bin to .local/bin
committed
[llama.cpp] Drop qwen3.8-27b preset; unreliable at tool calls and file inspection
committed
[llama.cpp] Document flash-next MTP depth-1 bug fix; add dlmodel.mjs helper
committed